Authors: Lorenzo Porcelli
Abstract: Instant payment infrastructures have stringent performance requirements, processing millions of transactions daily with zero-downtime expectations. Traditional monitoring approaches fail to bridge the gap between technical infrastructure metrics and business process visibility. We introduce a novel feature engineering approach based on processing times computed between consecutive ISO 20022 message exchanges, creating a compact representation of system state. By applying anomaly detection to these features, we enable early failure detection and localization, allowing incident classification. Experimental evaluation on the TARGET Instant Payment Settlement (TIPS) system, using both real-world incidents and controlled simulations, demonstrates the approach's effectiveness in detecting diverse anomaly patterns and provides inherently interpretable explanations that enable operators to understand the business impact. By mapping features to distinct processing phases, the resulting framework differentiates between internal and external payment system issues, significantly reduces investigation time, and bridges observability gaps in distributed systems where transaction state is fragmented across multiple entities.
Authors: Jinwoo Baek
Abstract: Transformers trained in low precision can suffer forward-error amplification. We give a first-order, module-wise theory that predicts when and where errors grow. For self-attention we derive a per-layer bound that factorizes into three interpretable diagnostics: a score-scale ratio $\kappa_{\rm score}$, a rowwise softmax sensitivity $\kappa_{\rm softmax}$, and value conditioning $\kappa(V)$. We prove a residual relaxation inequality showing that residual blocks attenuate depth-wise accumulation, and we introduce a precision- and width-aware LayerNorm indicator $\rho_{\rm LN}$ with a matching first-order bound in the $\epsilon$-dominated regime. These pieces yield a unified forward-stability bound whose right-hand side is directly estimable during training. On Tiny-ViT/CIFAR-10 we evaluate the bound and components. (1) The combined predictor $\kappa_{\rm softmax},(1+\kappa_{\rm score}),\kappa(V),|W_O|2+\kappa{\rm eff}+C_{\rm LN}$ tracks FP32$\leftrightarrow$LP mismatches across seeds, widths, and precisions; scaling by $\epsilon_{\rm mach}$ collapses mixed-precision points. (2) The time-series maximum of $\kappa_{\rm softmax}$ acts as an early-warning signal, leading error spikes by 16-24 steps (corr. 0.65-0.82; permutation $p!\approx!10^{-3}$; Precision@K 0.89-1.00). (3) Guided by $\rho_{\rm LN}$, a small LayerNorm-$\epsilon$ tweak targeting $\rho_\star$ gives consistent stabilization (mean tail-loss $\downarrow\ \approx0.010$ at $\rho_\star!=!0.6$, cap$=10^{-2}$) with negligible overhead. Overall, our theory supplies actionable, unitless diagnostics that (i) explain when self-attention is fragile, (ii) forecast instability, and (iii) motivate a minimally invasive mitigation.
Authors: Jinwoo Baek
Abstract: We introduce \textbf{Chebyshev Moment Regularization (CMR)}, a simple, architecture-agnostic loss that directly optimizes layer spectra. CMR jointly controls spectral edges via a log-condition proxy and shapes the interior via Chebyshev moments, with a decoupled, capped mixing rule that preserves task gradients. We prove strictly monotone descent for the condition proxy, bounded moment gradients, and orthogonal invariance. In an adversarial ``$\kappa$-stress'' setting (MNIST, 15-layer MLP), \emph{compared to vanilla training}, CMR reduces mean layer condition numbers by $\sim\!10^3$ (from $\approx3.9\!\times\!10^3$ to $\approx3.4$ in 5 epochs), increases average gradient magnitude, and restores test accuracy ( $\approx10\%\!\to\!\approx86\%$ ). These results support \textbf{optimization-driven spectral preconditioning}: directly steering models toward well-conditioned regimes for stable, accurate learning.
Authors: Supriya Nagesh, Karina Covarrubias, Robert El-Kareh, Shiva Prasad Kasiviswanathan, Nina Mishra
Abstract: Background: Aspiration, the inhalation of foreign material into the lungs, significantly impacts surgical patient morbidity and mortality. This study develops a machine learning (ML) model to predict postoperative aspiration, enabling timely preventative interventions. Methods: From the MIMIC-IV database of over 400,000 hospital admissions, we identified 826 surgical patients (mean age: 62, 55.7\% male) who experienced aspiration within seven days post-surgery, along with a matched non-aspiration cohort. Three ML models: XGBoost, Multilayer Perceptron, and Random Forest were trained using pre-surgical hospitalization data to predict postoperative aspiration. To investigate causation, we estimated Average Treatment Effects (ATE) using Augmented Inverse Probability Weighting. Results: Our ML model achieved an AUROC of 0.86 and 77.3\% sensitivity on a held-out test set. Maximum daily opioid dose, length of stay, and patient age emerged as the most important predictors. ATE analysis identified significant causative factors: opioids (0.25 +/- 0.06) and operative site (neck: 0.20 +/- 0.13, head: 0.19 +/- 0.13). Despite equal surgery rates across genders, men were 1.5 times more likely to aspirate and received 27\% higher maximum daily opioid dosages compared to women. Conclusion: ML models can effectively predict postoperative aspiration risk, enabling targeted preventative measures. Maximum daily opioid dosage and operative site significantly influence aspiration risk. The gender disparity in both opioid administration and aspiration rates warrants further investigation. These findings have important implications for improving postoperative care protocols and aspiration prevention strategies.
Authors: Larkin Liu, Jalal Etesami
Abstract: We explore the use of expert-guided bandit learning, which we refer to as online mixture-of-experts (OMoE). In this setting, given a context, a candidate committee of experts must determine how to aggregate their outputs to achieve optimal results in terms of aggregate accuracy. We propose two algorithms to address this problem. The first algorithm combines aggregate voting with UCB-driven successive elimination, efficiently pruning suboptimal exploration actions. The second algorithm employs an online weighted-majority-voting mechanism, leveraging the respective voting power of each expert proportional to their predictive power. We derive theoretical guarantees for the regret properties in the bandit setting under ideal circumstances, and empirical results are provided accordingly. As a modern study on applications, these methods are applied to the online fine-tuning of a set of expert large language models (LLMs), where after each response, the generative LLM dynamically reweighs its set of experts and/or selects the optimal committee of experts to generate the most accurate response. Our results introduce new methodologies and no-regret guarantees for combining multiple experts to improve on the performance of the an aggregate model overall.
Authors: Shifeng Xu, Yanzhu Liu, Adams Wai-Kin Kong
Abstract: Diffusion models have become emerging generative models. Their sampling process involves multiple steps, and in each step the models predict the noise from a noisy sample. When the models make prediction, the output deviates from the ground truth, and we call such a deviation as \textit{prediction error}. The prediction error accumulates over the sampling process and deteriorates generation quality. This paper introduces a novel technique for statistically measuring the prediction error and proposes the Variance-Reduction Guidance (VRG) method to mitigate this error. VRG does not require model fine-tuning or modification. Given a predefined sampling trajectory, it searches for a new trajectory which has the same number of sampling steps but produces higher quality results. VRG is applicable to both conditional and unconditional generation. Experiments on various datasets and baselines demonstrate that VRG can significantly improve the generation quality of diffusion models. Source code is available at https://github.com/shifengxu/VRG.
Authors: Xiao Zhou, Yuze Sun, Jie Wu, Xiaomeng Huang
Abstract: The Madden-Julian Oscillation (MJO) is an important driver of global weather and climate extremes, but its prediction in operational dynamical models remains challenging, with skillful forecasts typically limited to 3-4 weeks. Here, we introduce a novel deep learning framework, the Physics-guided Cascaded Corrector for MJO (PCC-MJO), which acts as a universal post-processor to correct MJO forecasts from dynamical models. This two-stage model first employs a physics-informed 3D U-Net to correct spatial-temporal field errors, then refines the MJO's RMM index using an LSTM optimized for forecast skill. When applied to three different operational forecasts from CMA, ECMWF and NCEP, our unified framework consistently extends the skillful forecast range (bivariate correlation > 0.5) by 2-8 days. Crucially, the model effectively mitigates the "Maritime Continent barrier", enabling more realistic eastward propagation and amplitude. Explainable AI analysis quantitatively confirms that the model's decision-making is spatially congruent with observed MJO dynamics (correlation > 0.93), demonstrating that it learns physically meaningful features rather than statistical fittings. Our work provides a promising physically consistent, computationally efficient, and highly generalizable pathway to break through longstanding barriers in subseasonal forecasting.
Authors: Zhaocheng Liu, Zhiwen Yu, Xiaoqing Liu
Abstract: Current mainstream approaches to addressing multimodal imbalance primarily focus on architectural modifications and optimization-based, often overlooking a quantitative analysis of the imbalance degree between modalities. To address this gap, our work introduces a novel method for the quantitative analysis of multi-modal imbalance, which in turn informs the design of a sample-level adaptive loss function.We begin by defining the "Modality Gap" as the difference between the Softmax scores of different modalities (e.g., audio and visual) for the ground-truth class prediction. Analysis of the Modality Gap distribution reveals that it can be effectively modeled by a bimodal Gaussian Mixture Model (GMM). These two components are found to correspond respectively to "modality-balanced" and "modality-imbalanced" data samples. Subsequently, we apply Bayes' theorem to compute the posterior probability of each sample belonging to these two distinct distributions.Informed by this quantitative analysis, we design a novel adaptive loss function with three objectives: (1) to minimize the overall Modality Gap; (2) to encourage the imbalanced sample distribution to shift towards the balanced one; and (3) to apply greater penalty weights to imbalanced samples. We employ a two-stage training strategy consisting of a warm-up phase followed by an adaptive training phase.Experimental results demonstrate that our approach achieves state-of-the-art (SOTA) performance on the public CREMA-D and AVE datasets, attaining accuracies of $80.65\%$ and $70.90\%$, respectively. This validates the effectiveness of our proposed methodology.
Authors: Yifeng Liu, Angela Yuan, Quanquan Gu
Abstract: Matrix-based preconditioned optimizers, such as Muon, have recently been shown to be more efficient than scalar-based optimizers for training large-scale neural networks, including large language models (LLMs). On the other hand, recent benchmarks on optimizers for LLM pre-training have demonstrated that variance-reduction techniques such as MARS can achieve substantial speedups over standard optimizers that do not employ variance reduction. In this paper, to achieve the best of both worlds, we introduce MARS-M, a new optimizer that integrates the variance reduction technique in MARS with Muon. Under standard regularity conditions, we prove that Muon-M converges to a first-order stationary point at a rate of $\tilde{\mathcal{O}}(T^{-1/3})$, which improves upon $\tilde{\mathcal{O}}(T^{-1/4})$ rate attained by Muon. Our empirical results on language modeling and computer vision tasks demonstrate that MARS-M consistently yields lower losses and improved performance across various downstream benchmarks. The implementation of MARS-M is available at https://github.com/AGI-Arena/MARS/MARS_M.
Authors: Shilaj Baral, Youngkyu Lee, Sangam Khanal, Joongoo Jeon
Abstract: Purely data-driven surrogates for fluid dynamics often fail catastrophically from error accumulation, while existing hybrid methods have lacked the automation and robustness for practical use. To solve this, we developed XRePIT, a novel hybrid simulation strategy that synergizes machine learning (ML) acceleration with solver-based correction. We specifically designed our method to be fully automated and physics-aware, ensuring the stability and practical applicability that previous approaches lacked. We demonstrate that this new design overcomes long-standing barriers, achieving the first stable, accelerated rollouts for over 10,000 timesteps. The method also generalizes robustly to unseen boundary conditions and, crucially, scales to 3D flows. Our approach delivers speedups up to 4.98$\times$ while maintaining high physical fidelity, resolving thermal fields with relative errors of ~1E-3 and capturing low magnitude velocity dynamics with errors below 1E-2 ms-1. This work thus establishes a mature and scalable hybrid method, paving the way for its use in real-world engineering.
Authors: Marcelo Cerda Castillo
Abstract: Short-term forecasting of airport fog (visibility < 1.0 km) presents challenges in geographic generalization because many machine learning models rely on location-specific features and fail to transfer across sites. This study investigates whether fundamental thermodynamic and radiative processes can be encoded in a coordinate-free (location-independent) feature set to enable geographic transferability. A gradient boosting classifier (XGBoost) trained on Santiago, Chile (SCEL, 33S) data from 2002-2009 was evaluated on a 2010-2012 holdout set and under strict zero-shot tests at Puerto Montt (SCTE), San Francisco (KSFO), and London (EGLL). The model achieved AUC values of 0.923-0.947 across distances up to 11,650 km and different fog regimes (radiative, advective, marine). Consistent SHAP feature rankings show that visibility persistence, solar angle, and thermal gradients dominate predictions, suggesting the model learned transferable physical relationships rather than site-specific patterns. Results suggest that physics-informed, coordinate-free feature engineering can yield geographically transferable atmospheric forecasting tools.
Authors: Rekha R Nair, Tina Babu, Alavikunhu Panthakkan, Hussain Al-Ahmad, Balamurugan Balusamy
Abstract: The proliferation of high-dimensional datasets in fields such as genomics, healthcare, and finance has created an urgent need for machine learning models that are both highly accurate and inherently interpretable. While traditional deep learning approaches deliver strong predictive performance, their lack of transparency often impedes their deployment in critical, decision-sensitive applications. In this work, we introduce the Hierarchical Attention-based Interpretable Network (HAIN), a novel architecture that unifies multi-level attention mechanisms, dimensionality reduction, and explanation-driven loss functions to deliver interpretable and robust analysis of complex biomedical data. HAIN provides feature-level interpretability via gradientweighted attention and offers global model explanations through prototype-based representations. Comprehensive evaluation on The Cancer Genome Atlas (TCGA) dataset demonstrates that HAIN achieves a classification accuracy of 94.3%, surpassing conventional post-hoc interpretability approaches such as SHAP and LIME in both transparency and explanatory power. Furthermore, HAIN effectively identifies biologically relevant cancer biomarkers, supporting its utility for clinical and research applications. By harmonizing predictive accuracy with interpretability, HAIN advances the development of transparent AI solutions for precision medicine and regulatory compliance.
Authors: Azim Ahmadzadeh, Mahsa Khazaei, Elaina Rohlfing
Abstract: Time series are high-dimensional and complex data objects, making their efficient search and indexing a longstanding challenge in data mining. Building on a recently introduced similarity measure, namely Multiscale Dubuc Distance (MDD), this paper investigates its comparative strengths and limitations relative to the widely used Dynamic Time Warping (DTW). MDD is novel in two key ways: it evaluates time series similarity across multiple temporal scales and avoids point-to-point alignment. We demonstrate that in many scenarios where MDD outperforms DTW, the gains are substantial, and we provide a detailed analysis of the specific performance gaps it addresses. We provide simulations, in addition to the 95 datasets from the UCR archive, to test our hypotheses. Finally, we apply both methods to a challenging real-world classification task and show that MDD yields a significant improvement over DTW, underscoring its practical utility.
Authors: Jianqing Zhang, Zhezheng Hao, Wei Xia, Hande Dong, Hong Wang, Chenxing Wei, Yuyan Zhou, Yubin Qi, Qiang Lin, Jian Cao
Abstract: Reinforcement learning (RL) is widely used for post-training large language models (LLMs) in code editing, where group-relative methods like GRPO are popular for their critic-free, normalized advantage estimation. However, in real-world code-editing scenarios, reward distributions are often skewed with unpredictable outliers, leading to distorted advantage computation and increased noise. To address this issue, we propose Group Adaptive Policy Optimization (GAPO), which adaptively finds an outlier-free highest-density interval (HDI) per prompt and then uses the median of that interval as an adaptive Q to replace the group mean in advantage calculation. This adaptive Q robustly handles skewed distributions while remaining plug-and-play and efficient. We validate GAPO on nine instruction-tuned LLMs (3B-14B) using a large internal dataset of 51,844 real-world, history-aware code-editing tasks across 10 languages, demonstrating consistent improvements in exact match accuracy over GRPO and its variant DAPO. Code is publicly available.
Authors: Zijian Feng, Hanzhang Zhou, Zixiao Zhu, Tianjiao Li, Jia Jim Deryl Chua, Lee Onn Mak, Gee Wah Ng, Kezhi Mao
Abstract: Pruning is a widely used technique to reduce the size and inference cost of large language models (LLMs), but it often causes performance degradation. To mitigate this, existing restoration methods typically employ parameter-efficient fine-tuning (PEFT), such as LoRA, to recover the pruned model's performance. However, most PEFT methods are designed for dense models and overlook the distinct properties of pruned models, often resulting in suboptimal recovery. In this work, we propose a targeted restoration strategy for pruned models that restores performance while preserving their low cost and high efficiency. We observe that pruning-induced information loss is reflected in attention activations, and selectively reintroducing components of this information can significantly recover model performance. Based on this insight, we introduce RestoreLCC (Restoring Pruned LLMs via Lost Component Compensation), a plug-and-play method that contrastively probes critical attention heads via activation editing, extracts lost components from activation differences, and finally injects them back into the corresponding pruned heads for compensation and recovery. RestoreLCC is compatible with structured, semi-structured, and unstructured pruning schemes. Extensive experiments demonstrate that RestoreLCC consistently outperforms state-of-the-art baselines in both general and task-specific performance recovery, without compromising the sparsity or inference efficiency of pruned models.
Authors: Nayan Kumar Singh
Abstract: Manually generating catchy descriptions and names is labor intensive and a slow process for retailers. Although generative AI provides an automation solution in form of Vision to Language Models (VLM), the current VLMs are prone to factual "hallucinations". Siloed, single task models are not only inefficient but also fail to capture interdependent relationships between features. To address these challenges, we propose an end to end, multi task system that generates factually grounded textual listings from a single image. The contributions of this study are two proposals for the model architecture. First, application of multi task learning approach for fine tuning a vision encoder where a single vision backbone is jointly trained on attribute prediction such as color, hemline and neck style and price regression. Second, introduction of a hierarchical generation process where the model's own predicted attributes are embedded in a prompt and fed to the text decoder to improve factual consistency. The experiments demonstrate the superiority of this architecture. The multi tasking approach outperforms both the independent price regression, with a 3.6% better R2 Value and attribute classification, with a 6.6% improvement F1 score. Critically, the hierarchical generation process proves highly effective, slashing the factual hallucination rate from 12.7% to 7.1%, a 44.5% relative reduction, compared to a non hierarchical ablation. The hierarchical approach also reduces the latency of the autoregressive text generation process by a factor of 3.5 when compared to direct vision to language model of similar size. One minor caveat is that the model does perform 3.5% worse than direct vision-to-language model on ROUGE-L score.
Authors: Jaya Krishna Mandivarapu
Abstract: Learning a set of tasks over time, also known as continual learning (CL), is one of the most challenging problems in artificial intelligence due to catastrophic forgetting. Large language models (LLMs) are often impractical to frequent re-training and continual learning , due to high cost of computational resources for training. Moreover, LLM are not suitable for continual learning as updating these models over time for acquiring new knowledge leads to overwrites existing knowledge leading to common phenomenon know as \textit{catastrophic forgetting}. In this paper, we aim to address these concerns using a novel framework , COLA that employs an autoencoder to learn capture low-dimensional embeddings of the weights associated with various tasks. Our approach facilitates the transfer of knowledge to new tasks while preventing catastrophic forgetting, all without using data replay or a substantial set of task-specific parameters. Our approach, COLA, makes the LLM efficiently learn new tasks with minimal training, insignificant performance degradation on previous tasks, and eliminates the need for retaining earlier training data. Empirical evaluation on different datasets ranging from task oriented dialouge system to intent classsfication datasets showcases that our method not only overcomes catastrophic forgetting but also achieves significant reduction in parameter usage and memory size, across multiple tasks and outperforming the existing state of the art methods across multiple datasets.
Authors: Azree Nazri
Abstract: Large Language Models (LLMs) like ChatGPT and LLaMA drive rapid progress in generative AI, yet their huge parameter scales create severe computational and environmental burdens. High training costs, energy use, and limited device deployment hinder accessibility. Existing compression - pruning, distillation, low-rank, and quantization - reduces size but ignores complex inter-layer correlations. We propose KARIPAP, a quantum-inspired tensor network compression using Infinite Projected Entangled Pair States (iPEPS) and Tensor Renormalization Group (TRG) contraction. Unlike 1D Matrix Product States, iPEPS captures multi-directional entanglement in attention and deep transformer layers. TRG ensures polynomial-time contraction, making tensorization feasible while preserving key correlation geometry. Experiments on LLaMA-2 7B show up to 93% memory and 70% parameter reduction, with 50% faster training, 25% faster inference, and only 2-3% accuracy loss. Layer-wise entanglement profiling reveals redundancy in deeper layers, confirming their suitability for tensor factorization. KARIPAP demonstrates that modern LLMs occupy low-dimensional entanglement manifolds, enabling scalable, energy-efficient, and quantum-aware AI architectures.
Authors: Yongchao Huang, Pengfei Zhang, Shahzad Mumtaz
Abstract: Membership inference attacks (MIAs) test whether a data point was part of a model's training set, posing serious privacy risks. Existing methods often depend on shadow models or heavy query access, which limits their practicality. We propose GP-MIA, an efficient and interpretable approach based on Gaussian process (GP) meta-modeling. Using post-hoc metrics such as accuracy, entropy, dataset statistics, and optional sensitivity features (e.g. gradients, NTK measures) from a single trained model, GP-MIA trains a GP classifier to distinguish members from non-members while providing calibrated uncertainty estimates. Experiments on synthetic data, real-world fraud detection data, CIFAR-10, and WikiText-2 show that GP-MIA achieves high accuracy and generalizability, offering a practical alternative to existing MIAs.
Authors: Kaiyi Xu, Junchao Gong, Wenlong Zhang, Ben Fei, Lei Bai, Wanli Ouyang
Abstract: Precipitation nowcasting based on radar echoes plays a crucial role in monitoring extreme weather and supporting disaster prevention. Although deep learning approaches have achieved significant progress, they still face notable limitations. For example, deterministic models tend to produce over-smoothed predictions, which struggle to capture extreme events and fine-scale precipitation patterns. Probabilistic generative models, due to their inherent randomness, often show fluctuating performance across different metrics and rarely achieve consistently optimal results. Furthermore, precipitation nowcasting is typically evaluated using multiple metrics, some of which are inherently conflicting. For instance, there is often a trade-off between the Critical Success Index (CSI) and the False Alarm Ratio (FAR), making it challenging for existing models to deliver forecasts that perform well on both metrics simultaneously. To address these challenges, we introduce preference optimization into precipitation nowcasting for the first time, motivated by the success of reinforcement learning from human feedback in large language models. Specifically, we propose SynCast, a method that employs the two-stage post-training framework of Diffusion Sequential Preference Optimization (Diffusion-SPO), to progressively align conflicting metrics and consistently achieve superior performance. In the first stage, the framework focuses on reducing FAR, training the model to effectively suppress false alarms. Building on this foundation, the second stage further optimizes CSI with constraints that preserve FAR alignment, thereby achieving synergistic improvements across these conflicting metrics.
Authors: Andr\'e G. Viveiros, Patrick Fernandes, Saul Santos, Sonal Sannigrahi, Emmanouil Zaranis, Nuno M. Guerreiro, Amin Farajian, Pierre Colombo, Graham Neubig, Andr\'e F. T. Martins
Abstract: Despite significant advances in vision-language models (VLMs), most existing work follows an English-centric design process, limiting their effectiveness in multilingual settings. In this work, we provide a comprehensive empirical study analyzing the impact of several multilingual design choices, such as training data composition, encoder selection, and text backbones. The result is TowerVision, a family of open multilingual VLMs for both image-text and video-text tasks, built upon the multilingual text-only model Tower+. TowerVision achieves competitive performance on multiple multimodal multilingual benchmarks and shows particular strength in culturally grounded tasks and multimodal translation. By incorporating visual and cultural context during fine-tuning, our models surpass existing approaches trained on substantially larger datasets, as demonstrated on ALM-Bench and Multi30K (image tasks) and ViMUL-Bench (video tasks). Alongside the models, we release VisionBlocks, a high-quality, curated vision-language dataset. Our findings highlight that multilingual vision-language training data substantially improves cross-lingual generalization -- both from high-resource to underrepresented languages and vice versa -- and that instruction-tuned LLMs are not always the optimal initialization point. To support further research, we publicly release all models, data, and training recipes.
Authors: Hojin Kim, Romit Maulik
Abstract: We present a differentiable framework that leverages the Discrete Empirical Interpolation Method (DEIM) for interpretable deep learning and dynamical system analysis. Although DEIM efficiently approximates nonlinear terms in projection-based reduced-order models (POD-ROM), its fixed interpolation points limit the adaptability to complex and time-varying dynamics. To address this limitation, we first develop a differentiable adaptive DEIM formulation for the one-dimensional viscous Burgers equation, which allows neural networks to dynamically select interpolation points in a computationally efficient and physically consistent manner. We then apply DEIM as an interpretable analysis tool for examining the learned dynamics of a pre-trained Neural Ordinary Differential Equation (NODE) on a two-dimensional vortex-merging problem. The DEIM trajectories reveal physically meaningful features in the learned dynamics of NODE and expose its limitations when extrapolating to unseen flow configurations. These findings demonstrate that DEIM can serve not only as a model reduction tool but also as a diagnostic framework for understanding and improving the generalization behavior of neural differential equation models.
Authors: Yangze Zhou, Ruiyang Yao, Dalin Qin, Yixiong Jia, Yi Wang
Abstract: Decision-making for multi-energy system (MES) dispatch depends on accurate load forecasting. Traditionally, load forecasting and decision-making for MES are implemented separately. Forecasting models are typically trained to minimize forecasting errors, overlooking their impact on downstream decision-making. To address this, decision-focused learning (DFL) has been studied to minimize decision-making costs instead. However, practical adoption of DFL in MES faces significant challenges: the process requires sharing sensitive load data and model parameters across multiple sectors, raising serious privacy issues. To this end, we propose a privacy-preserving DFL framework tailored for MES. Our approach introduces information masking to safeguard private data while enabling recovery of decision variables and gradients required for model training. To further enhance security for DFL, we design a safety protocol combining matrix decomposition and homomorphic encryption, effectively preventing collusion and unauthorized data access. Additionally, we developed a privacy-preserving load pattern recognition algorithm, enabling the training of specialized DFL models for heterogeneous load patterns. Theoretical analysis and comprehensive case studies, including real-world MES data, demonstrate that our framework not only protects privacy but also consistently achieves lower average daily dispatch costs compared to existing methods.
Authors: Shuang Wang, Xuben Wang, Fei Deng, Peifan Jiang, Jian Chen, Gianluca Fiandaca
Abstract: With the remarkable success of deep learning, applying such techniques to EM methods has emerged as a promising research direction to overcome the limitations of conventional approaches. The effectiveness of deep learning methods depends heavily on the quality of datasets, which directly influences model performance and generalization ability. Existing application studies often construct datasets from random one-dimensional or structurally simple three-dimensional models, which fail to represent the complexity of real geological environments. Furthermore, the absence of standardized, publicly available three-dimensional geoelectric datasets continues to hinder progress in deep learning based EM exploration. To address these limitations, we present OpenEM, a large scale, multi structural three dimensional geoelectric dataset that encompasses a broad range of geologically plausible subsurface structures. OpenEM consists of nine categories of geoelectric models, spanning from simple configurations with anomalous bodies in half space to more complex structures such as flat layers, folded layers, flat faults, curved faults, and their corresponding variants with anomalous bodies. Since three-dimensional forward modeling in electromagnetics is extremely time-consuming, we further developed a deep learning based fast forward modeling approach for OpenEM, enabling efficient and reliable forward modeling across the entire dataset. This capability allows OpenEM to be rapidly deployed for a wide range of tasks. OpenEM provides a unified, comprehensive, and large-scale dataset for common EM exploration systems to accelerate the application of deep learning in electromagnetic methods. The complete dataset, along with the forward modeling codes and trained models, is publicly available at https://doi.org/10.5281/zenodo.17141981.
Authors: Bentley DeVilling (Course Correct Labs, Independent Research Group)
Abstract: Large language models are often described as capable of reflective reasoning, yet recursive self-evaluation without external feedback frequently yields reformulation rather than progress. We test this prediction in a cross-provider study of 144 reasoning sequences across three models (OpenAI GPT-4o-mini, Anthropic Claude 3 Haiku, and Google Gemini 2.0 Flash) and four task families (arithmetic, code, explanation, reflection), each iterated ten times under two conditions: ungrounded self-critique and a minimal grounding intervention (a single verification step at iteration three). Mean informational change (delta I, measured via normalized edit distance) declined by 55% from early (0.193) to late (0.087) iterations in ungrounded runs, with consistent patterns across all three providers. Grounded runs showed a +28% rebound in informational change immediately after the intervention and sustained non-zero variance thereafter. Complementary measures-n-gram novelty, embedding drift, and character-level entropy-converged on the same pattern: reflection without contact tends toward informational closure. We interpret this as evidence for a structural limit on self-correction in generative reasoning: without an exchange of information with an independent verifier or environment, recursive inference approaches an attractor state of epistemic stasis. Minimal grounding functions as dissipative coupling, reintroducing informational flux. The cross-architecture consistency suggests the mirror loop arises from shared autoregressive training objectives rather than provider-specific alignment schemes. The results delineate when reflection is performative rather than epistemic and motivate design principles for grounded, cooperative reasoning. Materials and code are publicly available.
Authors: Chieh-Hsin Lai, Yang Song, Dongjun Kim, Yuki Mitsufuji, Stefano Ermon
Abstract: This monograph presents the core principles that have guided the development of diffusion models, tracing their origins and showing how diverse formulations arise from shared mathematical ideas. Diffusion modeling starts by defining a forward process that gradually corrupts data into noise, linking the data distribution to a simple prior through a continuum of intermediate distributions. The goal is to learn a reverse process that transforms noise back into data while recovering the same intermediates. We describe three complementary views. The variational view, inspired by variational autoencoders, sees diffusion as learning to remove noise step by step. The score-based view, rooted in energy-based modeling, learns the gradient of the evolving data distribution, indicating how to nudge samples toward more likely regions. The flow-based view, related to normalizing flows, treats generation as following a smooth path that moves samples from noise to data under a learned velocity field. These perspectives share a common backbone: a time-dependent velocity field whose flow transports a simple prior to the data. Sampling then amounts to solving a differential equation that evolves noise into data along a continuous trajectory. On this foundation, the monograph discusses guidance for controllable generation, efficient numerical solvers, and diffusion-motivated flow-map models that learn direct mappings between arbitrary times. It provides a conceptual and mathematically grounded understanding of diffusion models for readers with basic deep-learning knowledge.
Authors: Fadi Dornaika, Ahmad Khoder, Abdelmalik Moujahid, Wassim Khoder
Abstract: The performance of machine learning and pattern recognition algorithms generally depends on data representation. That is why, much of the current effort in performing machine learning algorithms goes into the design of preprocessing frameworks and data transformations able to support effective machine learning. The method proposed in this work consists of a hybrid linear feature extraction scheme to be used in supervised multi-class classification problems. Inspired by two recent linear discriminant methods: robust sparse linear discriminant analysis (RSLDA) and inter-class sparsitybased discriminative least square regression (ICS_DLSR), we propose a unifying criterion that is able to retain the advantages of these two powerful methods. The resulting transformation relies on sparsity-promoting techniques both to select the features that most accurately represent the data and to preserve the row-sparsity consistency property of samples from the same class. The linear transformation and the orthogonal matrix are estimated using an iterative alternating minimization scheme based on steepest descent gradient method and different initialization schemes. The proposed framework is generic in the sense that it allows the combination and tuning of other linear discriminant embedding methods. According to the experiments conducted on several datasets including faces, objects, and digits, the proposed method was able to outperform competing methods in most cases.
Authors: Mahavir Dabas, Tran Huynh, Nikhil Reddy Billa, Jiachen T. Wang, Peng Gao, Charith Peris, Yao Ma, Rahul Gupta, Ming Jin, Prateek Mittal, Ruoxi Jia
Abstract: Large language models remain vulnerable to jailbreak attacks that bypass safety guardrails to elicit harmful outputs. Defending against novel jailbreaks represents a critical challenge in AI safety. Adversarial training -- designed to make models robust against worst-case perturbations -- has been the dominant paradigm for adversarial robustness. However, due to optimization challenges and difficulties in defining realistic threat models, adversarial training methods often fail on newly developed jailbreaks in practice. This paper proposes a new paradigm for improving robustness against unseen jailbreaks, centered on the Adversarial D\'ej\`a Vu hypothesis: novel jailbreaks are not fundamentally new, but largely recombinations of adversarial skills from previous attacks. We study this hypothesis through a large-scale analysis of 32 attack papers published over two years. Using an automated pipeline, we extract and compress adversarial skills into a sparse dictionary of primitives, with LLMs generating human-readable descriptions. Our analysis reveals that unseen attacks can be effectively explained as sparse compositions of earlier skills, with explanatory power increasing monotonically as skill coverage grows. Guided by this insight, we introduce Adversarial Skill Compositional Training (ASCoT), which trains on diverse compositions of skill primitives rather than isolated attack instances. ASCoT substantially improves robustness to unseen attacks, including multi-turn jailbreaks, while maintaining low over-refusal rates. We also demonstrate that expanding adversarial skill coverage, not just data scale, is key to defending against novel attacks. \textcolor{red}{\textbf{Warning: This paper contains content that may be harmful or offensive in nature.
Authors: Fardin Gankhanloo, Emmett Springer, Erik H. Hoyer, Daniel L. Young, Kimia Ghobadi
Abstract: Risk assessment tools in healthcare commonly employ point-based scoring systems that map patients to ordinal risk categories via thresholds. While electronic health record (EHR) data presents opportunities for data-driven optimization of these tools, two fundamental challenges impede standard supervised learning: (1) partial supervision arising from intervention-censored outcomes, where only extreme categories can be reliably labeled, and (2) asymmetric misclassification costs that increase with ordinal distance. We propose a mixed-integer programming (MIP) framework that jointly optimizes scoring weights and category thresholds under these constraints. Our approach handles partial supervision through per-instance feasible label sets, incorporates asymmetric distance-aware objectives, and prevents middle-category collapse via minimum threshold gaps. We further develop a CSO relaxation using softplus losses that preserves the ordinal structure while enabling efficient optimization. The framework supports governance constraints including sign restrictions, sparsity, and minimal modifications to incumbent tools, ensuring practical deployability in clinical workflows.
Authors: Samuel Bright-Thonney, Christina Reissel, Gaia Grosso, Nathaniel Woodward, Katya Govorkova, Andrzej Novak, Sang Eon Park, Eric Moreno, Philip Harris
Abstract: Novelty detection in large scientific datasets faces two key challenges: the noisy and high-dimensional nature of experimental data, and the necessity of making statistically robust statements about any observed outliers. While there is a wealth of literature on anomaly detection via dimensionality reduction, most methods do not produce outputs compatible with quantifiable claims of scientific discovery. In this work we directly address these challenges, presenting the first step towards a unified pipeline for novelty detection adapted for the rigorous statistical demands of science. We introduce AutoSciDACT (Automated Scientific Discovery with Anomalous Contrastive Testing), a general-purpose pipeline for detecting novelty in scientific data. AutoSciDACT begins by creating expressive low-dimensional data representations using a contrastive pre-training, leveraging the abundance of high-quality simulated data in many scientific domains alongside expertise that can guide principled data augmentation strategies. These compact embeddings then enable an extremely sensitive machine learning-based two-sample test using the New Physics Learning Machine (NPLM) framework, which identifies and statistically quantifies deviations in observed data relative to a reference distribution (null hypothesis). We perform experiments across a range of astronomical, physical, biological, image, and synthetic datasets, demonstrating strong sensitivity to small injections of anomalous data across all domains.
Authors: Antoine Ledent, Rodrigo Alves, Yunwen Lei
Abstract: It has been recently observed in much of the literature that neural networks exhibit a bottleneck rank property: for larger depths, the activation and weights of neural networks trained with gradient-based methods tend to be of approximately low rank. In fact, the rank of the activations of each layer converges to a fixed value referred to as the ``bottleneck rank'', which is the minimum rank required to represent the training data. This perspective is in line with the observation that regularizing linear networks (without activations) with weight decay is equivalent to minimizing the Schatten $p$ quasi norm of the neural network. In this paper we investigate the implications of this phenomenon for generalization. More specifically, we prove generalization bounds for neural networks which exploit the approximate low rank structure of the weight matrices if present. The final results rely on the Schatten $p$ quasi norms of the weight matrices: for small $p$, the bounds exhibit a sample complexity $ \widetilde{O}(WrL^2)$ where $W$ and $L$ are the width and depth of the neural network respectively and where $r$ is the rank of the weight matrices. As $p$ increases, the bound behaves more like a norm-based bound instead.
Authors: J. Jon Ryu, Samuel Zhou, Gregory W. Wornell
Abstract: Spectral decomposition of linear operators plays a central role in many areas of machine learning and scientific computing. Recent work has explored training neural networks to approximate eigenfunctions of such operators, enabling scalable approaches to representation learning, dynamical systems, and partial differential equations (PDEs). In this paper, we revisit a classical optimization framework from the computational physics literature known as the \emph{orbital minimization method} (OMM), originally proposed in the 1990s for solving eigenvalue problems in computational chemistry. We provide a simple linear-algebraic proof of the consistency of the OMM objective, and reveal connections between this method and several ideas that have appeared independently across different domains. Our primary goal is to justify its broader applicability in modern learning pipelines. We adapt this framework to train neural networks to decompose positive semidefinite operators, and demonstrate its practical advantages across a range of benchmark tasks. Our results highlight how revisiting classical numerical methods through the lens of modern theory and computation can provide not only a principled approach for deploying neural networks in numerical simulation, but also effective and scalable tools for machine learning.
Authors: Armin Gerami, Ramani Duraiswami
Abstract: The original softmax-based attention mechanism (regular attention) in the extremely successful Transformer architecture computes attention between $N$ tokens, each embedded in a $D$-dimensional head, with a time complexity of $O(N^2D)$. Given the success of Transformers, improving their runtime during both training and inference is a popular research area. One such approach is the introduction of the linear attention (LA) mechanisms, which offers a linear time complexity of $O(ND^2)$ and have demonstrated comparable accuracy to regular attention. However, LA in practice lags behind its theoretical efficiency. We propose a novel method for LA's forward and backward passes, along with a highly-optimized CUDA implementation. Our approach outperforms the state-of-the-art by 3.3 times in speed and reduces memory consumption by 3.6 times. We validate these improvements in both single-layer and end-to-end settings by training a 1.4 billion parameter language model, which demonstrates similar expressivity to regular attention on major reasoning benchmarks.
Authors: Iskander Azangulov, Teodora Pandeva, Niranjani Prasad, Javier Zazo, Sushrut Karmalkar
Abstract: Masked diffusion models (MDMs) offer a compelling alternative to autoregressive models (ARMs) for discrete text generation because they enable parallel token sampling, rather than sequential, left-to-right generation. This means potentially much faster inference. However, effective parallel sampling faces two competing requirements: (i) simultaneously updated tokens must be conditionally independent, and (ii) updates should prioritise high-confidence predictions. These goals conflict because high-confidence predictions often cluster and depend on each other, opportunities for parallel updates. We present PUNT, a model-agnostic sampler that reconciles this trade-off. Our method identifies token dependencies and removes lower-confidence tokens from conflicting groups. This produces sets of indices for unmasking that satisfy both independence and confidence criteria. Our approach ensures improved parallel unmasking through approximate conditional independence testing. Our experiments show that PUNT delivers a superior trade-off between accuracy and compute when compared to other strong training-free baselines, especially for generation of longer sequences. On the IFEval benchmark, it achieves up to 16\% higher accuracy over baseline methods, including sequential generation (one-by-one). These gains hold across different values of hyperparameters, mitigating the need for brittle hyperparameter tuning. Moreover, we observe that PUNT induces an emergent hierarchical generation strategy, where the model first establishes high-level paragraph structure before local refinement, suggesting a planning-like generation process that contributes to strong alignment performance.
Authors: Yang Xu, Chiwoo Park
Abstract: We introduce Deep Jump Gaussian Processes (DJGP), a novel method for surrogate modeling of high-dimensional piecewise continuous functions. DJGP overcomes the limitations of conventional Jump Gaussian Processes in high-dimensional input spaces by adding a locally linear projection layer to Jump Gaussian Processes. This projection uses region-specific matrices to capture local subspace structures, naturally complementing the localized nature of JGP, a variant of local Gaussian Processes. To control model complexity, we place a Gaussian Process prior on the projection matrices, allowing them to evolve smoothly across the input space. The projected inputs are then modeled with a JGP to capture piecewise continuous relationships with the response. This yields a distinctive two-layer deep learning of GP/JGP. We further develop a scalable variational inference algorithm to jointly learn the projection matrices and JGP hyperparameters. Experiments on synthetic and benchmark datasets demonstrate that DJGP delivers superior predictive accuracy and more reliable uncertainty quantification compared to existing approaches.
Authors: Hoang Phan, Xianjun Yang, Kevin Yao, Jingyu Zhang, Shengjie Bi, Xiaocheng Tang, Madian Khabsa, Lijuan Liu, Deren Lei
Abstract: Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning and has become a standard post-training paradigm for contemporary language and vision-language models. However, the RLVR recipe introduces a significant risk of capability regression, where models forget foundational skills after prolonged training without employing regularization strategies. We empirically confirm this concern, observing that open-source reasoning models suffer performance degradation on core capabilities such as perception and faithfulness. While imposing regularization terms like KL divergence can help prevent deviation from the base model, these terms are calculated on the current task, thus they do not guarantee broader knowledge. Meanwhile, commonly used experience replay across heterogeneous domains makes it nontrivial to decide how much training focus each objective should receive. To address this, we propose RECAP-a replay strategy with dynamic objective reweighting for general knowledge preservation. Our reweighting mechanism adapts in an online manner using short-horizon signals of convergence and instability, shifting the post-training focus away from saturated objectives and toward underperforming or volatile ones. Our method is end-to-end and readily applicable to existing RLVR pipelines without training additional models or heavy tuning. Extensive experiments on benchmarks based on Qwen2.5-VL-3B and Qwen2.5-VL-7B demonstrate the effectiveness of our method, which not only preserves general capabilities but also improves reasoning by enabling more flexible trade-offs among in-task rewards.
Authors: Starlika Bauskar, Jade Jiao, Narayanan Kannan, Alexander Kimm, Justin M. Baker, Matthew J. Tyler, Andrea L. Bertozzi, Anne M. Andrews
Abstract: Machine-learning methods in biochemistry commonly represent molecules as graphs of pairwise intermolecular interactions for property and structure predictions. Most methods operate on a single graph, typically the minimal free energy (MFE) structure, for low-energy ensembles (conformations) representative of structures at thermodynamic equilibrium. We introduce a thermodynamically parameterized exponential-family random graph (ERGM) embedding that models molecules as Boltzmann-weighted ensembles of interaction graphs. We evaluate this embedding on SELEX datasets, where experimental biases (e.g., PCR amplification or sequencing noise) can obscure true aptamer-ligand affinity, producing anomalous candidates whose observed abundance diverges from their actual binding strength. We show that the proposed embedding enables robust community detection and subgraph-level explanations for aptamer ligand affinity, even in the presence of biased observations. This approach may be used to identify low-abundance aptamer candidates for further experimental evaluation.
Authors: Emanuele Rossi
Abstract: Graph Neural Networks (GNNs) have become a central tool for learning on graph-structured data, yet their applicability to real-world systems remains limited by key challenges such as scalability, temporality, directionality, data incompleteness, and structural uncertainty. This thesis introduces a series of models addressing these limitations: SIGN for scalable graph learning, TGN for temporal graphs, Dir-GNN for directed and heterophilic networks, Feature Propagation (FP) for learning with missing node features, and NuGget for game-theoretic structural inference. Together, these contributions bridge the gap between academic benchmarks and industrial-scale graphs, enabling the use of GNNs in domains such as social and recommender systems.
Authors: Micha{\l} Bortkiewicz, W{\l}adys{\l}aw Pa{\l}ucki, Mateusz Ostaszewski, Benjamin Eysenbach
Abstract: Reinforcement learning (RL) promises to solve long-horizon tasks even when training data contains only short fragments of the behaviors. This experience stitching capability is often viewed as the purview of temporal difference (TD) methods. However, outside of small tabular settings, trajectories never intersect, calling into question this conventional wisdom. Moreover, the common belief is that Monte Carlo (MC) methods should not be able to recombine experience, yet it remains unclear whether function approximation could result in a form of implicit stitching. The goal of this paper is to empirically study whether the conventional wisdom about stitching actually holds in settings where function approximation is used. We empirically demonstrate that Monte Carlo (MC) methods can also achieve experience stitching. While TD methods do achieve slightly stronger capabilities than MC methods (in line with conventional wisdom), that gap is significantly smaller than the gap between small and large neural networks (even on quite simple tasks). We find that increasing critic capacity effectively reduces the generalization gap for both the MC and TD methods. These results suggest that the traditional TD inductive bias for stitching may be less necessary in the era of large models for RL and, in some cases, may offer diminishing returns. Additionally, our results suggest that stitching, a form of generalization unique to the RL setting, might be achieved not through specialized algorithms (temporal difference learning) but rather through the same recipe that has provided generalization in other machine learning settings (via scale). Project website: https://michalbortkiewicz.github.io/golden-standard/
Authors: Inwoo Hwang, Yushu Pan, Elias Bareinboim
Abstract: Understanding the predictions made by deep learning models remains a central challenge, especially in high-stakes applications. A promising approach is to equip models with the ability to answer counterfactual questions -- hypothetical ``what if?'' scenarios that go beyond the observed data and provide insight into a model reasoning. In this work, we introduce the notion of causal interpretability, which formalizes when counterfactual queries can be evaluated from a specific class of models and observational data. We analyze two common model classes -- blackbox and concept-based predictors -- and show that neither is causally interpretable in general. To address this gap, we develop a framework for building models that are causally interpretable by design. Specifically, we derive a complete graphical criterion that determines whether a given model architecture supports a given counterfactual query. This leads to a fundamental tradeoff between causal interpretability and predictive accuracy, which we characterize by identifying the unique maximal set of features that yields an interpretable model with maximal predictive expressiveness. Experiments corroborate the theoretical findings.
Authors: T. Tony Cai, Xiang Li, Qi Long, Weijie J. Su, Garrett G. Wen
Abstract: Text watermarking plays a crucial role in ensuring the traceability and accountability of large language model (LLM) outputs and mitigating misuse. While promising, most existing methods assume perfect pseudorandomness. In practice, repetition in generated text induces collisions that create structured dependence, compromising Type I error control and invalidating standard analyses. We introduce a statistical framework that captures this structure through a hierarchical two-layer partition. At its core is the concept of minimal units -- the smallest groups treatable as independent across units while permitting dependence within. Using minimal units, we define a non-asymptotic efficiency measure and cast watermark detection as a minimax hypothesis testing problem. Applied to Gumbel-max and inverse-transform watermarks, our framework produces closed-form optimal rules. It explains why discarding repeated statistics often improves performance and shows that within-unit dependence must be addressed unless degenerate. Both theory and experiments confirm improved detection power with rigorous Type I error control. These results provide the first principled foundation for watermark detection under imperfect pseudorandomness, offering both theoretical insight and practical guidance for reliable tracing of model outputs.
Authors: Md Saiful Islam Sajol, Magesh Rajasekaran, Hayden Gemeinhardt, Adam Bess, Chris Alvin, Supratik Mukhopadhyay
Abstract: Computationally predicting protein-protein interactions (PPIs) is challenging due to the lack of integrated, multimodal protein representations. DPEB is a curated collection of 22,043 human proteins that integrates four embedding types: structural (AlphaFold2), transformer-based sequence (BioEmbeddings), contextual amino acid patterns (ESM-2: Evolutionary Scale Modeling), and sequence-based n-gram statistics (ProtVec]). AlphaFold2 protein structures are available through public databases (e.g., AlphaFold2 Protein Structure Database), but the internal neural network embeddings are not. DPEB addresses this gap by providing AlphaFold2-derived embeddings for computational modeling. Our benchmark evaluations show GraphSAGE with BioEmbedding achieved the highest PPI prediction performance (87.37% AUROC, 79.16% accuracy). The framework also achieved 77.42% accuracy for enzyme classification and 86.04% accuracy for protein family classification. DPEB supports multiple graph neural network methods for PPI prediction, enabling applications in systems biology, drug target identification, pathway analysis, and disease mechanism studies.
Authors: Pierangelo Lombardo, Antonio Casoli, Cristian Cingolani, Shola Oshodi, Michele Zanatta
Abstract: Selecting an appropriate evaluation metric for classifiers is crucial for model comparison and parameter optimization, yet there is not consensus on a universally accepted metric that serves as a definitive standard. Moreover, there is often a misconception about the perceived need to mitigate imbalance in datasets used to train classification models. Since the final goal in classifier optimization is typically maximizing the return of investment or, equivalently, minimizing the Total Classification Cost (TCC), we define Weighted Accuracy (WA), an evaluation metric for binary classifiers with a straightforward interpretation as a weighted version of the well-known accuracy metric, coherent with the need of minimizing TCC. We clarify the conceptual framework for handling class imbalance in cost-sensitive scenarios, providing an alternative to rebalancing techniques. This framework can be applied to any metric that, like WA, can be expressed as a linear combination of example-dependent quantities and allows for comparing the results obtained in different datasets and for addressing discrepancies between the development dataset, used to train and validate the model, and the target dataset, where the model will be deployed. It also specifies in which scenarios using UCCs-unaware class rebalancing techniques or rebalancing metrics aligns with TCC minimization and when it is instead counterproductive. Finally, we propose a procedure to estimate the WA weight parameter in the absence of fully specified UCCs and demonstrate the robustness of WA by analyzing its correlation with TCC in example-dependent scenarios.
Authors: Naina Balepur, Xingrui Pei, Hari Sundaram
Abstract: Many governmental bodies are adopting AI policies for decision-making. In particular, Reinforcement Learning has been used to design policies that citizens would be expected to follow if implemented. Much RL work assumes that citizens follow these policies, and evaluate them with this in mind. However, we know from prior work that without institutional trust, citizens will not follow policies put in place by governments. In this work, we develop a trust-aware RL algorithm for resource allocation in communities. We consider the case of humanitarian engineering, where the organization is aiming to distribute some technology or resource to community members. We use a Deep Deterministic Policy Gradient approach to learn a resource allocation that fits the needs of the organization. Then, we simulate resource allocation according to the learned policy, and model the changes in institutional trust of community members. We investigate how this incorporation of institutional trust affects outcomes, and ask how effectively an organization can learn policies if trust values are private. We find that incorporating trust into RL algorithms can lead to more successful policies, specifically when the organization's goals are less certain. We find more conservative trust estimates lead to increased fairness and average community trust, though organization success suffers. Finally, we explore a strategy to prevent unfair outcomes to communities. We implement a quota system by an external entity which decreases the organization's utility when it does not serve enough community members. We find this intervention can improve fairness and trust among communities in some cases, while decreasing the success of the organization. This work underscores the importance of institutional trust in algorithm design and implementation, and identifies a tension between organization success and community well-being.
Authors: Masoud Ataei, Vikas Dhiman, Mohammad Javad Khojasteh
Abstract: Neural networks are parametric and powerful tools for function approximation, and the choice of architecture heavily influences their interpretability, efficiency, and generalization. In contrast, Gaussian processes (GPs) are nonparametric probabilistic models that define distributions over functions using a kernel to capture correlations among data points. However, these models become computationally expensive for large-scale problems, as they require inverting a large covariance matrix. Kolmogorov- Arnold networks (KANs), semi-parametric neural architectures, have emerged as a prominent approach for modeling complex functions with structured and efficient representations through spline layers. Kurkova Kolmogorov-Arnold networks (KKANs) extend this idea by reducing the number of spline layers in KAN and replacing them with Chebyshev layers and multi-layer perceptrons, thereby mapping inputs into higher-dimensional spaces before applying spline-based transformations. Compared to KANs, KKANs perform more stable convergence during training, making them a strong architecture for estimating operators and system modeling in dynamical systems. By enhancing the KKAN architecture, we develop a novel learning algorithm, distance-aware error for Kurkova-Kolmogorov networks (K-DAREK), for efficient and interpretable function approximation with uncertainty quantification. Our approach establishes robust error bounds that are distance-aware; this means they reflect the proximity of a test point to its nearest training points. Through case studies on a safe control task, we demonstrate that K-DAREK is about four times faster and ten times higher computationally efficiency than Ensemble of KANs, 8.6 times more scalable than GP by increasing the data size, and 50% safer than our previous work distance-aware error for Kolmogorov networks (DAREK).
Authors: Nikita Karagodin, Shu Ge, Yury Polyanskiy, Philippe Rigollet
Abstract: We study the effect of normalization schemes on token representations in deep transformers. Modeling their evolution as interacting particles on the sphere, we show that normalization acts as a form of speed regulation. This perspective enables a unified analysis of several schemes -- including Post-LN, Pre-LN, Mix-LN, Peri-LN, nGPT, and LN-Scaling -- revealing how they influence clustering dynamics and representation collapse. Our framework clarifies how different schemes shape token representations across layers and provides a principled basis for comparing them, identifying Peri-LN as a particularly effective choice.
Authors: Yassine Chemingui, Aryan Deshwal, Alan Fern, Thanh Nguyen-Tang, Janardhan Rao Doppa
Abstract: We study the problem of Offline Safe Reinforcement Learning (OSRL), where the goal is to learn a reward-maximizing policy from fixed data under a cumulative cost constraint. We propose a novel OSRL approach that frames the problem as a minimax objective and solves it by combining offline RL with online optimization algorithms. We prove the approximate optimality of this approach when integrated with an approximate offline RL oracle and no-regret online optimization. We also present a practical approximation that can be combined with any offline RL algorithm, eliminating the need for offline policy evaluation. Empirical results on the DSRL benchmark demonstrate that our method reliably enforces safety constraints under stringent cost budgets, while achieving high rewards. The code is available at https://github.com/yassineCh/O3SRL.
Authors: Jincheng Zhou, Mengbo Wang, Anqi He, Yumeng Zhou, Hessam Olya, Murat Kocaoglu, Bruno Ribeiro
Abstract: Causal discovery from observational data is a fundamental task in artificial intelligence, with far-reaching implications for decision-making, predictions, and interventions. Despite significant advances, existing methods can be broadly categorized as constraint-based or score-based approaches. Constraint-based methods offer rigorous causal discovery but are often hindered by small sample sizes, while score-based methods provide flexible optimization but typically forgo explicit conditional independence testing. This work explores a third avenue: developing differentiable $d$-separation scores, obtained through a percolation theory using soft logic. This enables the implementation of a new type of causal discovery method: gradient-based optimization of conditional independence constraints. Empirical evaluations demonstrate the robust performance of our approach in low-sample regimes, surpassing traditional constraint-based and score-based baselines on a real-world dataset. Code and data of the proposed method are publicly available at https://github$.$com/PurdueMINDS/DAGPA.
Authors: Tianxiang Wang, Yingtong Ke, Dhananjay Bhaskar, Smita Krishnaswamy, Alexander Cloninger
Abstract: Single-cell technologies generate high-dimensional point clouds of cells, enabling detailed characterization of complex patient states and treatment responses. Yet each patient is represented by an irregular point cloud rather than a simple vector, making it difficult to directly quantify and compare biological differences between individuals. Nonlinear methods such as kernels and neural networks achieve predictive accuracy but act as black boxes, offering little biological interpretability. To address these limitations, we adapt the Linear Optimal Transport (LOT) framework to this setting, embedding irregular point clouds into a fixed-dimensional Euclidean space while preserving distributional structure. This embedding provides a principled linear representation that preserves optimal transport geometry while enabling downstream analysis. It also forms a registration between any two patients, enabling direct comparison of their cellular distributions. Within this space, LOT enables: (i) \textbf{accurate and interpretable classification} of COVID-19 patient states, where classifier weights map back to specific markers and spatial regions driving predictions; and (ii) \textbf{synthetic data generation} for patient-derived organoids, exploiting the linearity of the LOT embedding. LOT barycenters yield averaged cellular profiles representing combined conditions or samples, supporting drug interaction testing. Together, these results establish LOT as a unified framework that bridges predictive performance, interpretability, and generative modeling. By transforming heterogeneous point clouds into structured embeddings directly traceable to the original data, LOT opens new opportunities for understanding immune variation and treatment effects in high-dimensional biological systems.
Authors: Shahrzad Haddadan, Sara Ahmadian
Abstract: The classic Mallows model is a foundational tool for modeling user preferences. However, it has limitations in capturing real-world scenarios, where users often focus only on a limited set of preferred items and are indifferent to the rest. To address this, extensions such as the top-k Mallows model have been proposed, aligning better with practical applications. In this paper, we address several challenges related to the generalized top-k Mallows model, with a focus on analyzing buyer choices. Our key contributions are: (1) a novel sampling scheme tailored to generalized top-k Mallows models, (2) an efficient algorithm for computing choice probabilities under this model, and (3) an active learning algorithm for estimating the model parameters from observed choice data. These contributions provide new tools for analysis and prediction in critical decision-making scenarios. We present a rigorous mathematical analysis for the performance of our algorithms. Furthermore, through extensive experiments on synthetic data and real-world data, we demonstrate the scalability and accuracy of our proposed methods, and we compare the predictive power of Mallows model for top-k lists compared to the simpler Multinomial Logit model.
Authors: Kijung Jeon, Michael Muehlebach, Molei Tao
Abstract: Sampling from constrained statistical distributions is a fundamental task in various fields including Bayesian statistics, computational chemistry, and statistical physics. This article considers the cases where the constrained distribution is described by an unconstrained density, as well as additional equality and/or inequality constraints, which often make the constraint set nonconvex. Existing methods for nonconvex constraint set $\Sigma \subset \mathbb{R}^d$ defined by equality or inequality constraints commonly rely on costly projection steps. Moreover, they cannot handle equality and inequality constraints simultaneously as each method only specialized in one case. In addition, rigorous and quantitative convergence guarantee is often lacking. In this paper, we introduce Overdamped Langevin with LAnding (OLLA), a new framework that can design overdamped Langevin dynamics accommodating both equality and inequality constraints. The proposed dynamics also deterministically corrects trajectories along the normal direction of the constraint surface, thus obviating the need for explicit projections. We show that, under suitable regularity conditions on the target density and $\Sigma$, OLLA converges exponentially fast in $W_2$ distance to the constrained target density $\rho_\Sigma(x) \propto \exp(-f(x))d\sigma_\Sigma$. Lastly, through experiments, we demonstrate the efficiency of OLLA compared to projection-based constrained Langevin algorithms and their slack variable variants, highlighting its favorable computational cost and reasonable empirical mixing.
Authors: Ana K. Rivera, Anvita Bhagavathula, Alvaro Carbonero, Priya Donti
Abstract: Power flow (PF) calculations are the backbone of real-time grid operations, across workflows such as contingency analysis (where repeated PF evaluations assess grid security under outages) and topology optimization (which involves PF-based searches over combinatorially large action spaces). Running these calculations at operational timescales or across large evaluation spaces remains a major computational bottleneck. Additionally, growing uncertainty in power system operations from the integration of renewables and climate-induced extreme weather also calls for tools that can accurately and efficiently simulate a wide range of scenarios and operating conditions. Machine learning methods offer a potential speedup over traditional solvers, but their performance has not been systematically assessed on benchmarks that capture real-world variability. This paper introduces PF$\Delta$, a benchmark dataset for power flow that captures diverse variations in load, generation, and topology. PF$\Delta$ contains 859,800 solved power flow instances spanning six different bus system sizes, capturing three types of contingency scenarios (N , N -1, and N -2), and including close-to-infeasible cases near steady-state voltage stability limits. We evaluate traditional solvers and GNN-based methods, highlighting key areas where existing approaches struggle, and identifying open problems for future research. Our dataset is available at https://huggingface.co/datasets/pfdelta/pfdelta/tree/main and our code with data generation scripts and model implementations is at https://github.com/MOSSLab-MIT/pfdelta.
URLs: https://huggingface.co/datasets/pfdelta/pfdelta/tree/main, https://github.com/MOSSLab-MIT/pfdelta.
Authors: James Thiering, Tarun Sethupat Radha Krishna, Dylan Zelkin, Ashis Kumer Biswas
Abstract: With the rise of online and virtual learning, monitoring and enhancing student engagement have become an important aspect of effective education. Traditional methods of assessing a student's involvement might not be applicable directly to virtual environments. In this study, we focused on this problem and addressed the need to develop an automated system to detect student engagement levels during online learning. We proposed a novel training method which can discourage a model from leveraging sensitive features like gender for its predictions. The proposed method offers benefits not only in the enforcement of ethical standards, but also to enhance interpretability of the model predictions. We applied an attribute-orthogonal regularization technique to a split-model classifier, which uses multiple transfer learning strategies to achieve effective results in reducing disparity in the distribution of prediction for sensitivity groups from a Pearson correlation coefficient of 0.897 for the unmitigated model, to 0.999 for the mitigated model. The source code for this project is available on https://github.com/ashiskb/elearning-engagement-study .
Authors: Khatoon Khedri, Reza Rawassizadeh, Qifu Wen, Mehdi Hosseinzadeh
Abstract: Graph neural networks (GNNs) are known to operate with high accuracy on learning from graph-structured data, but they suffer from high computational and resource costs. Neural network compression methods are used to reduce the model size while maintaining reasonable accuracy. Two of the common neural network compression techniques include pruning and quantization. In this research, we empirically examine the effects of three pruning methods and three quantization methods on different GNN models, including graph classification tasks, node classification tasks, and link prediction. We conducted all experiments on three graph datasets, including Cora, Proteins, and BBBP. Our findings demonstrate that unstructured fine-grained and global pruning can significantly reduce the model's size(50\%) while maintaining or even improving precision after fine-tuning the pruned model. The evaluation of different quantization methods on GNN shows diverse impacts on accuracy, inference time, and model size across different datasets.
Authors: Matthew Lowery, Zhitong Xu, Da Long, Keyan Chen, Daniel S. Johnson, Yang Bai, Varun Shankar, Shandian Zhe
Abstract: Learning mappings between functional spaces, also known as function-on-function regression, plays a crucial role in functional data analysis and has broad applications, e.g. spatiotemporal forecasting, curve prediction, and climate modeling. Existing approaches, such as functional linear models and neural operators, either fall short of capturing complex nonlinearities or lack reliable uncertainty quantification under noisy, sparse, and irregularly sampled data. To address these issues, we propose Deep Gaussian Processes for Functional Maps (DGPFM). Our method designs a sequence of GP-based linear and nonlinear transformations, leveraging integral transforms of kernels, GP interpolation, and nonlinear activations sampled from GPs. A key insight simplifies implementation: under fixed locations, discrete approximations of kernel integral transforms collapse into direct functional integral transforms, enabling flexible incorporation of various integral transform designs. To achieve scalable probabilistic inference, we use inducing points and whitening transformations to develop a variational learning algorithm. Empirical results on real-world and PDE benchmark datasets demonstrate that the advantage of DGPFM in both predictive performance and uncertainty calibration.
Authors: Himadri S. Pandey, Kai Wang, Gian-Gabriel P. Garcia
Abstract: Restless multi-armed bandits (RMABs) provide a scalable framework for sequential decision-making under uncertainty, but classical formulations assume binary actions and a single global budget. Real-world settings, such as healthcare, often involve multiple interventions with heterogeneous costs and constraints, where such assumptions break down. We introduce a Neural Index Policy (NIP) for multi-action RMABs with heterogeneous budget constraints. Our approach learns to assign budget-aware indices to arm--action pairs using a neural network, and converts them into feasible allocations via a differentiable knapsack layer formulated as an entropy-regularized optimal transport (OT) problem. The resulting model unifies index prediction and constrained optimization in a single end-to-end differentiable framework, enabling gradient-based training directly on decision quality. The network is optimized to align its induced occupancy measure with the theoretical upper bound from a linear programming relaxation, bridging asymptotic RMAB theory with practical learning. Empirically, NIP achieves near-optimal performance within 5% of the oracle occupancy-measure policy while strictly enforcing heterogeneous budgets and scaling to hundreds of arms. This work establishes a general, theoretically grounded, and scalable framework for learning index-based policies in complex resource-constrained environments.
Authors: Luca Caldera, Giacomo Bottacini, Lara Cavinato
Abstract: Generative modeling has emerged as a powerful paradigm for representation learning, but its direct applicability to challenging fields like medical imaging remains limited: mere generation, without task alignment, fails to provide a robust foundation for clinical use. We propose MAGIC-Flow, a conditional multiscale normalizing flow architecture that performs generation and classification within a single modular framework. The model is built as a hierarchy of invertible and differentiable bijections, where the Jacobian determinant factorizes across sub-transformations. We show how this ensures exact likelihood computation and stable optimization, while invertibility enables explicit visualization of sample likelihoods, providing an interpretable lens into the model's reasoning. By conditioning on class labels, MAGIC-Flow supports controllable sample synthesis and principled class-probability estimation, effectively aiding both generative and discriminative objectives. We evaluate MAGIC-Flow against top baselines using metrics for similarity, fidelity, and diversity. Across multiple datasets, it addresses generation and classification under scanner noise, and modality-specific synthesis and identification. Results show MAGIC-Flow creates realistic, diverse samples and improves classification. MAGIC-Flow is an effective strategy for generation and classification in data-limited domains, with direct benefits for privacy-preserving augmentation, robust generalization, and trustworthy medical AI.
Authors: Siyu Zhu, Anastasiya Karpovich, Albert Chen, Jessica Koscheka, Shailesh Jannu, Di Wen, Yuqing Zhu, Rohit Jain, Alborz Geramifard
Abstract: We tackle the challenge of training reliable code-fixing agents in real repositories, where complex builds and shifting dependencies make evaluation unstable. We developed a verifiable pipeline with success defined as post-fix build validation and improved reproducibility across ~1K real issues by pinning dependencies and disabling automatic upgrades. Building on this, we introduced a scalable simplified pipeline for large-scale reinforcement learning (RL). Using this setup, we supervised fine-tuned Qwen3-32B in the full pipeline and applied RL on top of the SFT model in the simplified environment. The SFT model distilled from GPT-4.1 trajectories performs on par while being 56x smaller, and RL added 7-20% absolute gains under matched train-test conditions. "Thinking mode" was on par or worse in our experiments. Both SFT and RL models failed to generalize across environments, highlighting the importance of matching train-test environments for building reliable real-world code-fixing agents.
Authors: Thomas Bailie, S. Karthik Mukkavilli, Varvara Vetrova, Yun Sing Koh
Abstract: Climate events arise from intricate, multivariate dynamics governed by global-scale drivers, profoundly impacting food, energy, and infrastructure. Yet, accurate weather prediction remains elusive due to physical processes unfolding across diverse spatio-temporal scales, which fixed-resolution methods cannot capture. Hierarchical Graph Neural Networks (HGNNs) offer a multiscale representation, but nonlinear downward mappings often erase global trends, weakening the integration of physics into forecasts. We introduce HiFlowCast and its ensemble variant HiAntFlow, HGNNs that embed physics within a multiscale prediction framework. Two innovations underpin their design: a Latent-Memory-Retention mechanism that preserves global trends during downward traversal, and a Latent-to-Physics branch that integrates PDE solution fields across diverse scales. Our Flow models cut errors by over 5% at 13-day lead times and by 5-8% under 1st and 99th quantile extremes, improving reliability for rare events. Leveraging pretrained model weights, they converge within a single epoch, reducing training cost and their carbon footprint. Such efficiency is vital as the growing scale of machine learning challenges sustainability and limits research accessibility. Code and model weights are in the supplementary materials.
Authors: Su Liu, Xin Hu, Shurong Wen, Jiaqi Liu, Jiexi Xu, Lanruo Wang
Abstract: Physiologically Based Pharmacokinetic (PBPK) modeling plays a critical role in drug development by predicting drug concentration dynamics across organs. Traditional approaches rely on ordinary differential equations with strong simplifying assumptions, which limit their adaptability to nonlinear physiological interactions. In this study, we explore data-driven alternatives for PBPK prediction using deep learning. Two baseline architectures - a multilayer perceptron (MLP) and a long short-term memory (LSTM) network - are implemented to capture molecular and temporal dependencies, respectively. To incorporate inter-organ interactions, we propose a Dynamic Graph Neural Network (Dynamic GNN) that models physiological connections as recurrent message-passing processes between organs. Experimental results demonstrate that the proposed Dynamic GNN achieves the highest predictive performance among all models, with an R^2 of 0.9342, an RMSE of 0.0159, and an MAE of 0.0116. In comparison, the MLP baseline obtains an R^2 of 0.8705 and the LSTM achieves 0.8059. These results highlight that explicitly modeling the spatial and temporal dependencies of organ interactions enables more accurate and generalizable drug concentration prediction. The Dynamic GNN provides a scalable, equation-free alternative to traditional PBPK formulations and demonstrates strong potential for data-driven pharmacokinetic modeling in preclinical and clinical research.
Authors: Xixian Liu, Rui Jiao, Zhiyuan Liu, Yurou Liu, Yang Liu, Ziheng Lu, Wenbing Huang, Yang Zhang, Yixin Cao
Abstract: Coordinate denoising has emerged as a promising method for 3D molecular pretraining due to its theoretical connection to learning molecular force field. However, existing denoising methods rely on oversimplied molecular dynamics that assume atomic motions to be isotropic and homoscedastic. To address these limitations, we propose a novel denoising framework AniDS: Anisotropic Variational Autoencoder for 3D Molecular Denoising. AniDS introduces a structure-aware anisotropic noise generator that can produce atom-specific, full covariance matrices for Gaussian noise distributions to better reflect directional and structural variability in molecular systems. These covariances are derived from pairwise atomic interactions as anisotropic corrections to an isotropic base. Our design ensures that the resulting covariance matrices are symmetric, positive semi-definite, and SO(3)-equivariant, while providing greater capacity to model complex molecular dynamics. Extensive experiments show that AniDS outperforms prior isotropic and homoscedastic denoising models and other leading methods on the MD17 and OC22 benchmarks, achieving average relative improvements of 8.9% and 6.2% in force prediction accuracy. Our case study on a crystal and molecule structure shows that AniDS adaptively suppresses noise along the bonding direction, consistent with physicochemical principles. Our code is available at https://github.com/ZeroKnighting/AniDS.
Authors: Shiji Zhou (Institute of Artificial Intelligence, Beihang University, Beijing Advanced Innovation Center for Future Blockchain,Privacy Computing, Beihang University), Tianbai Yu (University of Illinois at Urbana-Champaign), Zhi Zhang (University of Amsterdam), Heng Chang (Tsinghua University), Xiao Zhou (Tsinghua University), Dong Wu (YanTron Technology Co. Ltd), Han Zhao (University of Illinois at Urbana-Champaign)
Abstract: Machine unlearning (MU) aims to efficiently remove sensitive or harmful memory from a pre-trained model. The key challenge is to balance the potential tradeoff between unlearning efficacy and utility preservation, which involves forgetting undesirable information as defined while maintaining the model's original performance. One potential way to tackle this problem is to use multi-objective optimization to jointly optimize both the unlearning and utility preservation objectives. However, existing multi-objective methods only guarantee finding a Pareto-optimal solution without fine-grained control, which causes under-optimization of the unlearning objective. To this end, we first model MU as a constrained optimization problem, that is, optimizing the unlearning objective under the constraint of a bounded increase for utility loss. We then show that solving this optimization problem is equivalent to unilateral gradient surgery on the unlearning objective. To resolve the additional computational cost brought by gradient surgery, we propose an implicit gradient surgery method, which approximates the solution to the aforementioned constrained optimization problem via only one backpropagation, thereby achieving efficient utility-preserving MU. Theoretically, we provide a tight convergence analysis of the algorithm. Empirically, our extensive experiments show that the proposed algorithm achieves better tradeoff results than existing baselines. Codes are available at https://github.com/anseryuer/EUPMU-Efficient-Utility-Preserving-Machine-Unlearning.
URLs: https://github.com/anseryuer/EUPMU-Efficient-Utility-Preserving-Machine-Unlearning.
Authors: Zhiqin Zhang, Yining Ma, Zhiguang Cao, Hoong Chuin Lau
Abstract: Neural combinatorial optimization (NCO) has achieved remarkable performance, yet its learned model representations and decision rationale remain a black box. This impedes both academic research and practical deployment, since researchers and stakeholders require deeper insights into NCO models. In this paper, we take the first critical step towards interpreting NCO models by investigating their representations through various probing tasks. Moreover, we introduce a novel probing tool named Coefficient Significance Probing (CS-Probing) to enable deeper analysis of NCO representations by examining the coefficients and statistical significance during probing. Extensive experiments and analysis reveal that NCO models encode low-level information essential for solution construction, while capturing high-level knowledge to facilitate better decisions. Using CS-Probing, we find that prevalent NCO models impose varying inductive biases on their learned representations, uncover direct evidence related to model generalization, and identify key embedding dimensions associated with specific knowledge. These insights can be potentially translated into practice, for example, with minor code modifications, we improve the generalization of the analyzed model. Our work represents a first systematic attempt to interpret black-box NCO models, showcasing probing as a promising tool for analyzing their internal mechanisms and revealing insights for the NCO community. The source code is publicly available.
Authors: Farzaneh Heidari, Chao Li, Farzaneh Heidari
Abstract: We show how to replace the O(2^n) coalition enumeration over n features behind Shapley values and Shapley-style interaction indices with a few-evaluation scheme on a tensor-network (TN) surrogate: TN-SHAP. The key idea is to represent a predictor's local behavior as a factorized multilinear map, so that coalitional quantities become linear probes of a coefficient tensor. TN-SHAP replaces exhaustive coalition sweeps with just a small number of targeted evaluations to extract order-k Shapley interactions. In particular, both order-1 (single-feature) and order-2 (pairwise) computations have cost O(n*poly(chi) + n^2), where chi is the TN's maximal cut rank. We provide theoretical guarantees on the approximation error and tractability of TN-SHAP. On UCI datasets, our method matches enumeration on the fitted surrogate while reducing evaluation by orders of magnitude and achieves 25-1000x wall-clock speedups over KernelSHAP-IQ at comparable accuracy, while amortizing training across local cohorts.
Authors: Jinzhe Liu, Junshu Sun, Shufan Shen, Chenxue Yang, Shuhui Wang
Abstract: Lifelong knowledge editing enables continuous, precise updates to outdated knowledge in large language models (LLMs) without computationally expensive full retraining. However, existing methods often accumulate errors throughout the editing process, causing a gradual decline in both editing accuracy and generalization. To tackle this problem, we propose Neuron-Specific Masked Knowledge Editing (NMKE), a novel fine-grained editing framework that combines neuron-level attribution with dynamic sparse masking. Leveraging neuron functional attribution, we identify two key types of knowledge neurons, with knowledge-general neurons activating consistently across prompts and knowledge-specific neurons activating to specific prompts. NMKE further introduces an entropy-guided dynamic sparse mask, locating relevant neurons to the target knowledge. This strategy enables precise neuron-level knowledge editing with fewer parameter modifications. Experimental results from thousands of sequential edits demonstrate that NMKE outperforms existing methods in maintaining high editing success rates and preserving model general capabilities in lifelong editing.
Authors: Mohammadsajad Alipour, Mohammad Mohammadi Amiri
Abstract: Federated learning (FL) has emerged as a promising paradigm for decentralized model training, enabling multiple clients to collaboratively learn a shared model without exchanging their local data. However, the decentralized nature of FL also introduces vulnerabilities, as malicious clients can compromise or manipulate the training process. In this work, we introduce dictator clients, a novel, well-defined, and analytically tractable class of malicious participants capable of entirely erasing the contributions of all other clients from the server model, while preserving their own. We propose concrete attack strategies that empower such clients and systematically analyze their effects on the learning process. Furthermore, we explore complex scenarios involving multiple dictator clients, including cases where they collaborate, act independently, or form an alliance in order to ultimately betray one another. For each of these settings, we provide a theoretical analysis of their impact on the global model's convergence. Our theoretical algorithms and findings about the complex scenarios including multiple dictator clients are further supported by empirical evaluations on both computer vision and natural language processing benchmarks.
Authors: Lorenzo Magnino, Kai Shao, Zida Wu, Jiacheng Shen, Mathieu Lauri\`ere
Abstract: Mean field games (MFGs) have emerged as a powerful framework for modeling interactions in large-scale multi-agent systems. Despite recent advancements in reinforcement learning (RL) for MFGs, existing methods are typically limited to finite spaces or stationary models, hindering their applicability to real-world problems. This paper introduces a novel deep reinforcement learning (DRL) algorithm specifically designed for non-stationary continuous MFGs. The proposed approach builds upon a Fictitious Play (FP) methodology, leveraging DRL for best-response computation and supervised learning for average policy representation. Furthermore, it learns a representation of the time-dependent population distribution using a Conditional Normalizing Flow. To validate the effectiveness of our method, we evaluate it on three different examples of increasing complexity. By addressing critical limitations in scalability and density approximation, this work represents a significant advancement in applying DRL techniques to complex MFG problems, bringing the field closer to real-world multi-agent systems.
Authors: Nadav Dym, Matthias Wellershoff, Efstratios Tsoukanis, Daniel Levy, Radu Balan
Abstract: We study the sorting-based embedding $\beta_{\mathbf A} : \mathbb R^{n \times d} \to \mathbb R^{n \times D}$, $\mathbf X \mapsto {\downarrow}(\mathbf X \mathbf A)$, where $\downarrow$ denotes column wise sorting of matrices. Such embeddings arise in graph deep learning where outputs should be invariant to permutations of graph nodes. Previous work showed that for large enough $D$ and appropriate $\mathbf A$, the mapping $\beta_{\mathbf A}$ is injective, and moreover satisfies a bi-Lipschitz condition. However, two gaps remain: firstly, the optimal size $D$ required for injectivity is not yet known, and secondly, no estimates of the bi-Lipschitz constants of the mapping are known. In this paper, we make substantial progress in addressing both of these gaps. Regarding the first gap, we improve upon the best known upper bounds for the embedding dimension $D$ necessary for injectivity, and also provide a lower bound on the minimal injectivity dimension. Regarding the second gap, we construct matrices $\mathbf A$, so that the bi-Lipschitz distortion of $\beta_{\mathbf A} $ depends quadratically on $n$, and is completely independent of $d$. We also show that the distortion of $\beta_{\mathbf A}$ is necessarily at least in $\Omega(\sqrt{n})$. Finally, we provide similar results for variants of $\beta_{\mathbf A}$ obtained by applying linear projections to reduce the output dimension of $\beta_{\mathbf A}$.
Authors: Qingzhu Zhang, Jiani Zhong, Zongsheng Li, Xinke Shen, Quanying Liu
Abstract: Task-specific pre-training is essential when task representations diverge from generic pre-training features. Existing task-general pre-training EEG models struggle with complex tasks like emotion recognition due to mismatches between task-specific features and broad pre-training approaches. This work aims to develop a task-specific multi-dataset joint pre-training framework for cross-dataset emotion recognition, tackling problems of large inter-dataset distribution shifts, inconsistent emotion category definitions, and substantial inter-subject variability. We introduce a cross-dataset covariance alignment loss to align second-order statistical properties across datasets, enabling robust generalization without the need for extensive labels or per-subject calibration. To capture the long-term dependency and complex dynamics of EEG, we propose a hybrid encoder combining a Mamba-like linear attention channel encoder and a spatiotemporal dynamics model. Our method outperforms state-of-the-art large-scale EEG models by an average of 4.57% in AUROC for few-shot emotion recognition and 11.92% in accuracy for zero-shot generalization to a new dataset. Performance scales with the increase of datasets used in pre-training. Multi-dataset joint pre-training achieves a performance gain of 8.55% over single-dataset training. This work provides a scalable framework for task-specific pre-training and highlights its benefit in generalizable affective computing. Our code is available at https://github.com/ncclab-sustech/mdJPT_nips2025.
Authors: Nnamdi Aghanya, Jun Li, Kewei Wang
Abstract: Large Language Models (LLMs) can achieve near-optimal lossless compression by acting as powerful probability models. We investigate their use in the lossy domain, where reconstruction fidelity is traded for higher compression ratios. This paper introduces Error-Bounded Predictive Coding (EPC), a lossy text codec that leverages a Masked Language Model (MLM) as a decompressor. Instead of storing a subset of original tokens, EPC allows the model to predict masked content and stores minimal, rank-based corrections only when the model's top prediction is incorrect. This creates a residual channel that offers continuous rate-distortion control. We compare EPC to a simpler Predictive Masking (PM) baseline and a transform-based Vector Quantisation with a Residual Patch (VQ+RE) approach. Through an evaluation that includes precise bit accounting and rate-distortion analysis, we demonstrate that EPC consistently dominates PM, offering superior fidelity at a significantly lower bit rate by more efficiently utilising the model's intrinsic knowledge.
Authors: Siddharth Jain, Shyamgopal Karthik, Vineet Gandhi
Abstract: Pretrained models are ubiquitous in the current deep learning landscape, offering strong results on a broad range of tasks. Recent works have shown that models differing in various design choices exhibit categorically diverse generalization behavior, resulting in one model grasping distinct data-specific insights unavailable to the other. In this paper, we propose to leverage large publicly available model repositories as an auxiliary source of model improvements. We introduce a data partitioning strategy where pretrained models autonomously adopt either the role of a student, seeking knowledge, or that of a teacher, imparting knowledge. Experiments across various tasks demonstrate the effectiveness of our proposed approach. In image classification, we improved the performance of ViT-B by approximately 1.4% through bidirectional knowledge transfer with ViT-T. For semantic segmentation, our method boosted all evaluation metrics by enabling knowledge transfer both within and across backbone architectures. In video saliency prediction, our approach achieved a new state-of-the-art. We further extend our approach to knowledge transfer between multiple models, leading to considerable performance improvements for all model participants.
Authors: Sofoklis Kitharidis, Cor J. Veenman, Thomas B\"ack, Niki van Stein
Abstract: In the context of algorithmic decision-making, fair machine learning methods often yield multiple models that balance predictive fairness and performance in varying degrees. This diversity introduces a challenge for stakeholders who must select a model that aligns with their specific requirements and values. To address this, we propose an interactive framework that assists in navigating and interpreting the trade-offs across a portfolio of models. Our approach leverages weakly supervised metric learning to learn a Mahalanobis distance that reflects similarity in fairness and performance outcomes, effectively structuring the feature importance space of the models according to stakeholder-relevant criteria. We then apply clustering technique (k-means) to group models based on their transformed representations of feature importances, allowing users to explore clusters of models with similar predictive behaviors and fairness characteristics. This facilitates informed decision-making by helping users understand how models differ not only in their fairness-performance balance but also in the features that drive their predictions.
Authors: Keyu Wang, Tian Lyu, Guinan Su, Jonas Geiping, Lu Yin, Marco Canini, Shiwei Liu
Abstract: Layer pruning has emerged as a widely adopted technique for improving the efficiency of large language models (LLMs). Although existing methods demonstrate strong performance retention on general knowledge tasks, their effect on long-chain reasoning, a more brittle yet crucial capability, remains largely unexplored. In this work, we study the impact of layer pruning on long-chain reasoning through the lens of test-time scaling, a key mechanism in modern LLMs that enables strong reasoning capacity by allocating more computation at inference time. With extensive experiments, we demonstrate that pruning even one or two layers can severely impair test-time scaling, with performance collapsing drastically on long reasoning benchmarks even when performance on knowledge-intensive and shallow reasoning tasks remains stable. Furthermore, we find that standard supervised fine-tuning remedies fail to recover test-time scaling once it has deteriorated. Through in-depth analyses, we identify the mechanisms underlying this fragility of test-time scaling and highlight the fundamental risks of applying layer pruning to reasoning-intensive LLMs. These findings call for a rethinking of layer pruning strategies and provide insights for developing methods that preserve the robustness of reasoning. We open-source the codebase in \href{https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling}{https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling}.
URLs: https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling, https://github.com/keyu-wang-2002/Layer-Pruning-Harms-Inference-Scaling
Authors: Berkay D\"oner, Thorir Mar Ingolfsson, Luca Benini, Yawei Li
Abstract: Electroencephalography (EEG) offers a non-invasive lens into human brain activity, but building large-scale models is hampered by topological heterogeneity: each public EEG data defines its own electrode layout, limiting generalization. We introduce LUNA (Latent Unified Network Architecture), a self-supervised foundation model that reconciles disparate electrode geometries while scaling linearly -- not quadratically -- with channel count. LUNA compresses multi-channel EEG into a fixed-size, topology-agnostic latent space via learned queries and cross-attention. Downstream transformer blocks then operate exclusively on this latent representation using patch-wise temporal self-attention, decoupling computation from electrode count. Pre-trained on TUEG and Siena (over 21,000 hours of raw EEG across diverse montages) using a masked-patch reconstruction objective, LUNA transfers effectively to four downstream tasks: abnormality detection, artifact rejection, slowing classification, and emotion recognition. It demonstrates highly competitive performance across several benchmarks, achieving state-of-the-art results on TUAR and TUSL, e.g., 0.921 AUROC on TUAR, while reducing FLOPs by 300x and trimming GPU memory use by up to 10x. Critically, these gains are consistent across all evaluated electrode configurations. Code is available at https://github.com/pulp-bio/BioFoundation
Authors: Shireen Kudukkil Manchingal
Abstract: Machine learning has achieved remarkable successes, yet its deployment in safety-critical domains remains hindered by an inherent inability to manage uncertainty, resulting in overconfident and unreliable predictions when models encounter out-of-distribution data, adversarial perturbations, or naturally fluctuating environments. This thesis, titled Epistemic Deep Learning: Enabling Machine Learning Models to 'Know When They Do Not Know', addresses these critical challenges by advancing the paradigm of Epistemic Artificial Intelligence, which explicitly models and quantifies epistemic uncertainty: the uncertainty arising from limited, biased, or incomplete training data, as opposed to the irreducible randomness of aleatoric uncertainty, thereby empowering models to acknowledge their limitations and refrain from overconfident decisions when uncertainty is high. Central to this work is the development of the Random-Set Neural Network (RS-NN), a novel methodology that leverages random set theory to predict belief functions over sets of classes, capturing the extent of epistemic uncertainty through the width of associated credal sets, applications of RS-NN, including its adaptation to Large Language Models (LLMs) and its deployment in weather classification for autonomous racing. In addition, the thesis proposes a unified evaluation framework for uncertainty-aware classifiers. Extensive experiments validate that integrating epistemic awareness into deep learning not only mitigates the risks associated with overconfident predictions but also lays the foundation for a paradigm shift in artificial intelligence, where the ability to 'know when it does not know' becomes a hallmark of robust and dependable systems. The title encapsulates the core philosophy of this work, emphasizing that true intelligence involves recognizing and managing the limits of one's own knowledge.
Authors: Rodrigo Tertulino, Ricardo Almeida
Abstract: Identifying the factors that influence student performance in basic education is a central challenge for formulating effective public policies in Brazil. This study introduces a multi-level machine learning approach to classify the proficiency of 9th-grade and high school students using microdata from the System of Assessment of Basic Education (SAEB). Our model uniquely integrates four data sources: student socioeconomic characteristics, teacher professional profiles, school indicators, and director management profiles. A comparative analysis of four ensemble algorithms confirmed the superiority of a Random Forest model, which achieved 90.2% accuracy and an Area Under the Curve (AUC) of 96.7%. To move beyond prediction, we applied Explainable AI (XAI) using SHAP, which revealed that the school's average socioeconomic level is the most dominant predictor, demonstrating that systemic factors have a greater impact than individual characteristics in isolation. The primary conclusion is that academic performance is a systemic phenomenon deeply tied to the school's ecosystem. This study provides a data-driven, interpretable tool to inform policies aimed at promoting educational equity by addressing disparities between schools.
Authors: Laxmi pant, Syed Ali Reza, Md Khalilor Rahman, MD Saifur Rahman, Shamima Sharmin, Md Fazlul Huq Mithu, Kazi Nehal Hasnain, Adnan Farabi, Mahamuda khanom, Raisul Kabir
Abstract: The growing instability of both global and domestic economic environments has increased the risk of financial distress at the household level. However, traditional econometric models often rely on delayed and aggregated data, limiting their effectiveness. This study introduces a machine learning-based early warning system that utilizes real-time digital and macroeconomic signals to identify financial distress in near real-time. Using a panel dataset of 750 households tracked over three monitoring rounds spanning 13 months, the framework combines socioeconomic attributes, macroeconomic indicators (such as GDP growth, inflation, and foreign exchange fluctuations), and digital economy measures (including ICT demand and market volatility). Through data preprocessing and feature engineering, we introduce lagged variables, volatility measures, and interaction terms to capture both gradual and sudden changes in financial stability. We benchmark baseline classifiers, such as logistic regression and decision trees, against advanced ensemble models including random forests, XGBoost, and LightGBM. Our results indicate that the engineered features from the digital economy significantly enhance predictive accuracy. The system performs reliably for both binary distress detection and multi-class severity classification, with SHAP-based explanations identifying inflation volatility and ICT demand as key predictors. Crucially, the framework is designed for scalable deployment in national agencies and low-bandwidth regional offices, ensuring it is accessible for policymakers and practitioners. By implementing machine learning in a transparent and interpretable manner, this study demonstrates the feasibility and impact of providing near-real-time early warnings of financial distress. This offers actionable insights that can strengthen household resilience and guide preemptive intervention strategies.
Authors: Yan Jiang, Ruihong Qiu, Zi Huang
Abstract: Homophily, the tendency of nodes from the same class to connect, is a fundamental property of real-world graphs, underpinning structural and semantic patterns in domains such as citation networks and social networks. Existing methods exploit homophily through designing homophily-aware GNN architectures or graph structure learning strategies, yet they primarily focus on GNN learning with training graphs. However, in real-world scenarios, test graphs often suffer from data quality issues and distribution shifts, such as domain shifts across users from different regions in social networks and temporal evolution shifts in citation network graphs collected over varying time periods. These factors significantly compromise the pre-trained model's robustness, resulting in degraded test-time performance. With empirical observations and theoretical analysis, we reveal that transforming the test graph structure by increasing homophily in homophilic graphs or decreasing it in heterophilic graphs can significantly improve the robustness and performance of pre-trained GNNs on node classifications, without requiring model training or update. Motivated by these insights, a novel test-time graph structural transformation method grounded in homophily, named GrapHoST, is proposed. Specifically, a homophily predictor is developed to discriminate test edges, facilitating adaptive test-time graph structural transformation by the confidence of predicted homophily scores. Extensive experiments on nine benchmark datasets under a range of test-time data quality issues demonstrate that GrapHoST consistently achieves state-of-the-art performance, with improvements of up to 10.92%. Our code has been released at https://github.com/YanJiangJerry/GrapHoST.
Authors: Mary E. An, Paul Griffin, Jonathan G. Stine, Ramakrishna Balakrishnan, Ram Sriram, Soundar Kumara
Abstract: Background: Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) affects ~33% of U.S. adults and is the most common chronic liver disease. Although often asymptomatic, progression can lead to cirrhosis. Early detection is important, as lifestyle interventions can prevent disease progression. We developed a fair, rigorous, and reproducible MASLD prediction model and compared it to prior methods using a large electronic health record database. Methods: We evaluated LASSO logistic regression, random forest, XGBoost, and a neural network for MASLD prediction using clinical feature subsets, including the top 10 SHAP-ranked features. To reduce disparities in true positive rates across racial and ethnic subgroups, we applied an equal opportunity postprocessing method. Results: This study included 59,492 patients in the training data, 24,198 in the validating data, and 25,188 in the testing data. The LASSO logistic regression model with the top 10 features was selected for its interpretability and comparable performance. Before fairness adjustment, the model achieved AUROC of 0.84, accuracy of 78%, sensitivity of 72%, specificity of 79%, and F1-score of 0.617. After equal opportunity postprocessing, accuracy modestly increased to 81% and specificity to 94%, while sensitivity decreased to 41% and F1-score to 0.515, reflecting the fairness trade-off. Conclusions: We developed the MASER prediction model (MASLD Static EHR Risk Prediction), a LASSO logistic regression model which achieved competitive performance for MASLD prediction (AUROC 0.836, accuracy 77.6%), comparable to previously reported ensemble and tree-based models. Overall, this approach demonstrates that interpretable models can achieve a balance of predictive performance and fairness in diverse patient populations.
Authors: Yujie Xiao, Gongzhen Tang, Wenhui Liu, Jun Li, Guangkun Nie, Zhuoran Kan, Deyun Zhang, Qinghao Zhao, Shenda Hong
Abstract: Timely access to laboratory values is critical for clinical decision-making, yet current approaches rely on invasive venous sampling and are intrinsically delayed. Electrocardiography (ECG), as a non-invasive and widely available signal, offers a promising modality for rapid laboratory estimation. Recent progress in deep learning has enabled the extraction of latent hematological signatures from ECGs. However, existing models are constrained by low signal-to-noise ratios, substantial inter-individual variability, limited data diversity, and suboptimal generalization, especially when adapted to low-lead wearable devices. In this work, we conduct an exploratory study leveraging transfer learning to fine-tune ECGFounder, a large-scale pre-trained ECG foundation model, on the Multimodal Clinical Monitoring in the Emergency Department (MC-MED) dataset from Stanford. We generated a corpus of more than 20 million standardized ten-second ECG segments to enhance sensitivity to subtle biochemical correlates. On internal validation, the model demonstrated strong predictive performance (area under the curve above 0.65) for thirty-three laboratory indicators, moderate performance (between 0.55 and 0.65) for fifty-nine indicators, and limited performance (below 0.55) for sixteen indicators. This study provides an efficient artificial-intelligence driven solution and establishes the feasibility scope for real-time, non-invasive estimation of laboratory values.
Authors: Hongyu Guo
Abstract: Analogical reasoning, the transfer of relational structures across contexts (e.g., planet is to sun as electron is to nucleus), is fundamental to scientific discovery. Yet human insight is often constrained by domain expertise and surface-level biases, limiting access to deeper, structure-driven analogies both within and across disciplines. Large language models (LLMs), trained on vast cross-domain data, present a promising yet underexplored tool for analogical reasoning in science. Here, we demonstrate that LLMs can generate novel battery materials by (1) retrieving cross-domain analogs and analogy-guided exemplars to steer exploration beyond conventional dopant substitutions, and (2) constructing in-domain analogical templates from few labeled examples to guide targeted exploitation. These explicit analogical reasoning strategies yield candidates outside established compositional spaces and outperform standard prompting baselines. Our findings position LLMs as interpretable, expert-like hypothesis generators that leverage analogy-driven generalization for scientific innovation.
Authors: Kumar Saurav, Ness B. Shroff, Yingbin Liang
Abstract: We consider a node-monitor pair, where the node's state varies with time. The monitor needs to track the node's state at all times; however, there is a fixed cost for each state query. So the monitor may instead predict the state using time-series forecasting methods, including time-series foundation models (TSFMs), and query only when prediction uncertainty is high. Since query decisions influence prediction accuracy, determining when to query is nontrivial. A natural approach is a greedy policy that predicts when the expected prediction loss is below the query cost and queries otherwise. We analyze this policy in a Markovian setting, where the optimal (OPT) strategy is a state-dependent threshold policy minimizing the time-averaged sum of query cost and prediction losses. We show that, in general, the greedy policy is suboptimal and can have an unbounded competitive ratio, but under common conditions such as identically distributed transition probabilities, it performs close to OPT. For the case of unknown transition probabilities, we further propose a projected stochastic gradient descent (PSGD)-based learning variant of the greedy policy, which achieves a favorable predict-query tradeoff with improved computational efficiency compared to OPT.
Authors: Mengyu Ye, Jun Suzuki, Tatsuro Inaba, Tatsuki Kuribayashi
Abstract: Recent interpretability work on large language models (LLMs) has been increasingly dominated by a feature-discovery approach with the help of proxy modules. Then, the quality of features learned by, e.g., sparse auto-encoders (SAEs), is evaluated. This paradigm naturally raises a critical question: do such learned features have better properties than those already represented within the original model parameters, and unfortunately, only a few studies have made such comparisons systematically so far. In this work, we revisit the interpretability of feature vectors stored in feed-forward (FF) layers, given the perspective of FF as key-value memories, with modern interpretability benchmarks. Our extensive evaluation revealed that SAE and FFs exhibits a similar range of interpretability, although SAEs displayed an observable but minimal improvement in some aspects. Furthermore, in certain aspects, surprisingly, even vanilla FFs yielded better interpretability than the SAEs, and features discovered in SAEs and FFs diverged. These bring questions about the advantage of SAEs from both perspectives of feature quality and faithfulness, compared to directly interpreting FF feature vectors, and FF key-value parameters serve as a strong baseline in modern interpretability research.
Authors: David Anton, Henning Wessels, Ulrich R\"omer, Alexander Henkes, Jorge-Humberto Urrea-Quintero
Abstract: Constitutive model discovery refers to the task of identifying an appropriate model structure, usually from a predefined model library, while simultaneously inferring its material parameters. The data used for model discovery are measured in mechanical tests and are thus inevitably affected by noise which, in turn, induces uncertainties. Previously proposed methods for uncertainty quantification in model discovery either require the selection of a prior for the material parameters, are restricted to the linear coefficients of the model library or are limited in the flexibility of the inferred parameter probability distribution. We therefore propose a four-step partially Bayesian framework for uncertainty quantification in model discovery that does not require prior selection for the material parameters and also allows for the discovery of non-linear constitutive models: First, we augment the available stress-deformation data with a Gaussian process. Second, we approximate the parameter distribution by a normalizing flow, which allows for capturing complex joint distributions. Third, we distill the parameter distribution by matching the distribution of stress-deformation functions induced by the parameters with the Gaussian process posterior. Fourth, we perform a Sobol' sensitivity analysis to obtain a sparse and interpretable model. We demonstrate the capability of our framework for both isotropic and anisotropic experimental data as well as linear and non-linear model libraries.
Authors: Jiazheng Li, Andreas Damianou, J Rosser, Jos\'e Luis Redondo Garc\'ia, Konstantina Palla
Abstract: Chain-of-thought (CoT) traces promise transparency for reasoning language models, but prior work shows they are not always faithful reflections of internal computation. This raises challenges for oversight: practitioners may misinterpret decorative reasoning as genuine. We introduce Concept Walk, a general framework for tracing how a model's internal stance evolves with respect to a concept direction during reasoning. Unlike surface text, Concept Walk operates in activation space, projecting each reasoning step onto the concept direction learned from contrastive data. This allows us to observe whether reasoning traces shape outcomes or are discarded. As a case study, we apply Concept Walk to the domain of Safety using Qwen 3-4B. We find that in 'easy' cases, perturbed CoTs are quickly ignored, indicating decorative reasoning, whereas in 'hard' cases, perturbations induce sustained shifts in internal activations, consistent with faithful reasoning. The contribution is methodological: Concept Walk provides a lens to re-examine faithfulness through concept-specific internal dynamics, helping identify when reasoning traces can be trusted and when they risk misleading practitioners.
Authors: Jan Simson, Alessandro Fabris, Cosima Fr\"ohner, Frauke Kreuter, Christoph Kern
Abstract: As machine learning (ML) systems are increasingly adopted in high-stakes decision-making domains, ensuring fairness in their outputs has become a central challenge. At the core of fair ML research are the datasets used to investigate bias and develop mitigation strategies. Yet, much of the existing work relies on a narrow selection of datasets--often arbitrarily chosen, inconsistently processed, and lacking in diversity--undermining the generalizability and reproducibility of results. To address these limitations, we present FairGround: a unified framework, data corpus, and Python package aimed at advancing reproducible research and critical data studies in fair ML classification. FairGround currently comprises 44 tabular datasets, each annotated with rich fairness-relevant metadata. Our accompanying Python package standardizes dataset loading, preprocessing, transformation, and splitting, streamlining experimental workflows. By providing a diverse and well-documented dataset corpus along with robust tooling, FairGround enables the development of fairer, more reliable, and more reproducible ML models. All resources are publicly available to support open and collaborative research.
Authors: Zirui Pang, Hao Zheng, Zhijie Deng, Ling Li, Zixin Zhong, Jiaheng Wei
Abstract: LLM unlearning has emerged as a promising approach, aiming to enable models to forget hazardous/undesired knowledge at low cost while preserving as much model utility as possible. Among existing techniques, the most straightforward method is performing Gradient Ascent (GA) w.r.t. the forget data, thereby forcing the model to unlearn the forget dataset. However, GA suffers from severe instability, as it drives updates in a divergent direction, often resulting in drastically degraded model utility. To address this issue, we propose Smoothed Gradient Ascent (SGA). SGA combines the forget data with multiple constructed normal data through a tunable smoothing rate. Intuitively, this extends GA from learning solely on the forget data to jointly learning across both forget and normal data, enabling more stable unlearning while better preserving model utility. Theoretically, we provide the theoretical guidance on the selection of the optimal smoothing rate. Empirically, we evaluate SGA on three benchmarks: TOFU, Harry Potter, and MUSE-NEWS. Experimental results demonstrate that SGA consistently outperforms the original Gradient Ascent (GA) method across all metrics and achieves top-2 performance among all baseline methods on several key metrics.
Authors: David Freire-Obreg\'on, Jos\'e Salas-C\'aceres, Modesto Castrill\'on-Santana
Abstract: Regularization techniques play a crucial role in preventing overfitting and improving the generalization performance of neural networks. Dropout, a widely used regularization technique, randomly deactivates units during training to introduce redundancy and prevent co-adaptation among neurons. Despite its effectiveness, dropout has limitations, such as its static nature and lack of interpretability. In this paper, we propose a novel approach to regularization by substituting dropout with Conway's Game of Life (GoL), a cellular automata with simple rules that govern the evolution of a grid of cells. We introduce dynamic unit deactivation during training by representing neural network units as cells in a GoL grid and applying the game's rules to deactivate units. This approach allows for the emergence of spatial patterns that adapt to the training data, potentially enhancing the network's ability to generalize. We demonstrate the effectiveness of our approach on the CIFAR-10 dataset, showing that dynamic unit deactivation using GoL achieves comparable performance to traditional dropout techniques while offering insights into the network's behavior through the visualization of evolving patterns. Furthermore, our discussion highlights the applicability of our proposal in deeper architectures, demonstrating how it enhances the performance of different dropout techniques.
Authors: Yasas Senarath, Hemant Purohit
Abstract: User behavior on online platforms is evolving, reflecting real-world changes in how people post, whether it's helpful messages or hate speech. Models that learn to capture this content can experience a decrease in performance over time due to data drift, which can lead to ineffective behavioral analytics systems. However, fine-tuning such a model over time with new data can be detrimental due to catastrophic forgetting. Replay-based approaches in continual learning offer a simple yet efficient method to update such models, minimizing forgetting by maintaining a buffer of important training instances from past learned tasks. However, the main limitation of this approach is the fixed size of the buffer. External knowledge bases can be utilized to overcome this limitation through data augmentation. We propose a novel augmentation-based approach to incorporate external knowledge in the replay-based continual learning framework. We evaluate several strategies with three datasets from prior studies related to deviant behavior classification to assess the integration of external knowledge in continual learning and demonstrate that augmentation helps outperform baseline replay-based approaches.
Authors: Sanjoy Dasgupta, Syamantak Kumar, Shourya Pandey, Purnamrita Sarkar
Abstract: Low-precision streaming PCA estimates the top principal component in a streaming setting under limited precision. We establish an information-theoretic lower bound on the quantization resolution required to achieve a target accuracy for the leading eigenvector. We study Oja's algorithm for streaming PCA under linear and nonlinear stochastic quantization. The quantized variants use unbiased stochastic quantization of the weight vector and the updates. Under mild moment and spectral-gap assumptions on the data distribution, we show that a batched version achieves the lower bound up to logarithmic factors under both schemes. This leads to a nearly dimension-free quantization error in the nonlinear quantization setting. Empirical evaluations on synthetic streams validate our theoretical findings and demonstrate that our low-precision methods closely track the performance of standard Oja's algorithm.
Authors: Amin Omidvar
Abstract: The choice of activation function plays a critical role in neural networks, yet most architectures still rely on fixed, uniform activation functions across all neurons. We introduce SmartMixed, a two-phase training strategy that allows networks to learn optimal per-neuron activation functions while preserving computational efficiency at inference. In the first phase, neurons adaptively select from a pool of candidate activation functions (ReLU, Sigmoid, Tanh, Leaky ReLU, ELU, SELU) using a differentiable hard-mixture mechanism. In the second phase, each neuron's activation function is fixed according to the learned selection, resulting in a computationally efficient network that supports continued training with optimized vectorized operations. We evaluate SmartMixed on the MNIST dataset using feedforward neural networks of varying depths. The analysis shows that neurons in different layers exhibit distinct preferences for activation functions, providing insights into the functional diversity within neural architectures.
Authors: Xingbo Fu, Zhenyu Lei, Zihan Chen, Binchi Zhang, Chuxu Zhang, Jundong Li
Abstract: Graph Neural Networks (GNNs) have revolutionized the field of graph learning by learning expressive graph representations from massive graph data. As a common pattern to train powerful GNNs, the "pre-training, adaptation" scheme first pre-trains GNNs over unlabeled graph data and subsequently adapts them to specific downstream tasks. In the adaptation phase, graph prompting is an effective strategy that modifies input graph data with learnable prompts while keeping pre-trained GNN models frozen. Typically, existing graph prompting studies mainly focus on *feature-oriented* methods that apply graph prompts to node features or hidden representations. However, these studies often achieve suboptimal performance, as they consistently overlook the potential of *topology-oriented* prompting, which adapts pre-trained GNNs by modifying the graph topology. In this study, we conduct a pioneering investigation of graph prompting in terms of graph topology. We propose the first **Graph** **T**opology-**O**riented **P**rompting (GraphTOP) framework to effectively adapt pre-trained GNN models for downstream tasks. More specifically, we reformulate topology-oriented prompting as an edge rewiring problem within multi-hop local subgraphs and relax it into the continuous probability space through reparameterization while ensuring tight relaxation and preserving graph sparsity. Extensive experiments on five graph datasets under four pre-training strategies demonstrate that our proposed GraphTOP outshines six baselines on multiple node classification datasets. Our code is available at https://github.com/xbfu/GraphTOP.
Authors: Jing Yang, Kaitong Cai, Yijia Fan, Yufeng Yang, Keze Wang
Abstract: Full fine-tuning of Large Language Models (LLMs) is notoriously memory-intensive, primarily because conventional optimizers such as SGD or Adam assume access to exact gradients derived from cached activations. Existing solutions either alter the model architecture (e.g., reversible networks) or trade memory for computation (e.g., activation checkpointing), but the optimizer itself remains untouched. In this work, we introduce GradLite, a backward-friendly optimizer that relaxes the requirement of exact gradients, enabling efficient training even when intermediate activations are aggressively discarded or approximated. GradLite leverages two key techniques: (i) low-rank Jacobian approximation, which reduces the dimensionality of backpropagated error signals, and (ii) error-feedback correction, which accumulates and compensates approximation errors across iterations to preserve convergence guarantees. We provide a theoretical analysis showing that GradLite maintains unbiased gradient estimates with bounded variance, ensuring convergence rates comparable to Adam. Empirically, GradLite reduces optimizer-state and activation memory consumption by up to 50\% without architectural changes, and achieves on-par or superior downstream performance on reasoning (MMLU, GSM8K), multilingual, and dialogue benchmarks compared to checkpointing and optimizer-centric baselines (LoMo, GaLore).
Authors: Pritish Chakraborty, Indradyumna Roy, Soumen Chakrabarti, Abir De
Abstract: Retrieving graphs from a large corpus, that contain a subgraph isomorphic to a given query graph, is a core operation in many real-world applications. While recent multi-vector graph representations and scores based on set alignment and containment can provide accurate subgraph isomorphism tests, their use in retrieval remains limited by their need to score corpus graphs exhaustively. We introduce CORGII (Contextual Representation of Graphs for Inverted Indexing), a graph indexing framework in which, starting with a contextual dense graph representation, a differentiable discretization module computes sparse binary codes over a learned latent vocabulary. This text document-like representation allows us to leverage classic, highly optimized inverted indices, while supporting soft (vector) set containment scores. Pushing this paradigm further, we replace the classical, fixed impact weight of a `token' on a graph (such as TFIDF or BM25) with a data-driven, trainable impact weight. Finally, we explore token expansion to support multi-probing the index for smoother accuracy-efficiency tradeoffs. To our knowledge, CORGII is the first indexer of dense graph representations using discrete tokens mapping to efficient inverted lists. Extensive experiments show that CORGII provides better trade-offs between accuracy and efficiency, compared to several baselines.
Authors: Ghadi Nehme, Yanxia Zhang, Dule Shu, Matt Klenk, Faez Ahmed
Abstract: Generating high-fidelity 3D geometries that satisfy specific parameter constraints has broad applications in design and engineering. However, current methods typically rely on large training datasets and struggle with controllability and generalization beyond the training distributions. To overcome these limitations, we introduce LAMP (Linear Affine Mixing of Parametric shapes), a data-efficient framework for controllable and interpretable 3D generation. LAMP first aligns signed distance function (SDF) decoders by overfitting each exemplar from a shared initialization, then synthesizes new geometries by solving a parameter-constrained mixing problem in the aligned weight space. To ensure robustness, we further propose a safety metric that detects geometry validity via linearity mismatch. We evaluate LAMP on two 3D parametric benchmarks: DrivAerNet++ and BlendedNet. We found that LAMP enables (i) controlled interpolation within bounds with as few as 100 samples, (ii) safe extrapolation by up to 100% parameter difference beyond training ranges, (iii) physics performance-guided optimization under fixed parameters. LAMP significantly outperforms conditional autoencoder and Deep Network Interpolation (DNI) baselines in both extrapolation and data efficiency. Our results demonstrate that LAMP advances controllable, data-efficient, and safe 3D generation for design exploration, dataset generation, and performance-driven optimization.
Authors: Ren Yin, Takashi Ishida, Masashi Sugiyama
Abstract: As artificial intelligence (AI) systems approach and surpass expert human performance across a broad range of tasks, obtaining high-quality human supervision for evaluation and training becomes increasingly challenging. Our focus is on tasks that require deep knowledge and skills of multiple domains. Unfortunately, even the best human experts are knowledgeable only in a single narrow area, and will not be able to evaluate the correctness of advanced AI systems on such superhuman tasks. However, based on their narrow expertise, humans may provide a weak signal, i.e., a complementary label indicating an option that is incorrect. For example, a cardiologist could state that "this is not related to cardiology,'' even if they cannot identify the true disease. Based on this weak signal, we propose a scalable oversight framework that enables us to evaluate frontier AI systems without the need to prepare the ground truth. We derive an unbiased estimator of top-1 accuracy from complementary labels and quantify how many complementary labels are needed to match the variance of ordinary labels. We further introduce two estimators to combine scarce ordinary labels with abundant complementary labels. We provide finite-sample deviation guarantees for both complementary-only and the mixed estimators. Empirically, we show that we can evaluate the output of large language models without the ground truth, if we have complementary labels. We further show that we can train an AI system with such weak signals: we show how we can design an agentic AI system automatically that can perform better with this partitioned human supervision. Our code is available at https://github.com/R-Yin-217/Scalable-Oversight-via-Human-Partitioned-Supervision.
URLs: https://github.com/R-Yin-217/Scalable-Oversight-via-Human-Partitioned-Supervision.
Authors: Nikhil Abhyankar, Sanchit Kabra, Saaketh Desai, Chandan K. Reddy
Abstract: Materials discovery requires navigating vast chemical and structural spaces while satisfying multiple, often conflicting, objectives. We present LLM-guided Evolution for MAterials design (LLEMA), a unified framework that couples the scientific knowledge embedded in large language models with chemistry-informed evolutionary rules and memory-based refinement. At each iteration, an LLM proposes crystallographically specified candidates under explicit property constraints; a surrogate-augmented oracle estimates physicochemical properties; and a multi-objective scorer updates success/failure memories to guide subsequent generations. Evaluated on 14 realistic tasks spanning electronics, energy, coatings, optics, and aerospace, LLEMA discovers candidates that are chemically plausible, thermodynamically stable, and property-aligned, achieving higher hit-rates and stronger Pareto fronts than generative and LLM-only baselines. Ablation studies confirm the importance of rule-guided generation, memory-based refinement, and surrogate prediction. By enforcing synthesizability and multi-objective trade-offs, LLEMA delivers a principled pathway to accelerate practical materials discovery. Code: https://github.com/scientific-discovery/LLEMA
Authors: Patrick Pynadath, Jiaxin Shi, Ruqi Zhang
Abstract: While continuous diffusion has shown remarkable success in continuous domains such as image generation, its direct application to discrete data has underperformed compared to purely discrete formulations. This gap is counterintuitive, given that continuous diffusion learns score functions that enable joint evolution across multiple positions. To understand this gap, we introduce token identifiability as an analytical framework for understanding how Gaussian noise corrupts discrete data through two mechanisms: discrete identity corruption and continuous rank degradation. We reveal that these mechanisms scale differently with vocabulary size, creating a temporal dissonance: at noise levels where discrete corruption preserves enough structure for conditional learning, continuous denoising is trivial; at noise levels where continuous denoising is meaningful, discrete corruption destroys nearly all conditional structure. To solve this, we propose CANDI (Continuous ANd DIscrete diffusion), a hybrid framework that decouples discrete and continuous corruption, enabling simultaneous learning of both conditional structure and continuous geometry. We empirically validate the temporal dissonance phenomenon and demonstrate that CANDI successfully avoids it. This unlocks the benefits of continuous diffusion for discrete spaces: on controlled generation, CANDI enables classifier-based guidance with off-the-shelf classifiers through simple gradient addition; on text generation, CANDI outperforms masked diffusion at low NFE, demonstrating the value of learning continuous gradients for discrete spaces.
Authors: Seohong Park, Aditya Oberai, Pranav Atreya, Sergey Levine
Abstract: In this work, we present Transitive Reinforcement Learning (TRL), a new value learning algorithm based on a divide-and-conquer paradigm. TRL is designed for offline goal-conditioned reinforcement learning (GCRL) problems, where the aim is to find a policy that can reach any state from any other state in the smallest number of steps. TRL converts a triangle inequality structure present in GCRL into a practical divide-and-conquer value update rule. This has several advantages compared to alternative value learning paradigms. Compared to temporal difference (TD) methods, TRL suffers less from bias accumulation, as in principle it only requires $O(\log T)$ recursions (as opposed to $O(T)$ in TD learning) to handle a length-$T$ trajectory. Unlike Monte Carlo methods, TRL suffers less from high variance as it performs dynamic programming. Experimentally, we show that TRL achieves the best performance in highly challenging, long-horizon benchmark tasks compared to previous offline GCRL algorithms.
Authors: Junran Wu, Beng Chin Ooi, Ke Xu
Abstract: Signed Graph Neural Networks (SGNNs) are widely adopted to analyze complex patterns in signed graphs with both positive and negative links. Given the noisy nature of real-world connections, the robustness of SGNN has also emerged as a pivotal research area. Under the supervision of empirical properties, graph structure learning has shown its robustness on signed graph representation learning, however, there remains a paucity of research investigating a robust SGNN with theoretical guidance. Inspired by the success of graph information bottleneck (GIB) in information extraction, we propose RIDGE, a novel framework for Robust sI gned graph learning through joint Denoising of Graph inputs and supervision targEts. Different from the basic GIB, we extend the GIB theory with the capability of target space denoising as the co-existence of noise in both input and target spaces. In instantiation, RIDGE effectively cleanses input data and supervision targets via a tractable objective function produced by reparameterization mechanism and variational approximation. We extensively validate our method on four prevalent signed graph datasets, and the results show that RIDGE clearly improves the robustness of popular SGNN models under various levels of noise.
Authors: Pedro Chumpitaz-Flores, My Duong, Cristobal Heredia, Kaixun Hua
Abstract: Constrained clustering leverages limited domain knowledge to improve clustering performance and interpretability, but incorporating pairwise must-link and cannot-link constraints is an NP-hard challenge, making global optimization intractable. Existing mixed-integer optimization methods are confined to small-scale datasets, limiting their utility. We propose Sample-Driven Constrained Group-Based Branch-and-Bound (SDC-GBB), a decomposable branch-and-bound (BB) framework that collapses must-linked samples into centroid-based pseudo-samples and prunes cannot-link through geometric rules, while preserving convergence and guaranteeing global optimality. By integrating grouped-sample Lagrangian decomposition and geometric elimination rules for efficient lower and upper bounds, the algorithm attains highly scalable pairwise k-Means constrained clustering via parallelism. Experimental results show that our approach handles datasets with 200,000 samples with cannot-link constraints and 1,500,000 samples with must-link constraints, which is 200 - 1500 times larger than the current state-of-the-art under comparable constraint settings, while reaching an optimality gap of less than 3%. In providing deterministic global guarantees, our method also avoids the search failures that off-the-shelf heuristics often encounter on large datasets.
Authors: Michael Ito, Danai Koutra, Jenna Wiens
Abstract: Random walk neural networks (RWNNs) have emerged as a promising approach for graph representation learning, leveraging recent advances in sequence models to process random walks. However, under realistic sampling constraints, RWNNs often fail to capture global structure even in small graphs due to incomplete node and edge coverage, limiting their expressivity. To address this, we propose \textit{random search neural networks} (RSNNs), which operate on random searches, each of which guarantees full node coverage. Theoretically, we demonstrate that in sparse graphs, only $O(\log |V|)$ searches are needed to achieve full edge coverage, substantially reducing sampling complexity compared to the $O(|V|)$ walks required by RWNNs (assuming walk lengths scale with graph size). Furthermore, when paired with universal sequence models, RSNNs are universal approximators. We lastly show RSNNs are probabilistically invariant to graph isomorphisms, ensuring their expectation is an isomorphism-invariant graph function. Empirically, RSNNs consistently outperform RWNNs on molecular and protein benchmarks, achieving comparable or superior performance with up to 16$\times$ fewer sampled sequences. Our work bridges theoretical and practical advances in random walk based approaches, offering an efficient and expressive framework for learning on sparse graphs.
Authors: Ashwin Ramachandran, Vaibhav Raj, Indrayumna Roy, Soumen Chakrabarti, Abir De
Abstract: Graph retrieval based on subgraph isomorphism has several real-world applications such as scene graph retrieval, molecular fingerprint detection and circuit design. Roy et al. [35] proposed IsoNet, a late interaction model for subgraph matching, which first computes the node and edge embeddings of each graph independently of paired graph and then computes a trainable alignment map. Here, we present IsoNet++, an early interaction graph neural network (GNN), based on several technical innovations. First, we compute embeddings of all nodes by passing messages within and across the two input graphs, guided by an injective alignment between their nodes. Second, we update this alignment in a lazy fashion over multiple rounds. Within each round, we run a layerwise GNN from scratch, based on the current state of the alignment. After the completion of one round of GNN, we use the last-layer embeddings to update the alignments, and proceed to the next round. Third, IsoNet++ incorporates a novel notion of node-pair partner interaction. Traditional early interaction computes attention between a node and its potential partners in the other graph, the attention then controlling messages passed across graphs. In contrast, we consider node pairs (not single nodes) as potential partners. Existence of an edge between the nodes in one graph and non-existence in the other provide vital signals for refining the alignment. Our experiments on several datasets show that the alignments get progressively refined with successive rounds, resulting in significantly better retrieval performance than existing methods. We demonstrate that all three innovations contribute to the enhanced accuracy. Our code and datasets are publicly available at https://github.com/structlearning/isonetpp.
Authors: Yuyang Ding, Chi Zhang, Juntao Li, Haibin Lin, Xin Liu, Min Zhang
Abstract: Reinforcement learning with verifiable rewards (RLVR) has emerged as a promising paradigm for enhancing the reasoning capabilities of large language models (LLMs). In this context, models explore reasoning trajectories and exploit rollouts with correct answers as positive signals for policy optimization. However, these rollouts might involve flawed patterns such as answer-guessing and jump-in-reasoning. Such flawed-positive rollouts are rewarded identically to fully correct ones, causing policy models to internalize these unreliable reasoning patterns. In this work, we first conduct a systematic study of flawed-positive rollouts in RL and find that they enable rapid capability gains during the early optimization stage, while constraining reasoning capability later by reinforcing unreliable patterns. Building on these insights, we propose Flawed-Aware Policy Optimization (FAPO), which presents a parameter-free reward penalty for flawed-positive rollouts, enabling the policy to leverage them as useful shortcuts in the warm-up stage, securing stable early gains, while gradually shifting optimization toward reliable reasoning in the later refinement stage. To accurately and comprehensively detect flawed-positive rollouts, we introduce a generative reward model (GenRM) with a process-level reward that precisely localizes reasoning errors. Experiments show that FAPO is effective in broad domains, improving outcome correctness, process reliability, and training stability without increasing the token budget.
Authors: Maximilian Matyash, Avigdor Gal, Arik Senderovich
Abstract: With recent technological advances, process logs, which were traditionally deterministic in nature, are being captured from non-deterministic sources, such as uncertain sensors or machine learning models (that predict activities using cameras). In the presence of stochastically-known logs, logs that contain probabilistic information, the need for stochastic trace recovery increases, to offer reliable means of understanding the processes that govern such systems. We design a novel deep learning approach for stochastic trace recovery, based on Diffusion Denoising Probabilistic Models (DDPM), which makes use of process knowledge (either implicitly by discovering a model or explicitly by injecting process knowledge in the training phase) to recover traces by denoising. We conduct an empirical evaluation demonstrating state-of-the-art performance with up to a 25% improvement over existing methods, along with increased robustness under high noise levels.
Authors: Eduard Popescu, Adrian Groza, Andreea Cernat
Abstract: The task here is to predict the toxicological activity of chemical compounds based on the Tox21 dataset, a benchmark in computational toxicology. After a domain-specific overview of chemical toxicity, we discuss current computational strategies, focusing on machine learning and deep learning. Several architectures are compared in terms of performance, robustness, and interpretability. This research introduces a novel image-based pipeline based on DenseNet121, which processes 2D graphical representations of chemical structures. Additionally, we employ Grad-CAM visualizations, an explainable AI technique, to interpret the model's predictions and highlight molecular regions contributing to toxicity classification. The proposed architecture achieves competitive results compared to traditional models, demonstrating the potential of deep convolutional networks in cheminformatics. Our findings emphasize the value of combining image-based representations with explainable AI methods to improve both predictive accuracy and model transparency in toxicology.
Authors: Dhruv Sarkar, Abhishek Sinha
Abstract: We propose an anytime online algorithm for the problem of learning a sequence of adversarial convex cost functions while approximately satisfying another sequence of adversarial online convex constraints. A sequential algorithm is called \emph{anytime} if it provides a non-trivial performance guarantee for any intermediate timestep $t$ without requiring prior knowledge of the length of the entire time horizon $T$. Our proposed algorithm achieves optimal performance bounds without resorting to the standard doubling trick, which has poor practical performance due to multiple restarts. Our core technical contribution is the use of time-varying Lyapunov functions to keep track of constraint violations. This must be contrasted with prior works that used a fixed Lyapunov function tuned to the known horizon length $T$. The use of time-varying Lyapunov function poses unique analytical challenges as properties, such as \emph{monotonicity}, on which the prior proofs rest, no longer hold. By introducing a new analytical technique, we show that our algorithm achieves $O(\sqrt{t})$ regret and $\tilde{O}(\sqrt{t})$ cumulative constraint violation bounds for any $t\geq 1$. We extend our results to the dynamic regret setting, achieving bounds that adapt to the path length of the comparator sequence without prior knowledge of its total length. We also present an adaptive algorithm in the optimistic setting, whose performance gracefully scales with the cumulative prediction error. We demonstrate the practical utility of our algorithm through numerical experiments involving the online shortest path problem.
Authors: Noa Shoham, Ron Dorfman, Shalev Shaer, Kfir Y. Levy, Yaniv Romano
Abstract: Prediction-Powered Inference (PPI) is a recently proposed statistical inference technique for parameter estimation that leverages pseudo-labels on both labeled and unlabeled data to construct an unbiased, low-variance estimator. In this work, we extend its core idea to semi-supervised learning (SSL) for model training, introducing a novel unbiased gradient estimator. This extension addresses a key challenge in SSL: while unlabeled data can improve model performance, its benefit heavily depends on the quality of pseudo-labels. Inaccurate pseudo-labels can introduce bias, leading to suboptimal models.To balance the contributions of labeled and pseudo-labeled data, we utilize an interpolation parameter and tune it on the fly, alongside the model parameters, using a one-dimensional online learning algorithm. We verify the practical advantage of our approach through experiments on both synthetic and real datasets, demonstrating improved performance over classic SSL baselines and PPI methods that tune the interpolation parameter offline.
Authors: Yasharth Yadav, Kelin Xia
Abstract: Geometric data analysis and learning has emerged as a distinct and rapidly developing research area, increasingly recognized for its effectiveness across diverse applications. At the heart of this field lies curvature, a powerful and interpretable concept that captures intrinsic geometric structure and underpins numerous tasks, from community detection to geometric deep learning. A wide range of discrete curvature models have been proposed for various data representations, including graphs, simplicial complexes, cubical complexes, and point clouds sampled from manifolds. These models not only provide efficient characterizations of data geometry but also constitute essential components in geometric learning frameworks. In this paper, we present the first comprehensive review of existing discrete curvature models, covering their mathematical foundations, computational formulations, and practical applications in data analysis and learning. In particular, we discuss discrete curvature from both Riemannian and metric geometry perspectives and propose a systematic pipeline for curvature-driven data analysis. We further examine the corresponding computational algorithms across different data representations, offering detailed comparisons and insights. Finally, we review state-of-the-art applications of curvature in both supervised and unsupervised learning. This survey provides a conceptual and practical roadmap for researchers to gain a better understanding of discrete curvature as a fundamental tool for geometric understanding and learning.
Authors: Songhan Zhang, Yuanhao Lai, Pengfei Zheng, Boxi Yu, Xiaoying Tang, Qiuai Fu, Pinjia He
Abstract: Multivariate time series (MTS) anomaly detection is essential for maintaining the reliability of industrial systems, yet real-world deployment is hindered by two critical challenges: training data contamination (noises and hidden anomalies) and inefficient model inference. Existing unsupervised methods assume clean training data, but contamination distorts learned patterns and degrades detection accuracy. Meanwhile, complex deep models often overfit to contamination and suffer from high latency, limiting practical use. To address these challenges, we propose CLEANet, a robust and efficient anomaly detection framework in contaminated multivariate time series. CLEANet introduces a Contamination-Resilient Training Framework (CRTF) that mitigates the impact of corrupted samples through an adaptive reconstruction weighting strategy combined with clustering-guided contrastive learning, thereby enhancing robustness. To further avoid overfitting on contaminated data and improve computational efficiency, we design a lightweight conjugate MLP that disentangles temporal and cross-feature dependencies. Across five public datasets, CLEANet achieves up to 73.04% higher F1 and 81.28% lower runtime compared with ten state-of-the-art baselines. Furthermore, integrating CRTF into three advanced models yields an average 5.35% F1 gain, confirming its strong generalizability.
Authors: Divya Jyoti Bajpai, Manjesh Kumar Hanawal
Abstract: Vision-language Models (VLMs) have made significant strides in visual understanding and query response generation, but often face challenges of high computational cost and inference latency due to autoregressive decoding. In this work, we introduce an imitation-learning-based Self-Speculative Decoding (SSD) framework, named FastVLM, to address these limitations. Our approach employs a lightweight draft model for token generation in an autoregressive manner, while a full model verifies these tokens non-autoregressively. Accepted tokens proceed seamlessly, while rejected tokens are corrected by the full model and used to guide the draft model's refinement. Through an imitation network, FastVLM enhances the draft model by integrating deeper level insights from the full model's architecture. Also, it maintains the performance integrity of the full model while training the draft model, achieving a balance between efficiency and accuracy. Our method speeds up the inference process by 1.55-1.85x as compared to the final layer with minimal loss in performance.
Authors: Sofiane Ennadir, Oleg Smirnov, Yassine Abbahaddou, Lele Cao, Johannes F. Lutzeyer
Abstract: Graph Neural Networks (GNNs) have achieved strong performance across a range of graph representation learning tasks, yet their adversarial robustness in graph classification remains underexplored compared to node classification. While most existing defenses focus on the message-passing component, this work investigates the overlooked role of pooling operations in shaping robustness. We present a theoretical analysis of standard flat pooling methods (sum, average and max), deriving upper bounds on their adversarial risk and identifying their vulnerabilities under different attack scenarios and graph structures. Motivated by these insights, we propose \textit{Robust Singular Pooling (RS-Pool)}, a novel pooling strategy that leverages the dominant singular vector of the node embedding matrix to construct a robust graph-level representation. We theoretically investigate the robustness of RS-Pool and interpret the resulting bound leading to improved understanding of our proposed pooling operator. While our analysis centers on Graph Convolutional Networks (GCNs), RS-Pool is model-agnostic and can be implemented efficiently via power iteration. Empirical results on real-world benchmarks show that RS-Pool provides better robustness than the considered pooling methods when subject to state-of-the-art adversarial attacks while maintaining competitive clean accuracy. Our code is publicly available at:\href{https://github.com/king/rs-pool}{https://github.com/king/rs-pool}.
URLs: https://github.com/king/rs-pool, https://github.com/king/rs-pool
Authors: Lu Xu, Tsai Hor Chan, Kwok Fai Lam, Lequan Yu, Guosheng Yin
Abstract: Density estimation is essential for generative modeling, particularly with the rise of modern neural networks. While existing methods capture complex data distributions, they often lack interpretability and uncertainty quantification. Bayesian nonparametric methods, especially the \polya tree, offer a robust framework that addresses these issues by accurately capturing function behavior over small intervals. Traditional techniques like Markov chain Monte Carlo (MCMC) face high computational complexity and scalability limitations, hindering the use of Bayesian nonparametric methods in deep learning. To tackle this, we introduce the variational \polya tree (VPT) model, which employs stochastic variational inference to compute posterior distributions. This model provides a flexible, nonparametric Bayesian prior that captures latent densities and works well with stochastic gradient optimization. We also leverage the joint distribution likelihood for a more precise variational posterior approximation than traditional mean-field methods. We evaluate the model performance on both real data and images, and demonstrate its competitiveness with other state-of-the-art deep density estimation methods. We also explore its ability in enhancing interpretability and uncertainty quantification. Code is available at https://github.com/howardchanth/var-polya-tree.
Authors: Sofiane Ennadir, Johannes F. Lutzeyer, Michalis Vazirgiannis, El Houcine Bergou
Abstract: Graph Neural Networks (GNNs) have demonstrated remarkable performance across a spectrum of graph-related tasks, however concerns persist regarding their vulnerability to adversarial perturbations. While prevailing defense strategies focus primarily on pre-processing techniques and adaptive message-passing schemes, this study delves into an under-explored dimension: the impact of weight initialization and associated hyper-parameters, such as training epochs, on a model's robustness. We introduce a theoretical framework bridging the connection between initialization strategies and a network's resilience to adversarial perturbations. Our analysis reveals a direct relationship between initial weights, number of training epochs and the model's vulnerability, offering new insights into adversarial robustness beyond conventional defense mechanisms. While our primary focus is on GNNs, we extend our theoretical framework, providing a general upper-bound applicable to Deep Neural Networks. Extensive experiments, spanning diverse models and real-world datasets subjected to various adversarial attacks, validate our findings. We illustrate that selecting appropriate initialization not only ensures performance on clean datasets but also enhances model robustness against adversarial perturbations, with observed gaps of up to 50\% compared to alternative initialization approaches.
Authors: Ilgam Latypov, Alexandra Suvorikova, Alexey Kroshnin, Alexander Gasnikov, Yuriy Dorn
Abstract: In many modern applications, a system must dynamically choose between several adaptive learning algorithms that are trained online. Examples include model selection in streaming environments, switching between trading strategies in finance, and orchestrating multiple contextual bandit or reinforcement learning agents. At each round, a learner must select one predictor among $K$ adaptive experts to make a prediction, while being able to update at most $M \le K$ of them under a fixed training budget. We address this problem in the \emph{stochastic setting} and introduce \algname{M-LCB}, a computationally efficient UCB-style meta-algorithm that provides \emph{anytime regret guarantees}. Its confidence intervals are built directly from realized losses, require no additional optimization, and seamlessly reflect the convergence properties of the underlying experts. If each expert achieves internal regret $\tilde O(T^\alpha)$, then \algname{M-LCB} ensures overall regret bounded by $\tilde O\!\Bigl(\sqrt{\tfrac{KT}{M}} \;+\; (K/M)^{1-\alpha}\,T^\alpha\Bigr)$. To our knowledge, this is the first result establishing regret guarantees when multiple adaptive experts are trained simultaneously under per-round budget constraints. We illustrate the framework with two representative cases: (i) parametric models trained online with stochastic losses, and (ii) experts that are themselves multi-armed bandit algorithms. These examples highlight how \algname{M-LCB} extends the classical bandit paradigm to the more realistic scenario of coordinating stateful, self-learning experts under limited resources.
Authors: Berken Utku Demirel, Christian Holz
Abstract: Self-supervised learning (SSL) has emerged as a powerful paradigm for learning representations without labeled data. Most SSL approaches rely on strong, well-established, handcrafted data augmentations to generate diverse views for representation learning. However, designing such augmentations requires domain-specific knowledge and implicitly imposes representational invariances on the model, which can limit generalization. In this work, we propose an unsupervised representation learning method that replaces augmentations by generating views using orthonormal bases and overcomplete frames. We show that embeddings learned from orthonormal and overcomplete spaces reside on distinct manifolds, shaped by the geometric biases introduced by representing samples in different spaces. By jointly leveraging the complementary geometry of these distinct manifolds, our approach achieves superior performance without artificially increasing data diversity through strong augmentations. We demonstrate the effectiveness of our method on nine datasets across five temporal sequence tasks, where signal-specific characteristics make data augmentations particularly challenging. Without relying on augmentation-induced diversity, our method achieves performance gains of up to 15--20\% over existing self-supervised approaches. Source code: https://github.com/eth-siplab/Learning-with-FrameProjections
URLs: https://github.com/eth-siplab/Learning-with-FrameProjections
Authors: Shan Zhong, Shutong Ding, He Diao, Xiangyu Wang, Kah Chan Teh, Bei Peng
Abstract: Reliable value estimation serves as the cornerstone of reinforcement learning (RL) by evaluating long-term returns and guiding policy improvement, significantly influencing the convergence speed and final performance. Existing works improve the reliability of value function estimation via multi-critic ensembles and distributional RL, yet the former merely combines multi point estimation without capturing distributional information, whereas the latter relies on discretization or quantile regression, limiting the expressiveness of complex value distributions. Inspired by flow matching's success in generative modeling, we propose a generative paradigm for value estimation, named FlowCritic. Departing from conventional regression for deterministic value prediction, FlowCritic leverages flow matching to model value distributions and generate samples for value estimation.
Authors: Wei Chen, Linjun Peng, Zhiyi Huang, Haoyue Dai, Zhifeng Hao, Ruichu Cai, Kun Zhang
Abstract: Recovering causal structure in the presence of latent variables is an important but challenging task. While many methods have been proposed to handle it, most of them require strict and/or untestable assumptions on the causal structure. In real-world scenarios, observed variables may be affected by multiple latent variables simultaneously, which, generally speaking, cannot be handled by these methods. In this paper, we consider the linear, non-Gaussian case, and make use of the joint higher-order cumulant matrix of the observed variables constructed in a specific way. We show that, surprisingly, causal asymmetry between two observed variables can be directly seen from the rank deficiency properties of such higher-order cumulant matrices, even in the presence of an arbitrary number of latent confounders. Identifiability results are established, and the corresponding identification methods do not even involve iterative procedures. Experimental results demonstrate the effectiveness and asymptotic correctness of our proposed method.
Authors: Khai Le-Duc, Duy M. H. Nguyen, Phuong T. H. Trinh, Tien-Phat Nguyen, Nghiem T. Diep, An Ngo, Tung Vu, Trinh Vuong, Anh-Tien Nguyen, Mau Nguyen, Van Trung Hoang, Khai-Nguyen Nguyen, Hy Nguyen, Chris Ngo, Anji Liu, Nhat Ho, Anne-Christin Hauschild, Khanh Xuan Nguyen, Thanh Nguyen-Tang, Pengtao Xie, Daniel Sonntag, James Zou, Mathias Niepert, Anh Totti Nguyen
Abstract: Faithful reasoning in medical vision-language models (VLMs) requires not only accurate predictions but also transparent alignment between textual rationales and visual evidence. While Chain-of-Thought (CoT) prompting has shown promise in medical visual question answering (VQA), no large-scale expert-level dataset has captured stepwise reasoning with precise visual grounding. We introduce S-Chain, the first large-scale dataset of 12,000 expert-annotated medical images with bounding boxes and structured visual CoT (SV-CoT), explicitly linking visual regions to reasoning steps. The dataset further supports 16 languages, totaling over 700k VQA pairs for broad multilingual applicability. Using S-Chain, we benchmark state-of-the-art medical VLMs (ExGra-Med, LLaVA-Med) and general-purpose VLMs (Qwen2.5-VL, InternVL2.5), showing that SV-CoT supervision significantly improves interpretability, grounding fidelity, and robustness. Beyond benchmarking, we study its synergy with retrieval-augmented generation, revealing how domain knowledge and visual grounding interact during autoregressive reasoning. Finally, we propose a new mechanism that strengthens the alignment between visual evidence and reasoning, improving both reliability and efficiency. S-Chain establishes a new benchmark for grounded medical reasoning and paves the way toward more trustworthy and explainable medical VLMs.
Authors: Jiali Cheng, Anjishnu Kumar, Roshan Lal, Rishi Rajasekaran, Hani Ramezani, Omar Zia Khan, Oleg Rokhlenko, Sunny Chiu-Webster, Gang Hua, Hadi Amiri
Abstract: We observe that current state-of-the-art web-agents are unable to effectively adapt to new environments without neural network fine-tuning, without which they produce inefficient execution plans due to a lack of awareness of the structure and dynamics of the new environment. To address this limitation, we introduce ATLAS (Actor-Critic Task-completion with Look-ahead Action Simulation), a memory-augmented agent that is able to make plans grounded in a model of the environment by simulating the consequences of those actions in cognitive space. Our agent starts by building a "cognitive map" by performing a lightweight curiosity driven exploration of the environment. The planner proposes candidate actions; the simulator predicts their consequences in cognitive space; a critic analyzes the options to select the best roll-out and update the original plan; and a browser executor performs the chosen action. On the WebArena-Lite Benchmark, we achieve a 63% success rate compared to 53.9% success rate for the previously published state-of-the-art. Unlike previous systems, our modular architecture requires no website-specific LLM fine-tuning. Ablations show sizable drops without the world-model, hierarchical planner, and look-ahead-based replanner confirming their complementary roles within the design of our system
Authors: Yuanhao Lai, Pengfei Zheng, Chenpeng Ji, Yan Li, Songhan Zhang, Rutao Zhang, Zhengang Wang, Yunfei Du
Abstract: Gaussian-Process-based Bayesian optimization (GP-BO), is a prevailing model-based framework for DBMS auto-tuning. However, recent work shows GP-BO-based DBMS auto-tuners significantly outperformed auto-tuners based on SMAC, which features random forest surrogate models; such results motivate us to rethink and investigate the limitations of GP-BO in auto-tuner design. We find the fundamental assumptions of GP-BO are widely violated when modeling and optimizing DBMS performance, while tree-ensemble-BOs (e.g., SMAC) can avoid the assumption pitfalls and deliver improved tuning efficiency and effectiveness. Moreover, we argue that existing tree-ensemble-BOs restrict further advancement in DBMS auto-tuning. First, existing tree-ensemble-BOs can only achieve distribution-free point estimates, but still impose unrealistic distributional assumptions on uncertainty estimates, compromising surrogate modeling and distort the acquisition function. Second, recent advances in gradient boosting, which can further enhance surrogate modeling against vanilla GP and random forest counterparts, have rarely been applied in optimizing DBMS auto-tuners. To address these issues, we propose a novel model-based DBMS auto-tuner, Centrum. Centrum improves distribution-free point and interval estimation in surrogate modeling with a two-phase learning procedure of stochastic gradient boosting ensembles. Moreover, Centrum adopts a generalized SGBE-estimated locally-adaptive conformal prediction to facilitate a distribution-free uncertainty estimation and acquisition function. To our knowledge, Centrum is the first auto-tuner to realize distribution-freeness, enhancing BO's practicality in DBMS auto-tuning, and the first to seamlessly fuse gradient boosting ensembles and conformal inference in BO. Extensive physical and simulation experiments on two DBMSs and three workloads show Centrum outperforms 21 SOTA methods.
Authors: Jiaqi Wen, Jianyi Yang
Abstract: This paper studies Distributionally Robust Optimization (DRO), a fundamental framework for enhancing the robustness and generalization of statistical learning and optimization. An effective ambiguity set for DRO must involve distributions that remain consistent with the nominal distribution while being diverse enough to account for a variety of potential scenarios. Moreover, it should lead to tractable DRO solutions. To this end, we propose a diffusion-based ambiguity set design that captures various adversarial distributions beyond the nominal support space while maintaining consistency with the nominal distribution. Building on this ambiguity modeling, we propose Diffusion-based DRO (D-DRO), a tractable DRO algorithm that solves the inner maximization over the parameterized diffusion model space. We formally establish the stationary convergence performance of D-DRO and empirically demonstrate its superior Out-of-Distribution (OOD) generalization performance in a ML prediction task.
Authors: Omar Naim, Krish Sharma, Nicholas Asher
Abstract: In this paper we introduce Tale, Task-Aware Layer Elimination, an inference-time algorithm that prunes entire transformer layers in an LLM by directly optimizing task-specific validation performance. We evaluate TALE on 9 tasks and 5 models, including LLaMA 3.1 8B, Qwen 2.5 7B, Qwen 2.5 0.5B, Mistral 7B, and Lucie 7B, under both zero-shot and few-shot settings. Unlike prior approaches, TALE requires no retraining and consistently improves accuracy while reducing computational cost across all benchmarks. Furthermore, applying TALE during finetuning leads to additional performance gains. Finally, TALE provides flexible user control over trade-offs between accuracy and efficiency. Mutual information analysis shows that certain layers act as bottlenecks, degrading task-relevant representations. Tale's selective layer removal remedies this problem, producing smaller, faster, and more accurate models that are also faster to fine-tune while offering new insights into transformer interpretability.
Authors: Wenrui Cai, Defa Zhu, Qingjie Liu, Qiyang Min
Abstract: Normalization layer constitutes an essential component in neural networks. In transformers, the predominantly used RMSNorm constrains vectors to a unit hypersphere, followed by dimension-wise rescaling through a learnable scaling coefficient $\gamma$ to maintain the representational capacity of the model. However, RMSNorm discards the input norm information in forward pass and a static scaling factor $\gamma$ may be insufficient to accommodate the wide variability of input data and distributional shifts, thereby limiting further performance improvements, particularly in zero-shot scenarios that large language models routinely encounter. To address this limitation, we propose SeeDNorm, which enhances the representational capability of the model by dynamically adjusting the scaling coefficient based on the current input, thereby preserving the input norm information and enabling data-dependent, self-rescaled dynamic normalization. During backpropagation, SeeDNorm retains the ability of RMSNorm to dynamically adjust gradient according to the input norm. We provide a detailed analysis of the training optimization for SeedNorm and proposed corresponding solutions to address potential instability issues that may arise when applying SeeDNorm. We validate the effectiveness of SeeDNorm across models of varying sizes in large language model pre-training as well as supervised and unsupervised computer vision tasks. By introducing a minimal number of parameters and with neglligible impact on model efficiency, SeeDNorm achieves consistently superior performance compared to previously commonly used normalization layers such as RMSNorm and LayerNorm, as well as element-wise activation alternatives to normalization layers like DyT.
Authors: Florian Gr\"otschla, Elia Trachsel, Luca A. Lanzend\"orfer, Roger Wattenhofer
Abstract: Graph-based recommender systems are commonly trained in transductive settings, which limits their applicability to new users, items, or datasets. We propose NBF-Rec, a graph-based recommendation model that supports inductive transfer learning across datasets with disjoint user and item sets. Unlike conventional embedding-based methods that require retraining for each domain, NBF-Rec computes node embeddings dynamically at inference time. We evaluate the method on seven real-world datasets spanning movies, music, e-commerce, and location check-ins. NBF-Rec achieves competitive performance in zero-shot settings, where no target domain data is used for training, and demonstrates further improvements through lightweight fine-tuning. These results show that inductive transfer is feasible in graph-based recommendation and that interaction-level message passing supports generalization across datasets without requiring aligned users or items.
Authors: Christopher J. Hazard, Michael Resnick, Jacob Beel, Jack Xia, Cade Mack, Dominic Glennie, Matthew Fulp, David Maze, Andrew Bassett, Martin Koistinen
Abstract: Traditional machine learning relies on explicit models and domain assumptions, limiting flexibility and interpretability. We introduce a model-free framework using surprisal (information theoretic uncertainty) to directly analyze and perform inferences from raw data, eliminating distribution modeling, reducing bias, and enabling efficient updates including direct edits and deletion of training data. By quantifying relevance through uncertainty, the approach enables generalizable inference across tasks including generative inference, causal discovery, anomaly detection, and time series forecasting. It emphasizes traceability, interpretability, and data-driven decision making, offering a unified, human-understandable framework for machine learning, and achieves at or near state-of-the-art performance across most common machine learning tasks. The mathematical foundations create a ``physics'' of information, which enable these techniques to apply effectively to a wide variety of complex data types, including missing data. Empirical results indicate that this may be a viable alternative path to neural networks with regard to scalable machine learning and artificial intelligence that can maintain human understandability of the underlying mechanics.
Authors: Jingyuan Liu, Hao Qiu, Lin Yang, Mengfan Xu
Abstract: We study the distributed multi-agent multi-armed bandit problem with heterogeneous rewards over random communication graphs. Uniquely, at each time step $t$ agents communicate over a time-varying random graph $G_t$ generated by applying the Erd\H{o}s-R\'enyi model to a fixed connected base graph $G$ (for classical Erd\H{o}s-R\'enyi graphs, $G$ is a complete graph), where each potential edge in $G$ is randomly and independently present with the link probability $p$. Notably, the resulting random graph is not necessarily connected at each time step. Each agent's arm rewards follow time-invariant distributions, and the reward distribution for the same arm may differ across agents. The goal is to minimize the cumulative expected regret relative to the global mean reward of each arm, defined as the average of that arm's mean rewards across all agents. To this end, we propose a fully distributed algorithm that integrates the arm elimination strategy with the random gossip algorithm. We theoretically show that the regret upper bound is of order $\log T$ and is highly interpretable, where $T$ is the time horizon. It includes the optimal centralized regret $O\left(\sum_{k: \Delta_k>0} \frac{\log T}{\Delta_k}\right)$ and an additional term $O\left(\frac{N^2 \log T}{p \lambda_{N-1}(Lap(G))} + \frac{KN^2 \log T}{p}\right)$ where $N$ and $K$ denote the total number of agents and arms, respectively. This term reflects the impact of $G$'s algebraic connectivity $\lambda_{N-1}(Lap(G))$ and the link probability $p$, and thus highlights a fundamental trade-off between communication efficiency and regret. As a by-product, we show a nearly optimal regret lower bound. Finally, our numerical experiments not only show the superiority of our algorithm over existing benchmarks, but also validate the theoretical regret scaling with problem complexity.
Authors: Soham Pahari, Sandeep Chand Kumain
Abstract: Air pollution remains a critical environmental and public health concern in Indian megacities such as Delhi, Kolkata, and Mumbai, where sudden spikes in pollutant levels challenge timely intervention. Accurate Air Quality Index (AQI) forecasting is difficult due to the coexistence of linear trends, seasonal variations, and volatile nonlinear patterns. This paper proposes a hybrid forecasting framework that integrates LOESS decomposition, ARIMA modeling, and a multi-scale CNN-BiLSTM network with a residual-gated attention mechanism. The LOESS step separates the AQI series into trend, seasonal, and residual components, with ARIMA modeling the smooth components and the proposed deep learning module capturing multi-scale volatility in the residuals. Model hyperparameters are tuned via the Unified Adaptive Multi-Stage Metaheuristic Optimizer (UAMMO), combining multiple optimization strategies for efficient convergence. Experiments on 2021-2023 AQI datasets from the Central Pollution Control Board show that the proposed method consistently outperforms statistical, deep learning, and hybrid baselines across PM2.5, O3, CO, and NOx in three major cities, achieving up to 5-8% lower MSE and higher R^2 scores (>0.94) for all pollutants. These results demonstrate the framework's robustness, sensitivity to sudden pollution events, and applicability to urban air quality management.
Authors: Jingxin Zhan, Yuze Han, Zhihua Zhang
Abstract: The convergence analysis of online learning algorithms is central to machine learning theory, where last-iterate convergence is particularly important, as it captures the learner's actual decisions and describes the evolution of the learning process over time. However, in multi-armed bandits, most existing algorithmic analyses mainly focus on the order of regret, while the last-iterate (simple regret) convergence rate remains less explored -- especially for the widely studied Follow-the-Regularized-Leader (FTRL) algorithms. Recently, a growing line of work has established the Best-of-Both-Worlds (BOBW) property of FTRL algorithms in bandit problems, showing in particular that they achieve logarithmic regret in stochastic bandits. Nevertheless, their last-iterate convergence rate has not yet been studied. Intuitively, logarithmic regret should correspond to a $t^{-1}$ last-iterate convergence rate. This paper partially confirms this intuition through theoretical analysis, showing that the Bregman divergence, defined by the regular function $\Psi(p)=-4\sum_{i=1}^{d}\sqrt{p_i}$ associated with the BOBW FTRL algorithm $1/2$-Tsallis-INF (arXiv:1807.07623), between the point mass on the optimal arm and the probability distribution over the arm set obtained at iteration $t$, decays at a rate of $t^{-1/2}$.
Authors: Mirco A. Mannucci
Abstract: GANs promise indistinguishability, logic explains it. We put the two on a budget: a discriminator that can only ``see'' up to a logical depth $k$, and a generator that must look correct to that bounded observer. \textbf{LOGAN} (LOGical GANs) casts the discriminator as a depth-$k$ Ehrenfeucht--Fra\"iss\'e (EF) \emph{Opponent} that searches for small, legible faults (odd cycles, nonplanar crossings, directed bridges), while the generator plays \emph{Builder}, producing samples that admit a $k$-round matching to a target theory $T$. We ship a minimal toolkit -- an EF-probe simulator and MSO-style graph checkers -- and four experiments including real neural GAN training with PyTorch. Beyond verification, we score samples with a \emph{logical loss} that mixes budgeted EF round-resilience with cheap certificate terms, enabling a practical curriculum on depth. Framework validation demonstrates $92\%$--$98\%$ property satisfaction via simulation (Exp.~3), while real neural GAN training achieves $5\%$--$14\%$ improvements on challenging properties and $98\%$ satisfaction on connectivity (matching simulation) through adversarial learning (Exp.~4). LOGAN is a compact, reproducible path toward logic-bounded generation with interpretable failures, proven effectiveness (both simulated and real training), and dials for control.
Authors: Dominik Meier, Shixing Yu, Sagnik Nandy, Promit Ghosal, Kyra Gan
Abstract: Single-cell RNA sequencing (scRNA-seq) enables the study of cellular heterogeneity. Yet, clustering accuracy, and with it downstream analyses based on cell labels, remain challenging due to measurement noise and biological variability. In standard latent spaces (e.g., obtained through PCA), data from different cell types can be projected close together, making accurate clustering difficult. We introduce a latent plug-and-play diffusion framework that separates the observation and denoising space. This separation is operationalized through a novel Gibbs sampling procedure: the learned diffusion prior is applied in a low-dimensional latent space to perform denoising, while to steer this process, noise is reintroduced into the original high-dimensional observation space. This unique "input-space steering" ensures the denoising trajectory remains faithful to the original data structure. Our approach offers three key advantages: (1) adaptive noise handling via a tunable balance between prior and observed data; (2) uncertainty quantification through principled uncertainty estimates for downstream analysis; and (3) generalizable denoising by leveraging clean reference data to denoise noisier datasets, and via averaging, improve quality beyond the training set. We evaluate robustness on both synthetic and real single-cell genomics data. Our method improves clustering accuracy on synthetic data across varied noise levels and dataset shifts. On real-world single-cell data, our method demonstrates improved biological coherence in the resulting cell clusters, with cluster boundaries that better align with known cell type markers and developmental trajectories.
Authors: Divyesh Savaliya, Marius E. Yamakou
Abstract: Self-induced stochastic resonance (SISR) is the emergence of coherent oscillations in slow-fast excitable systems driven solely by noise, without external periodic forcing or proximity to a bifurcation. This work presents a physics-informed machine learning framework for modeling and predicting SISR in the stochastic FitzHugh-Nagumo neuron. We embed the governing stochastic differential equations and SISR-asymptotic timescale-matching constraints directly into a Physics-Informed Neural Network (PINN) based on a Noise-Augmented State Predictor architecture. The composite loss integrates data fidelity, dynamical residuals, and barrier-based physical constraints derived from Kramers' escape theory. The trained PINN accurately predicts the dependence of spike-train coherence on noise intensity, excitability, and timescale separation, matching results from direct stochastic simulations with substantial improvements in accuracy and generalization compared with purely data-driven methods, while requiring significantly less computation. The framework provides a data-efficient and interpretable surrogate model for simulating and analyzing noise-induced coherence in multiscale stochastic systems.
Authors: Marianne Arriola, Yair Schiff, Hao Phung, Aaron Gokaslan, Volodymyr Kuleshov
Abstract: Discrete diffusion models enable parallel token sampling for faster inference than autoregressive approaches. However, prior diffusion models use a decoder-only architecture, which requires sampling algorithms that invoke the full network at every denoising step and incur high computational cost. Our key insight is that discrete diffusion models perform two types of computation: 1) representing clean tokens and 2) denoising corrupted tokens, which enables us to use separate modules for each task. We propose an encoder-decoder architecture to accelerate discrete diffusion inference, which relies on an encoder to represent clean tokens and a lightweight decoder to iteratively refine a noised sequence. We also show that this architecture enables faster training of block diffusion models, which partition sequences into blocks for better quality and are commonly used in diffusion language model inference. We introduce a framework for Efficient Encoder-Decoder Diffusion (E2D2), consisting of an architecture with specialized training and sampling algorithms, and we show that E2D2 achieves superior trade-offs between generation quality and inference throughput on summarization, translation, and mathematical reasoning tasks. We provide the code, model weights, and blog post on the project page: https://m-arriola.com/e2d2
Authors: Yugong Zeng, Jonathan Wu
Abstract: Precipitation prediction has undergone a profound transformation -- from early symbolic and empirical methods rooted in divination and observation, to modern technologies based on atmospheric physics and artificial intelligence. This review traces the historical and technological evolution of precipitation forecasting, presenting a survey about end-to-end precipitation prediction technologies that spans ancient practices, the foundations of meteorological science, the rise of numerical weather prediction (NWP), and the emergence of machine learning (ML) and deep learning (DL) models. We first explore traditional and indigenous forecasting methods, then describe the development of physical modeling and statistical frameworks that underpin contemporary operational forecasting. Particular emphasis is placed on recent advances in neural network-based approaches, including automated deep learning, interpretability-driven design, and hybrid physical-data models. By compositing research across multiple eras and paradigms, this review not only depicts the history of end-to-end precipitation prediction but also outlines future directions in next generation forecasting systems.
Authors: Kaitong Cai, Jusheng Zhang, Jing Yang, Keze Wang
Abstract: Hybrid offline--online reinforcement learning (O2O RL) promises both sample efficiency and robust exploration, but suffers from instability due to distribution shift between offline and online data. We introduce RLPD-GX, a framework that decouples policy optimization from safety enforcement: a reward-seeking learner explores freely, while a projection-based guardian guarantees rule-consistent execution and safe value backups. This design preserves the exploratory value of online interactions without collapsing to conservative policies. To further stabilize training, we propose dynamic curricula that gradually extend temporal horizons and anneal offline--online data mixing. We prove convergence via a contraction property of the guarded Bellman operator, and empirically show state-of-the-art performance on Atari-100k, achieving a normalized mean score of 3.02 (+45\% over prior hybrid methods) with stronger safety and stability. Beyond Atari, ablations demonstrate consistent gains across safety-critical and long-horizon tasks, underscoring the generality of our design. Extensive and comprehensive results highlight decoupled safety enforcement as a simple yet principled route to robust O2O RL, suggesting a broader paradigm for reconciling exploration and safety in reinforcement learning.
Authors: Amirali Ataee Naeini, Arshia Ataee Naeini, Fatemeh Karami Mohammadi, Omid Ghaffarpasand
Abstract: Reliable long-term forecasting of PM2.5 concentrations is critical for public health early-warning systems, yet existing deep learning approaches struggle to maintain prediction stability beyond 48 hours, especially in cities with sparse monitoring networks. This paper presents a deep learning framework that combines Dynamic Time Warping (DTW) for intelligent station similarity selection with a CNN-GRU architecture to enable extended-horizon PM2.5 forecasting in Isfahan, Iran, a city characterized by complex pollution dynamics and limited monitoring coverage. Unlike existing approaches that rely on computationally intensive transformer models or external simulation tools, our method integrates three key innovations: (i) DTW-based historical sampling to identify similar pollution patterns across peer stations, (ii) a lightweight CNN-GRU architecture augmented with meteorological features, and (iii) a scalable design optimized for sparse networks. Experimental validation using multi-year hourly data from eight monitoring stations demonstrates superior performance compared to state-of-the-art deep learning methods, achieving R2 = 0.91 for 24-hour forecasts. Notably, this is the first study to demonstrate stable 10-day PM2.5 forecasting (R2 = 0.73 at 240 hours) without performance degradation, addressing critical early-warning system requirements. The framework's computational efficiency and independence from external tools make it particularly suitable for deployment in resource-constrained urban environments.
Authors: Nicholas I-Hsien Kuo, Blanca Gallego, Louisa Jorm
Abstract: Foundation models refer to architectures trained on vast datasets using autoregressive pre-training from natural language processing to capture intricate patterns and motifs. They were originally developed to transfer such learned knowledge to downstream predictive tasks. Recently, however, some studies repurpose these learned representations for phenotype discovery without rigorous validation, risking superficially realistic but clinically incoherent embeddings. To test this mismatch, we trained two autoregressive models -- a sequence-to-sequence LSTM and a reduced Transformer -- on longitudinal ART for HIV and Acute Hypotension datasets. Controlled irregularity was added during training via random inter-visit gaps, while test sequences stayed complete. Patient-trajectory synthesis evaluated distributional and correlational fidelity. Both reproduced feature distributions but failed to preserve cross-feature structure -- showing that generative pre-training yields local realism but limited clinical coherence. These results highlight the need for domain-specific evaluation and support trajectory synthesis as a practical probe before fine-tuning or deployment.
Authors: Duong M. Nguyen, Trong Nghia Hoang, Thanh Trung Huynh, Quoc Viet Hung Nguyen, Phi Le Nguyen
Abstract: Multimodal federated learning in real-world settings often encounters incomplete and heterogeneous data across clients. This results in misaligned local feature representations that limit the effectiveness of model aggregation. Unlike prior work that assumes either differing modality sets without missing input features or a shared modality set with missing features across clients, we consider a more general and realistic setting where each client observes a different subset of modalities and might also have missing input features within each modality. To address the resulting misalignment in learned representations, we propose a new federated learning framework featuring locally adaptive representations based on learnable client-side embedding controls that encode each client's data-missing patterns. These embeddings serve as reconfiguration signals that align the globally aggregated representation with each client's local context, enabling more effective use of shared information. Furthermore, the embedding controls can be algorithmically aggregated across clients with similar data-missing patterns to enhance the robustness of reconfiguration signals in adapting the global representation. Empirical results on multiple federated multimodal benchmarks with diverse data-missing patterns across clients demonstrate the efficacy of the proposed method, achieving up to 36.45\% performance improvement under severe data incompleteness. The method is also supported by a theoretical analysis with an explicit performance bound that matches our empirical observations. Our source codes are provided at https://github.com/nmduonggg/PEPSY
Authors: Saeed Najafi, Alona Fyshe
Abstract: Aligning Large Language Models (LLMs) with human preferences is crucial, but standard methods like Reinforcement Learning from Human Feedback (RLHF) are often complex and unstable. In this work, we propose a new, simpler approach that recasts alignment through the lens of Maximum Marginal Likelihood (MML) estimation. Our new MML based Preference Optimization (MMPO) maximizes the marginal log-likelihood of a preferred text output, using the preference pair as samples for approximation, and forgoes the need for both an explicit reward model and entropy maximization. We theoretically demonstrate that MMPO implicitly performs preference optimization, producing a weighted gradient that naturally up-weights chosen responses over rejected ones. Across models ranging from 135M to 8B parameters, we empirically show that MMPO: 1) is more stable with respect to the hyperparameter $\beta$ compared to alternative baselines, and 2) achieves competitive or superior preference alignment while better preserving the base model's general language capabilities. Through a series of ablation experiments, we show that this improved performance is indeed attributable to MMPO's implicit preference optimization within the gradient updates.
Authors: Yuqin Jiang, Song Jiang, Jacob Algrim, Trevor Harms, Maxwell Koenen, Xinya Lan, Xingyu Li, Chun-Han Lin, Jia Liu, Jiayang Sun, Henry Zenger
Abstract: Linguistic Landscape (LL) research traditionally relies on manual photography and annotation of public signages to examine distribution of languages in urban space. While such methods yield valuable findings, the process is time-consuming and difficult for large study areas. This study explores the use of AI powered language detection method to automate LL analysis. Using Honolulu Chinatown as a case study, we constructed a georeferenced photo dataset of 1,449 images collected by researchers and applied AI for optical character recognition (OCR) and language classification. We also conducted manual validations for accuracy checking. This model achieved an overall accuracy of 79%. Five recurring types of mislabeling were identified, including distortion, reflection, degraded surface, graffiti, and hallucination. The analysis also reveals that the AI model treats all regions of an image equally, detecting peripheral or background texts that human interpreters typically ignore. Despite these limitations, the results demonstrate the potential of integrating AI-assisted workflows into LL research to reduce such time-consuming processes. However, due to all the limitations and mis-labels, we recognize that AI cannot be fully trusted during this process. This paper encourages a hybrid approach combining AI automation with human validation for a more reliable and efficient workflow.
Authors: Darshana Priyasad, Tharindu Fernando, Maryam Haghighat, Harshala Gammulle, Clinton Fookes
Abstract: Natural disasters, such as volcanic eruptions, pose significant challenges to daily life and incur considerable global economic losses. The emergence of next-generation small-satellites, capable of constellation-based operations, offers unparalleled opportunities for near-real-time monitoring and onboard processing of such events. However, a major bottleneck remains the lack of extensive annotated datasets capturing volcanic activity, which hinders the development of robust detection systems. This paper introduces a novel dataset explicitly designed for volcanic activity and eruption detection, encompassing diverse volcanoes worldwide. The dataset provides binary annotations to identify volcanic anomalies or non-anomalies, covering phenomena such as temperature anomalies, eruptions, and volcanic ash emissions. These annotations offer a foundational resource for developing and evaluating detection models, addressing a critical gap in volcanic monitoring research. Additionally, we present comprehensive benchmarks using state-of-the-art models to establish baselines for future studies. Furthermore, we explore the potential for deploying these models onboard next-generation satellites. Using the Intel Movidius Myriad X VPU as a testbed, we demonstrate the feasibility of volcanic activity detection directly onboard. This capability significantly reduces latency and enhances response times, paving the way for advanced early warning systems. This paves the way for innovative solutions in volcanic disaster management, encouraging further exploration and refinement of onboard monitoring technologies.
Authors: Vaibhav Raj, Indradyumna Roy, Ashwin Ramachandran, Soumen Chakrabarti, Abir De
Abstract: Subgraph matching is vital in knowledge graph (KG) question answering, molecule design, scene graph, code and circuit search, etc. Neural methods have shown promising results for subgraph matching. Our study of recent systems suggests refactoring them into a unified design space for graph matching networks. Existing methods occupy only a few isolated patches in this space, which remains largely uncharted. We undertake the first comprehensive exploration of this space, featuring such axes as attention-based vs. soft permutation-based interaction between query and corpus graphs, aligning nodes vs. edges, and the form of the final scoring network that integrates neural representations of the graphs. Our extensive experiments reveal that judicious and hitherto-unexplored combinations of choices in this space lead to large performance benefits. Beyond better performance, our study uncovers valuable insights and establishes general design principles for neural graph representation and interaction, which may be of wider interest.
Authors: Andreas Floros, Seyed-Mohsen Moosavi-Dezfooli, Pier Luigi Dragotti
Abstract: We investigate the role of network architecture in shaping the inductive biases of modern score-based generative models. To this end, we introduce the Score Anisotropy Directions (SADs), architecture-dependent directions that reveal how different networks preferentially capture data structure. Our analysis shows that SADs form adaptive bases aligned with the architecture's output geometry, providing a principled way to predict generalization ability in score models prior to training. Through both synthetic data and standard image benchmarks, we demonstrate that SADs reliably capture fine-grained model behavior and correlate with downstream performance, as measured by Wasserstein metrics. Our work offers a new lens for explaining and predicting directional biases of generative models.
Authors: Daniel Sin, Milad Toutounchian
Abstract: In our article, we describe a method for generating counterfactual explanations in high-dimensional spaces using four steps that involve fitting our dataset to a model, finding the decision boundary, determining constraints on the problem, and computing the closest point (counterfactual explanation) from that boundary. We propose a discretized approach where we find many discrete points on the boundary and then identify the closest feasible counterfactual explanation. This method, which we later call $\textit{Segmented Sampling for Boundary Approximation}$ (SSBA), applies binary search to find decision boundary points and then searches for the closest boundary point. Across four datasets of varying dimensionality, we show that our method can outperform current methods for counterfactual generation with reductions in distance between $5\%$ to $50\%$ in terms of the $L_2$ norm. Our method can also handle real-world constraints by restricting changes to immutable and categorical features, such as age, gender, sex, height, and other related characteristics such as the case for a health-based dataset. In terms of runtime, the SSBA algorithm generates decision boundary points on multiple orders of magnitude in the same given time when we compare to a grid-based approach. In general, our method provides a simple and effective model-agnostic method that can compute nearest feasible (i.e. realistic with constraints) counterfactual explanations. All of our results and our code can be found here at this link: $\href{https://github.com/dsin85691/SSBA_For_Counterfactuals}{https://github.com/ dsin85691/SSBA\_For\_Counterfactuals}$
URLs: https://github.com/dsin85691/SSBA_For_Counterfactuals, https://github.com/
Authors: Huaisheng Zhu, Zhengyu Chen, Shijie Zhou, Zhihui Xie, Yige Yuan, Zhimeng Guo, Siyuan Xu, Hangfan Zhang, Vasant Honavar, Teng Xiao
Abstract: Diffusion models have recently been extended to language generation through Masked Diffusion Language Models (MDLMs), which achieve performance competitive with strong autoregressive models. However, MDLMs tend to degrade in the few-step regime and cannot directly adopt existing few-step distillation methods designed for continuous diffusion models, as they lack the intrinsic property of mapping from noise to data. Recent Uniform-state Diffusion Models (USDMs), initialized from a uniform prior, alleviate some limitations but still suffer from complex loss formulations that hinder scalability. In this work, we propose a simplified denoising-based loss for USDMs that optimizes only noise-replaced tokens, stabilizing training and matching ELBO-level performance. Furthermore, by framing denoising as self-supervised learning, we introduce a simple modification to our denoising loss with contrastive-inspired negative gradients, which is practical and yield additional improvements in generation quality.
Authors: Mingze Gong, Juan Du, Jianbang You
Abstract: Anomaly detection in complex, high-dimensional data, such as UAV sensor readings, is essential for operational safety but challenging for existing methods due to their limited sensitivity, scalability, and inability to capture intricate dependencies. We propose the Diffuse to Detect (DTD) framework, a novel approach that innovatively adapts diffusion models for anomaly detection, diverging from their conventional use in generative tasks with high inference time. By comparison, DTD employs a single-step diffusion process to predict noise patterns, enabling rapid and precise identification of anomalies without reconstruction errors. This approach is grounded in robust theoretical foundations that link noise prediction to the data distribution's score function, ensuring reliable deviation detection. By integrating Graph Neural Networks to model sensor relationships as dynamic graphs, DTD effectively captures spatial (inter-sensor) and temporal anomalies. Its two-branch architecture, with parametric neural network-based energy scoring for scalability and nonparametric statistical methods for interpretability, provides flexible trade-offs between computational efficiency and transparency. Extensive evaluations on UAV sensor data, multivariate time series, and images demonstrate DTD's superior performance over existing methods, underscoring its generality across diverse data modalities. This versatility, combined with its adaptability, positions DTD as a transformative solution for safety-critical applications, including industrial monitoring and beyond.
Authors: Xiaofan Zhou, Lu Cheng
Abstract: Continual Learning (CL) is essential for enabling self-evolving large language models (LLMs) to adapt and remain effective amid rapid knowledge growth. Yet, despite its importance, little attention has been given to establishing statistical reliability guarantees for LLMs under CL, particularly in the setting of continual domain pretraining (CDP). Conformal Prediction (CP) has shown promise in offering correctness guarantees for LLMs, but it faces major challenges in CDP: testing data often stems from unknown or shifting domain distributions, under which CP may no longer provide valid guarantees. Moreover, when high coverage is required, CP can yield excessively large prediction sets for unanswerable queries, reducing informativeness. To address these challenges, we introduce an adaptive rejection and non-exchangeable CP framework. Our method first estimates the distribution of questions across domains in the test set using transformer-based clustering, then reweights or resamples the calibration data accordingly. Building on this, adaptive rejection CP allows the LLM to selectively abstain from answering when its confidence or competence shifts significantly. Extensive experiments demonstrate that our framework enhances both the effectiveness and reliability of CP under CDP scenarios. Our code is available at: https://anonymous.4open.science/r/CPCL-8C12/
Authors: Judah Goldfeder, Matthew So, Hod Lipson
Abstract: Auxiliary Learning (AL) is a special case of Multi-task Learning (MTL) in which a network trains on auxiliary tasks to improve performance on its main task. This technique is used to improve generalization and, ultimately, performance on the network's main task. AL has been demonstrated to improve performance across multiple domains, including navigation, image classification, and natural language processing. One weakness of AL is the need for labeled auxiliary tasks, which can require human effort and domain expertise to generate. Meta Learning techniques have been used to solve this issue by learning an additional auxiliary task generation network that can create helpful tasks for the primary network. The most prominent techniques rely on Bi-Level Optimization, which incurs computational cost and increased code complexity. To avoid the need for Bi-Level Optimization, we present an RL-based approach to dynamically create auxiliary tasks. In this framework, an RL agent is tasked with selecting auxiliary labels for every data point in a training set. The agent is rewarded when their selection improves the performance on the primary task. We also experiment with learning optimal strategies for weighing the auxiliary loss per data point. On the 20-Superclass CIFAR100 problem, our RL approach outperforms human-labeled auxiliary tasks and performs as well as a prominent Bi-Level Optimization technique. Our weight learning approaches significantly outperform all of these benchmarks. For example, a Weight-Aware RL-based approach helps the VGG16 architecture achieve 80.9% test accuracy while the human-labeled auxiliary task setup achieved 75.53%. The goal of this work is to (1) prove that RL is a viable approach to dynamically generate auxiliary tasks and (2) demonstrate that per-sample auxiliary task weights can be learned alongside the auxiliary task labels and can achieve strong results.
Authors: Zhenglai Shen, Hongyu Zhou
Abstract: Compounding climate hazards, such as wildfire-induced outages and urban heatwaves, challenge the stability and equity of cities. We present a Hazard-Responsive Digital Twin (H-RDT) that combines physics-informed neural network modeling, multimodal data fusion, and equity-aware risk analytics for urban-scale response. In a synthetic district with diverse building archetypes and populations, a simulated wildfire-outage-heatwave cascade shows that H-RDT maintains stable indoor temperature predictions (approximately 31 to 33 C) under partial sensor loss, reproducing outage-driven surges and recovery. The reinforcement learning based fusion module adaptively reweights IoT, UAV, and satellite inputs to sustain spatiotemporal coverage, while the equity-adjusted mapping isolates high-vulnerability clusters (schools, clinics, low-income housing). Prospective interventions, such as preemptive cooling-center activation and microgrid sharing, reduce population-weighted thermal risk by 11 to 13 percent, shrink the 95th-percentile (tail) risk by 7 to 17 percent, and cut overheating hours by up to 9 percent. Beyond the synthetic demonstration, the framework establishes a transferable foundation for real-city implementation, linking physical hazard modeling with social equity and decision intelligence. The H-RDT advances digital urban resilience toward adaptive, learning-based, and equity-centered decision support for climate adaptation.
Authors: Paul Schwerdtner, Jules Berman, Benjamin Peherstorfer
Abstract: Deep neural networks using state space models as layers are well suited for long-range sequence tasks but can be challenging to compress after training. We use that regularizing the sum of Hankel singular values of state space models leads to a fast decay of these singular values and thus to compressible models. To make the proposed Hankel singular value regularization scalable, we develop an algorithm to efficiently compute the Hankel singular values during training iterations by exploiting the specific block-diagonal structure of the system matrices that is we use in our state space model parametrization. Experiments on Long Range Arena benchmarks demonstrate that the regularized state space layers are up to 10$\times$ more compressible than standard state space layers while maintaining high accuracy.
Authors: Mohammad Tariqul Islam, Du Liu, Deblina Sarkar
Abstract: Centered kernel alignment (CKA) is a popular metric for comparing representations, determining equivalence of networks, and neuroscience research. However, CKA does not account for the underlying manifold and relies on numerous heuristics that cause it to behave differently at different scales of data. In this work, we propose Manifold approximated Kernel Alignment (MKA), which incorporates manifold geometry into the alignment task. We derive a theoretical framework for MKA. We perform empirical evaluations on synthetic datasets and real-world examples to characterize and compare MKA to its contemporaries. Our findings suggest that manifold-aware kernel alignment provides a more robust foundation for measuring representations, with potential applications in representation learning.
Authors: Junhao Fan, Wenrui Liang, Wei-Qiang Zhang
Abstract: Accurate prediction of remaining useful life (RUL) is essential to enhance system reliability and reduce maintenance risk. Yet many strong contemporary models are fragile around fault onset and opaque to engineers: short, high-energy spikes are smoothed away or misread, fixed thresholds blunt sensitivity, and physics-based explanations are scarce. To remedy this, we introduce SARNet (Spike-Aware Consecutive Validation Framework), which builds on a Modern Temporal Convolutional Network (ModernTCN) and adds spike-aware detection to provide physics-informed interpretability. ModernTCN forecasts degradation-sensitive indicators; an adaptive consecutive threshold validates true spikes while suppressing noise. Failure-prone segments then receive targeted feature engineering (spectral slopes, statistical derivatives, energy ratios), and the final RUL is produced by a stacked RF--LGBM regressor. Across benchmark-ported datasets under an event-triggered protocol, SARNet consistently lowers error compared to recent baselines (RMSE 0.0365, MAE 0.0204) while remaining lightweight, robust, and easy to deploy.
Authors: Chenlong Yin, Zeyang Sha, Shiwen Cui, Changhua Meng
Abstract: Enhancing the reasoning capabilities of Large Language Models (LLMs) is a key strategy for building Agents that "think then act." However, recent observations, like OpenAI's o3, suggest a paradox: stronger reasoning often coincides with increased hallucination, yet no prior work has systematically examined whether reasoning enhancement itself causes tool hallucination. To address this gap, we pose the central question: Does strengthening reasoning increase tool hallucination? To answer this, we introduce SimpleToolHalluBench, a diagnostic benchmark measuring tool hallucination in two failure modes: (i) no tool available, and (ii) only distractor tools available. Through controlled experiments, we establish three key findings. First, we demonstrate a causal relationship: progressively enhancing reasoning through RL increases tool hallucination proportionally with task performance gains. Second, this effect transcends overfitting - training on non-tool tasks (e.g., mathematics) still amplifies subsequent tool hallucination. Third, the effect is method-agnostic, appearing when reasoning is instilled via supervised fine-tuning and when it is merely elicited at inference by switching from direct answers to step-by-step thinking. We also evaluate mitigation strategies including Prompt Engineering and Direct Preference Optimization (DPO), revealing a fundamental reliability-capability trade-off: reducing hallucination consistently degrades utility. Mechanistically, Reasoning RL disproportionately collapses tool-reliability-related representations, and hallucinations surface as amplified divergences concentrated in late-layer residual streams. These findings reveal that current reasoning enhancement methods inherently amplify tool hallucination, highlighting the need for new training objectives that jointly optimize for capability and reliability.
Authors: Bhavya Vasudeva, Puneesh Deora, Yize Zhao, Vatsal Sharan, Christos Thrampoulidis
Abstract: The growing adoption of spectrum-aware matrix-valued optimizers such as Muon and Shampoo in deep learning motivates a systematic study of their generalization properties and, in particular, when they might outperform competitive algorithms. We approach this question by introducing appropriate simplifying abstractions as follows: First, we use imbalanced data as a testbed. Second, we study the canonical form of such optimizers, which is Spectral Gradient Descent (SpecGD) -- each update step is $UV^T$ where $U\Sigma V^T$ is the truncated SVD of the gradient. Third, within this framework we identify a canonical setting for which we precisely quantify when SpecGD outperforms vanilla Euclidean GD. For a Gaussian mixture data model and both linear and bilinear models, we show that unlike GD, which prioritizes learning dominant principal components of the data first, SpecGD learns all principal components of the data at equal rates. We demonstrate how this translates to a growing gap in balanced accuracy favoring SpecGD early in training and further show that the gap remains consistent even when the GD counterpart uses adaptive step-sizes via normalization. By extending the analysis to deep linear models, we show that depth amplifies these effects. We empirically verify our theoretical findings on a variety of imbalanced datasets. Our experiments compare practical variants of spectral methods, like Muon and Shampoo, against their Euclidean counterparts and Adam. The results validate our findings that these spectral optimizers achieve superior generalization by promoting a more balanced learning of the data's underlying components.
Authors: Guanchen Du, Jianlong Xu, Mingtong Li, Ruiqi Wang, Qianqing Guo, Caiyi Chen, Qingcao Dai, Yuxiang Zeng
Abstract: With the rapid advancement of internet technologies, network services have become critical for delivering diverse and reliable applications to users. However, the exponential growth in the number of available services has resulted in many similar offerings, posing significant challenges in selecting optimal services. Predicting Quality of Service (QoS) accurately thus becomes a fundamental prerequisite for ensuring reliability and user satisfaction. However, existing QoS prediction methods often fail to capture rich contextual information and exhibit poor performance under extreme data sparsity and structural noise. To bridge this gap, we propose a novel architecture, QoSMGAA, specifically designed to enhance prediction accuracy in complex and noisy network service environments. QoSMGAA integrates a multi-order attention mechanism to aggregate extensive contextual data and predict missing QoS values effectively. Additionally, our method incorporates adversarial neural networks to perform autoregressive supervised learning based on transformed interaction matrices. To capture complex, higher-order interactions among users and services, we employ a discrete sampling technique leveraging the Gumbel-Softmax method to generate informative negative samples. Comprehensive experimental validation conducted on large-scale real-world datasets demonstrates that our proposed model significantly outperforms existing baseline methods, highlighting its strong potential for practical deployment in service selection and recommendation scenarios.
Authors: Chankyo Kim (University of Michigan), Sicheng Zhao (University of Michigan), Minghan Zhu (University of Michigan, University of Pennsylvania), Tzu-Yuan Lin (Massachusetts Institute of Technology), Maani Ghaffari (University of Michigan)
Abstract: Encoding symmetries is a powerful inductive bias for improving the generalization of deep neural networks. However, most existing equivariant models are limited to simple symmetries like rotations, failing to address the broader class of general linear transformations, GL(n), that appear in many scientific domains. We introduce Reductive Lie Neurons (ReLNs), a novel neural network architecture exactly equivariant to these general linear symmetries. ReLNs are designed to operate directly on a wide range of structured inputs, including general n-by-n matrices. ReLNs introduce a novel adjoint-invariant bilinear layer to achieve stable equivariance for both Lie-algebraic features and matrix-valued inputs, without requiring redesign for each subgroup. This architecture overcomes the limitations of prior equivariant networks that only apply to compact groups or simple vector data. We validate ReLNs' versatility across a spectrum of tasks: they outperform existing methods on algebraic benchmarks with sl(3) and sp(4) symmetries and achieve competitive results on a Lorentz-equivariant particle physics task. In 3D drone state estimation with geometric uncertainty, ReLNs jointly process velocities and covariances, yielding significant improvements in trajectory accuracy. ReLNs provide a practical and general framework for learning with broad linear group symmetries on Lie algebras and matrix-valued data. Project page: https://reductive-lie-neuron.github.io/
Authors: Dimitris Bertsimas, Yubing Cui
Abstract: Random Forests (RF) and Extreme Gradient Boosting (XGBoost) are two of the most widely used and highly performing classification and regression models. They aggregate equally weighted CART trees, generated randomly in RF or sequentially in XGBoost. In this paper, we propose Adaptive Forests (AF), a novel approach that adaptively selects the weights of the underlying CART models. AF combines (a) the Optimal Predictive-Policy Trees (OP2T) framework to prescribe tailored, input-dependent unequal weights to trees and (b) Mixed Integer Optimization (MIO) to refine weight candidates dynamically, enhancing overall performance. We demonstrate that AF consistently outperforms RF, XGBoost, and other weighted RF in binary and multi-class classification problems over 20+ real-world datasets.
Authors: Zidong Liu, Zhuoyan Xu, Zhenmei Shi, Yingyu Liang
Abstract: Composing basic skills from simple tasks to accomplish composite tasks is crucial for modern intelligent systems. We investigate the in-context composition ability of language models to perform composite tasks that combine basic skills demonstrated in in-context examples. This is more challenging than the standard setting, where skills and their composition can be learned in training. We conduct systematic experiments on various representative open-source language models, utilizing linguistic and logical tasks designed to probe composition abilities. The results reveal that simple task examples can have a surprising negative impact on the performance, because the models generally struggle to recognize and assemble the skills correctly, even with Chain-of-Thought examples. Theoretical analysis further shows that it is crucial to align examples with the corresponding steps in the composition. This inspires a method for the probing tasks, whose improved performance provides positive support for our insights.
Authors: Pravin Nair
Abstract: The softmax function is a basic operator in machine learning and optimization, used in classification, attention mechanisms, reinforcement learning, game theory, and problems involving log-sum-exp terms. Existing robustness guarantees of learning models and convergence analysis of optimization algorithms typically consider the softmax operator to have a Lipschitz constant of $1$ with respect to the $\ell_2$ norm. In this work, we prove that the softmax function is contractive with the Lipschitz constant $1/2$, uniformly across all $\ell_p$ norms with $p \ge 1$. We also show that the local Lipschitz constant of softmax attains $1/2$ for $p = 1$ and $p = \infty$, and for $p \in (1,\infty)$, the constant remains strictly below $1/2$ and the supremum $1/2$ is achieved only in the limit. To our knowledge, this is the first comprehensive norm-uniform analysis of softmax Lipschitz continuity. We demonstrate how the sharper constant directly improves a range of existing theoretical results on robustness and convergence. We further validate the sharpness of the $1/2$ Lipschitz constant of the softmax operator through empirical studies on attention-based architectures (ViT, GPT-2, Qwen3-8B) and on stochastic policies in reinforcement learning.
Authors: Han Wu, Jie Yin
Abstract: Few-shot knowledge graph relational learning seeks to perform reasoning over relations given only a limited number of training examples. While existing approaches largely adopt a meta-learning framework for enabling fast adaptation to new relations, they suffer from two key pitfalls. First, they learn relation meta-knowledge in isolation, failing to capture common relational patterns shared across tasks. Second, they struggle to effectively incorporate local, task-specific contexts crucial for rapid adaptation. To address these limitations, we propose MoEMeta, a novel meta-learning framework that disentangles globally shared knowledge from task-specific contexts to enable both effective generalization and rapid adaptation. MoEMeta introduces two key innovations: (i) a mixture-of-experts (MoE) model that learns globally shared relational prototypes to enhance generalization, and (ii) a task-tailored adaptation mechanism that captures local contexts for fast task-specific adaptation. By balancing global generalization with local adaptability, MoEMeta significantly advances few-shot relational learning. Extensive experiments and analyses on three KG benchmarks demonstrate that MoEMeta consistently outperforms existing baselines, achieving state-of-the-art performance.
Authors: Gurpreet Singh, Keshav Sood, P. Rajalakshmi, Yong Xiang
Abstract: Federated learning (FL) offers a privacy-preserving paradigm for machine learning, but its application in intrusion detection systems (IDS) within IoT networks is challenged by severe class imbalance, non-IID data, and high communication overhead.These challenges severely degrade the performance of conventional FL methods in real-world network traffic classification. To overcome these limitations, we propose Sentinel, a personalized federated IDS (pFed-IDS) framework that incorporates a dual-model architecture on each client, consisting of a personalized teacher and a lightweight shared student model. This design effectively balances deep local adaptation with efficient global model consensus while preserving client privacy by transmitting only the compact student model, thus reducing communication costs. Sentinel integrates three key mechanisms to ensure robust performance: bidirectional knowledge distillation with adaptive temperature scaling, multi-faceted feature alignment, and class-balanced loss functions. Furthermore, the server employs normalized gradient aggregation with equal client weighting to enhance fairness and mitigate client drift. Extensive experiments on the IoTID20 and 5GNIDD benchmark datasets demonstrate that Sentinel significantly outperforms state-of-the-art federated methods, establishing a new performance benchmark, especially under extreme data heterogeneity, while maintaining communication efficiency.
Authors: Di Zhang, Xun Wu, Shaohan Huang, Yaru Hao, Li Dong, Zewen Chi, Zhifang Sui, Furu Wei
Abstract: Recent advances in reinforcement learning (RL) have substantially improved the training of large-scale language models, leading to significant gains in generation quality and reasoning ability. However, most existing research focuses on dense models, while RL training for Mixture-of-Experts (MoE) architectures remains underexplored. To address the instability commonly observed in MoE training, we propose a novel router-aware approach to optimize importance sampling (IS) weights in off-policy RL. Specifically, we design a rescaling strategy guided by router logits, which effectively reduces gradient variance and mitigates training divergence. Experimental results demonstrate that our method significantly improves both the convergence stability and the final performance of MoE models, highlighting the potential of RL algorithmic innovations tailored to MoE architectures and providing a promising direction for efficient training of large-scale expert models.
Authors: Ved Danait, Srijan Das, Sujoy Bhore
Abstract: Approximate Nearest Neighbor (ANN) search and Approximate Kernel Density Estimation (A-KDE) are fundamental problems at the core of modern machine learning, with broad applications in data analysis, information systems, and large-scale decision making. In massive and dynamic data streams, a central challenge is to design compact sketches that preserve essential structural properties of the data while enabling efficient queries. In this work, we develop new sketching algorithms that achieve sublinear space and query time guarantees for both ANN and A-KDE for a dynamic stream of data. For ANN in the streaming model, under natural assumptions, we design a sublinear sketch that requires only $\mathcal{O}(n^{1+\rho-\eta})$ memory by storing only a sublinear ($n^{-\eta}$) fraction of the total inputs, where $\rho$ is a parameter of the LSH family, and $0<\eta<1$. Our method supports sublinear query time, batch queries, and extends to the more general Turnstile model. While earlier works have focused on Exact NN, this is the first result on ANN that achieves near-optimal trade-offs between memory size and approximation error. Next, for A-KDE in the Sliding-Window model, we propose a sketch of size $\mathcal{O}\left(RW \cdot \frac{1}{\sqrt{1+\epsilon} - 1} \log^2 N\right)$, where $R$ is the number of sketch rows, $W$ is the LSH range, $N$ is the window size, and $\epsilon$ is the approximation error. This, to the best of our knowledge, is the first theoretical sublinear sketch guarantee for A-KDE in the Sliding-Window model. We complement our theoretical results with experiments on various real-world datasets, which show that the proposed sketches are lightweight and achieve consistently low error in practice.
Authors: Subhojyoti Khastagir, Kishalay Das, Pawan Goyal, Seung-Cheol Lee, Satadeep Bhattacharjee, Niloy Ganguly
Abstract: Recent advances in generative modeling have shown significant promise in designing novel periodic crystal structures. Existing approaches typically rely on either large language models (LLMs) or equivariant denoising models, each with complementary strengths: LLMs excel at handling discrete atomic types but often struggle with continuous features such as atomic positions and lattice parameters, while denoising models are effective at modeling continuous variables but encounter difficulties in generating accurate atomic compositions. To bridge this gap, we propose CrysLLMGen, a hybrid framework that integrates an LLM with a diffusion model to leverage their complementary strengths for crystal material generation. During sampling, CrysLLMGen first employs a fine-tuned LLM to produce an intermediate representation of atom types, atomic coordinates, and lattice structure. While retaining the predicted atom types, it passes the atomic coordinates and lattice structure to a pre-trained equivariant diffusion model for refinement. Our framework outperforms state-of-the-art generative models across several benchmark tasks and datasets. Specifically, CrysLLMGen not only achieves a balanced performance in terms of structural and compositional validity but also generates more stable and novel materials compared to LLM-based and denoisingbased models Furthermore, CrysLLMGen exhibits strong conditional generation capabilities, effectively producing materials that satisfy user-defined constraints. Code is available at https://github.com/kdmsit/crysllmgen
Authors: Christos Thrampoulidis, Sadegh Mahdavi, Wenlong Deng
Abstract: This note reconciles two seemingly distinct approaches to policy gradient optimization for the Pass@K objective in reinforcement learning with verifiable rewards: (1) direct REINFORCE-style methods, and (2) advantage-shaping techniques that directly modify GRPO. We show that these are two sides of the same coin. By reverse-engineering existing advantage-shaping algorithms, we reveal that they implicitly optimize surrogate rewards. We specifically interpret practical ``hard-example up-weighting'' modifications to GRPO as reward-level regularization. Conversely, starting from surrogate reward objectives, we provide a simple recipe for deriving both existing and new advantage-shaping methods. This perspective provides a lens for RLVR policy gradient optimization beyond our original motivation of Pass@K.
Authors: Tengxue Zhang, Biao Ouyang, Yang Shu, Xinyang Chen, Chenjuan Guo, Bin Yang
Abstract: Pre-trained models exhibit strong generalization to various downstream tasks. However, given the numerous models available in the model hub, identifying the most suitable one by individually fine-tuning is time-consuming. In this paper, we propose \textbf{SwiftTS}, a swift selection framework for time series pre-trained models. To avoid expensive forward propagation through all candidates, SwiftTS adopts a learning-guided approach that leverages historical dataset-model performance pairs across diverse horizons to predict model performance on unseen datasets. It employs a lightweight dual-encoder architecture that embeds time series and candidate models with rich characteristics, computing patchwise compatibility scores between data and model embeddings for efficient selection. To further enhance the generalization across datasets and horizons, we introduce a horizon-adaptive expert composition module that dynamically adjusts expert weights, and the transferable cross-task learning with cross-dataset and cross-horizon task sampling to enhance out-of-distribution (OOD) robustness. Extensive experiments on 14 downstream datasets and 8 pre-trained models demonstrate that SwiftTS achieves state-of-the-art performance in time series pre-trained model selection.
Authors: Zhiyu Wang, Suman Raj, Rajkumar Buyya
Abstract: Multiple Unmanned Aerial Vehicles (UAVs) cooperative Mobile Edge Computing (MEC) systems face critical challenges in coordinating trajectory planning, task offloading, and resource allocation while ensuring Quality of Service (QoS) under dynamic and uncertain environments. Existing approaches suffer from limited scalability, slow convergence, and inefficient knowledge sharing among UAVs, particularly when handling large-scale IoT device deployments with stringent deadline constraints. This paper proposes AirFed, a novel federated graph-enhanced multi-agent reinforcement learning framework that addresses these challenges through three key innovations. First, we design dual-layer dynamic Graph Attention Networks (GATs) that explicitly model spatial-temporal dependencies among UAVs and IoT devices, capturing both service relationships and collaborative interactions within the network topology. Second, we develop a dual-Actor single-Critic architecture that jointly optimizes continuous trajectory control and discrete task offloading decisions. Third, we propose a reputation-based decentralized federated learning mechanism with gradient-sensitive adaptive quantization, enabling efficient and robust knowledge sharing across heterogeneous UAVs. Extensive experiments demonstrate that AirFed achieves 42.9% reduction in weighted cost compared to state-of-the-art baselines, attains over 99% deadline satisfaction and 94.2% IoT device coverage rate, and reduces communication overhead by 54.5%. Scalability analysis confirms robust performance across varying UAV numbers, IoT device densities, and system scales, validating AirFed's practical applicability for large-scale UAV-MEC deployments.
Authors: Sergei Kholkin, Francisco Vargas, Alexander Korotin
Abstract: We introduce the Target Concrete Score Identity Sampler (TCSIS), a method for sampling from unnormalized densities on discrete state spaces by learning the reverse dynamics of a Continuous-Time Markov Chain (CTMC). Our approach builds on a forward in time CTMC with a uniform noising kernel and relies on the proposed Target Concrete Score Identity, which relates the concrete score, the ratio of marginal probabilities of two states, to a ratio of expectations of Boltzmann factors under the forward uniform diffusion kernel. This formulation enables Monte Carlo estimation of the concrete score without requiring samples from the target distribution or computation of the partition function. We approximate the concrete score with a neural network and propose two algorithms: Self-Normalized TCSIS and Unbiased TCSIS. Finally, we demonstrate the effectiveness of TCSIS on problems from statistical physics.
Authors: Felix Koehler, Nils Thuerey
Abstract: Neural operators or emulators for PDEs trained on data from numerical solvers are conventionally assumed to be limited by their training data's fidelity. We challenge this assumption by identifying "emulator superiority," where neural networks trained purely on low-fidelity solver data can achieve higher accuracy than those solvers when evaluated against a higher-fidelity reference. Our theoretical analysis reveals how the interplay between emulator inductive biases, training objectives, and numerical error characteristics enables superior performance during multi-step rollouts. We empirically validate this finding across different PDEs using standard neural architectures, demonstrating that emulators can implicitly learn dynamics that are more regularized or exhibit more favorable error accumulation properties than their training data, potentially surpassing training data limitations and mitigating numerical artifacts. This work prompts a re-evaluation of emulator benchmarking, suggesting neural emulators might achieve greater physical fidelity than their training source within specific operational regimes. Project Page: https://tum-pbs.github.io/emulator-superiority
Authors: Omer Jauhar Khan, Sudais Khan, Hafeez Anwar
Abstract: Physics Informed Neural Networks (PINNs) are gaining attention for their ability to embed physical laws into deep learning models, which is particularly useful in structural engineering tasks with limited data. This paper aims to explore the use of PINNs to predict the weight of small scale spaghetti bridges, a task relevant to understanding load limits and potential failure modes in simplified structural models. Our proposed framework incorporates physics-based constraints to the prediction model for improved performance. In addition to standard PINNs, we introduce a novel architecture named Physics Informed Kolmogorov Arnold Network (PIKAN), which blends universal function approximation theory with physical insights. The structural parameters provided as input to the model are collected either manually or through computer vision methods. Our dataset includes 15 real bridges, augmented to 100 samples, and our best model achieves an $R^2$ score of 0.9603 and a mean absolute error (MAE) of 10.50 units. From applied perspective, we also provide a web based interface for parameter entry and prediction. These results show that PINNs can offer reliable estimates of structural weight, even with limited data, and may help inform early stage failure analysis in lightweight bridge designs. The complete data and code are available at https://github.com/OmerJauhar/PINNS-For-Spaghetti-Bridges.
URLs: https://github.com/OmerJauhar/PINNS-For-Spaghetti-Bridges.
Authors: Juan A. Lara, David Lizcano, V\'ictor Ramp\'erez, Javier Soriano
Abstract: Outlier detection is an important problem occurring in a wide range of areas. Outliers are the outcome of fraudulent behaviour, mechanical faults, human error, or simply natural deviations. Many data mining applications perform outlier detection, often as a preliminary step in order to filter out outliers and build more representative models. In this paper, we propose an outlier detection method based on a clustering process. The aim behind the proposal outlined in this paper is to overcome the specificity of many existing outlier detection techniques that fail to take into account the inherent dispersion of domain objects. The outlier detection method is based on four criteria designed to represent how human beings (experts in each domain) visually identify outliers within a set of objects after analysing the clusters. This has an advantage over other clustering-based outlier detection techniques that are founded on a purely numerical analysis of clusters. Our proposal has been evaluated, with satisfactory results, on data (particularly time series) from two different domains: stabilometry, a branch of medicine studying balance-related functions in human beings and electroencephalography (EEG), a neurological exploration used to diagnose nervous system disorders. To validate the proposed method, we studied method outlier detection and efficiency in terms of runtime. The results of regression analyses confirm that our proposal is useful for detecting outlier data in different domains, with a false positive rate of less than 2% and a reliability greater than 99%.
Authors: Chi Liu
Abstract: We provide a new perspective on GSPO's length-normalized importance ratios by establishing their connection to information-theoretic quantities. We show that GSPO's sequence-level weight $s(\theta) = (\pi_\theta/\pi_{\theta_{\text{old}}})^{1/|y|}$ can be equivalently expressed as the inverse perplexity ratio $\text{PPL}_{\theta_{\text{old}}}/\text{PPL}_\theta$ and as the exponential cross-entropy change $\exp(\Delta H)$. While the perplexity-entropy relationship follows from standard definitions, this observation provides a useful lens for understanding GSPO: the algorithm weights policy gradient updates by perplexity ratios, offering an information-theoretic interpretation of the importance weights. This perspective helps explain GSPO's empirical properties, including log-domain variance reduction through geometric averaging and stability in training mixture-of-experts models. We validate the mathematical equivalences and variance predictions through controlled experiments on mathematical reasoning tasks.
Authors: Aryan Mathur, Asaduddin Ahmed
Abstract: Deep reinforcement learning agents often struggle when tasks require understanding both vision and language. Conventional architectures typically isolate perception (for example, CNN-based visual encoders) from decision-making (policy networks). This separation can be inefficient, since the policy's failures do not directly help the perception module learn what is important. To address this, we implement the Perception-Decision Interleaving Transformer (PDiT) architecture introduced by Mao et al. (2023), a model that alternates between perception and decision layers within a single transformer. This interleaving allows feedback from decision-making to refine perceptual features dynamically. In addition, we integrate a contrastive loss inspired by CLIP to align textual mission embeddings with visual scene features. We evaluate the PDiT encoders on the BabyAI GoToLocal environment and find that the approach achieves more stable rewards and stronger alignment compared to a standard PPO baseline. The results suggest that interleaved transformer encoders are a promising direction for developing more integrated autonomous agents.
Authors: Koki Shibata, Tianheng Ling, Chao Qian, Tomokazu Matsui, Hirohiko Suwa, Keiichi Yasumoto, Gregor Schiele
Abstract: The growing demand for smart home interfaces has increased interest in non-intrusive sensing methods like vibration-based gesture recognition. While prior studies demonstrated feasibility, they often rely on complex preprocessing and large Neural Networks (NNs) requiring costly high-performance hardware, resulting in high energy usage and limited real-world deployability. This study proposes an energy-efficient solution deploying compact NNs on low-power Field-Programmable Gate Arrays (FPGAs) to enable real-time gesture recognition with competitive accuracy. We adopt a series of optimizations: (1) We replace complex spectral preprocessing with raw waveform input, eliminating complex on-board preprocessing while reducing input size by 21x without sacrificing accuracy. (2) We design two lightweight architectures (1D-CNN and 1D-SepCNN) tailored for embedded FPGAs, reducing parameters from 369 million to as few as 216 while maintaining comparable accuracy. (3) With integer-only quantization and automated RTL generation, we achieve seamless FPGA deployment. A ping-pong buffering mechanism in 1D-SepCNN further improves deployability under tight memory constraints. (4) We extend a hardware-aware search framework to support constraint-driven model configuration selection, considering accuracy, deployability, latency, and energy consumption. Evaluated on two swipe-direction datasets with multiple users and ordinary tables, our approach achieves low-latency, energy-efficient inference on the AMD Spartan-7 XC7S25 FPGA. Under the PS data splitting setting, the selected 6-bit 1D-CNN reaches 0.970 average accuracy across users with 9.22 ms latency. The chosen 8-bit 1D-SepCNN further reduces latency to 6.83 ms (over 53x CPU speedup) with slightly lower accuracy (0.949). Both consume under 1.2 mJ per inference, demonstrating suitability for long-term edge operation.
Authors: Timo Freiesleben, Sebastian Zezulka
Abstract: Predictive benchmarking, the evaluation of machine learning models based on predictive performance and competitive ranking, is a central epistemic practice in machine learning research and an increasingly prominent method for scientific inquiry. Yet, benchmark scores alone provide at best measurements of model performance relative to an evaluation dataset and a concrete learning problem. Drawing substantial scientific inferences from the results, say about theoretical tasks like image classification, requires additional assumptions about the theoretical structure of the learning problems, evaluation functions, and data distributions. We make these assumptions explicit by developing conditions of construct validity inspired by psychological measurement theory. We examine these assumptions in practice through three case studies, each exemplifying a typical intended inference: measuring engineering progress in computer vision with ImageNet; evaluating policy-relevant weather predictions with WeatherBench; and examining limitations of the predictability of life events with the Fragile Families Challenge. Our framework clarifies the conditions under which benchmark scores can support diverse scientific claims, bringing predictive benchmarking into perspective as an epistemological practice and a key site of conceptual and theoretical reasoning in machine learning.
Authors: Etienne Goffinet, Shane Bergsma, Avraham Sheinin, Natalia Vassilieva, Shaheer Muhammad, Preslav Nakov, Gurpreet Gosal
Abstract: Continual pre-training (CPT) for domain adaptation must balance target-domain gains with stability on the base domain. Existing CPT scaling laws typically assume a fixed pre-training budget, which limits their ability to forecast adaptation outcomes for models trained at different tokens-per-parameter (PTPP). We present \emph{PTPP-aware} adaptation scaling laws that make the pre-training budget an explicit variable, enabling accurate \emph{prediction} of adaptation loss at unseen \ptpp. On a multilingual setup (English/Arabic $\rightarrow$ French), PTPP-aware formulations trained on early stages (\ptpp{}=\{15,31\}) predict target loss at \ptpp{}=279 and outperform a PTPP-agnostic \dcpt{} transfer baseline on metrics (Huber-on-log, MAE$_\mathrm{rel}$, calibration slope); full diagnostics (RMSE, MAPE) are in the appendix. Beyond forecasting, we show a practical use case: planning replay ratios and adaptation token budgets that satisfy target and forgetting constraints under compute limits.
Authors: Amal Abed, Ivan Lukic, J\"org K. H. Franke, Frank Hutter
Abstract: Large language models (LLMs) have shown impressive promise in code generation, yet their progress remains limited by the shortage of large-scale datasets that are both diverse and well-aligned with human reasoning. Most existing resources pair problems with solutions, but omit the intermediate thought process that guides coding. To close this gap, we present a scalable synthetic data generation pipeline that produces nearly 800k instruction-reasoning-code-test quadruplets. Each sample combines a task, a step-by-step reasoning trace, a working solution, and executable tests, enabling models to learn not just the what but also the how of problem solving. Our pipeline combines four key components: curated contest problems, web-mined content filtered by relevance classifiers, data expansion guided by reasoning patterns, and multi-stage execution-based validation. A genetic mutation algorithm further increases task diversity while maintaining consistency between reasoning traces and code implementations. Our key finding is that fine-tuning LLMs on this dataset yields consistent improvements on coding benchmarks. Beyond raw accuracy, reasoning-aware data can substitute for model scaling, generalize across architectures, and outperform leading open-source alternatives under identical sample budgets. Our work establishes reasoning-centered synthetic data generation as an efficient approach for advancing coding capabilities in LLMs. We publish our dataset and generation pipeline to facilitate further research.
Authors: Hong Wang, Jie Wang, Jian Luo, huanshuo dong, Yeqiu Chen, Runmin Jiang, Zhen huang
Abstract: Eigenvalue problems are among the most important topics in many scientific disciplines. With the recent surge and development of machine learning, neural eigenvalue methods have attracted significant attention as a forward pass of inference requires only a tiny fraction of the computation time compared to traditional solvers. However, a key limitation is the requirement for large amounts of labeled data in training, including operators and their eigenvalues. To tackle this limitation, we propose a novel method, named Sorting Chebyshev Subspace Filter (SCSF), which significantly accelerates eigenvalue data generation by leveraging similarities between operators -- a factor overlooked by existing methods. Specifically, SCSF employs truncated fast Fourier transform sorting to group operators with similar eigenvalue distributions and constructs a Chebyshev subspace filter that leverages eigenpairs from previously solved problems to assist in solving subsequent ones, reducing redundant computations. To the best of our knowledge, SCSF is the first method to accelerate eigenvalue data generation. Experimental results show that SCSF achieves up to a $3.5\times$ speedup compared to various numerical solvers.
Authors: Anton Savostianov, Michael T. Schaub, Benjamin Stamm
Abstract: Low-pass graph filters are fundamental for signal processing on graphs and other non-Euclidean domains. However, the computation of such filters for parametric graph families can be prohibitively expensive as computation of the corresponding low-frequency subspaces, requires the repeated solution of an eigenvalue problem. We suggest a novel algorithm of low-pass graph filter interpolation based on Riemannian interpolation in normal coordinates on the Grassmann manifold. We derive an error bound estimate for the subspace interpolation and suggest two possible applications for induced parametric graph families. First, we argue that the temporal evolution of the node features may be translated to the evolving graph topology via a similarity correction to adjust the homophily degree of the network. Second, we suggest a dot product graph family induced by a given static graph which allows to infer improved message passing scheme for node classification facilitated by the filter interpolation.
Authors: Ning Ning
Abstract: Hidden Quantum Markov Models (HQMMs) extend classical Hidden Markov Models to the quantum domain, offering a powerful probabilistic framework for modeling sequential data with quantum coherence. However, existing HQMM learning algorithms are highly sensitive to data corruption and lack mechanisms to ensure robustness under adversarial perturbations. In this work, we introduce the Adversarially Corrupted HQMM (AC-HQMM), which formalizes robustness analysis by allowing a controlled fraction of observation sequences to be adversarially corrupted. To learn AC-HQMMs, we propose the Robust Iterative Learning Algorithm (RILA), a derivative-free method that integrates a Remove Corrupted Rows by Entropy Filtering (RCR-EF) module with an iterative stochastic resampling procedure for physically valid Kraus operator updates. RILA incorporates L1-penalized likelihood objectives to enhance stability, resist overfitting, and remain effective under non-differentiable conditions. Across multiple HQMM and HMM benchmarks, RILA demonstrates superior convergence stability, corruption resilience, and preservation of physical validity compared to existing algorithms, establishing a principled and efficient approach for robust quantum sequential learning.
Authors: Qi Li, Jun Wang
Abstract: Traditional clustering algorithms often struggle with high-dimensional and non-uniformly distributed data, where low-density boundary samples are easily disturbed by neighboring clusters, leading to unstable and distorted clustering results. To address this issue, we propose a Group-driven Clustering via Gravitational Attraction and Optimization (GCAO) algorithm. GCAO introduces a group-level optimization mechanism that aggregates low-density boundary points into collaboratively moving groups, replacing the traditional point-based contraction process. By combining local density estimation with neighborhood topology, GCAO constructs effective gravitational interactions between groups and their surroundings, enhancing boundary clarity and structural consistency. Using groups as basic motion units, a gravitational contraction strategy ensures globally stable and directionally consistent convergence. Experiments on multiple high-dimensional datasets demonstrate that GCAO outperforms 11 representative clustering methods, achieving average improvements of 37.13%, 52.08%, 44.98%, and 38.81% in NMI, ARI, Homogeneity, and ACC, respectively, while maintaining competitive efficiency and scalability. These results highlight GCAO's superiority in preserving cluster integrity, enhancing boundary separability, and ensuring robust performance on complex data distributions.
Authors: F\'elix Chavelli, Paul Boniol, Micha\"el Thomazo
Abstract: Time series segmentation is a fundamental task in analyzing temporal data across various domains, from human activity recognition to energy monitoring. While numerous state-of-the-art methods have been developed to tackle this problem, the evaluation of their performance remains critically limited. Existing measures predominantly focus on change point accuracy or rely on point-based measures such as Adjusted Rand Index (ARI), which fail to capture the quality of the detected segments, ignore the nature of errors, and offer limited interpretability. In this paper, we address these shortcomings by introducing two novel evaluation measures: WARI (Weighted Adjusted Rand Index), that accounts for the position of segmentation errors, and SMS (State Matching Score), a fine-grained measure that identifies and scores four fundamental types of segmentation errors while allowing error-specific weighting. We empirically validate WARI and SMS on synthetic and real-world benchmarks, showing that they not only provide a more accurate assessment of segmentation quality but also uncover insights, such as error provenance and type, that are inaccessible with traditional measures.
Authors: Xinhai Wang, Shu Yang, Liangyu Wang, Lin Zhang, Huanyi Xie, Lijie Hu, Di Wang
Abstract: Circuit discovery, which involves identifying sparse and task-relevant subnetworks in pre-trained language models, is a cornerstone of mechanistic interpretability. Automated Circuit Discovery (ACDC) has emerged as a pivotal methodology in circuit discovery, but its application to large language models is severely limited by computational inefficiency and prohibitively high memory requirements. Although several accelerated approaches have been proposed, they primarily rely on linear approximations to ACDC, which significantly compromises analytical faithfulness. Our proposed method for accelerating automated circuit discovery, Per Attention Head Quantization (PAHQ), takes a fundamentally different approach by optimizing the efficiency of each individual patching operation. PAHQ leverages a fundamental alignment between activation patching and mixed-precision quantization (MPQ): interpretability analysis through patching essentially performs targeted ablation studies. Therefore, we can maintain high precision exclusively for investigated components while safely reducing precision elsewhere in the network. PAHQ-accelerated ACDC reduces runtime by up to 80\% and memory consumption by up to 30\% compared to unaccelerated ACDC while maintaining faithfulness. Importantly, our method readily integrates with existing edge-based circuit discovery techniques by modifying the attention computation mechanism. This training-free approach provides a practical and novel pathway for accelerating mechanistic interpretability methods. Our code is available at https://github.com/626619403/PAHQ.
Authors: Runjie Zheng, Zhen Wang, Anjie Qiao, Jiancong Xie, Jiahua Rao, Yuedong Yang
Abstract: Accurate protein function prediction requires integrating heterogeneous intrinsic signals (e.g., sequence and structure) with noisy extrinsic contexts (e.g., protein-protein interactions and GO term annotations). However, two key challenges hinder effective fusion: (i) cross-modal distributional mismatch among embeddings produced by pre-trained intrinsic encoders, and (ii) noisy relational graphs of extrinsic data that degrade GNN-based information aggregation. We propose Diffused and Aligned Multi-modal Protein Embedding (DAMPE), a unified framework that addresses these through two core mechanisms. First, we propose Optimal Transport (OT)-based representation alignment that establishes correspondence between intrinsic embedding spaces of different modalities, effectively mitigating cross-modal heterogeneity. Second, we develop a Conditional Graph Generation (CGG)-based information fusion method, where a condition encoder fuses the aligned intrinsic embeddings to provide informative cues for graph reconstruction. Meanwhile, our theoretical analysis implies that the CGG objective drives this condition encoder to absorb graph-aware knowledge into its produced protein representations. Empirically, DAMPE outperforms or matches state-of-the-art methods such as DPFunc on standard GO benchmarks, achieving AUPR gains of 0.002-0.013 pp and Fmax gains 0.004-0.007 pp. Ablation studies further show that OT-based alignment contributes 0.043-0.064 pp AUPR, while CGG-based fusion adds 0.005-0.111 pp Fmax. Overall, DAMPE offers a scalable and theoretically grounded approach for robust multi-modal protein representation learning, substantially enhancing protein function prediction.
Authors: Daiyuan Li, Shreya Arya, Robert Ghrist
Abstract: Most equivariant neural networks rely on a single global symmetry, limiting their use in domains where symmetries are instead local. We introduce Torsor CNNs, a framework for learning on graphs with local symmetries encoded as edge potentials-- group-valued transformations between neighboring coordinate frames. We establish that this geometric construction is fundamentally equivalent to the classical group synchronization problem, yielding: (1) a Torsor Convolutional Layer that is provably equivariant to local changes in coordinate frames, and (2) the frustration loss--a standalone geometric regularizer that encourages locally equivariant representations when added to any NN's training objective. The Torsor CNN framework unifies and generalizes several architectures--including classical CNNs and Gauge CNNs on manifolds-- by operating on arbitrary graphs without requiring a global coordinate system or smooth manifold structure. We establish the mathematical foundations of this framework and demonstrate its applicability to multi-view 3D recognition, where relative camera poses naturally define the required edge potentials.
Authors: Yakup Emre \c{S}ahin, Niki Kilbertus, S\"oren Becker
Abstract: We introduce MIO, a transformer-based model for inferring symbolic ordinary differential equations (ODEs) from multiple observed trajectories of a dynamical system. By combining multiple instance learning with transformer-based symbolic regression, the model effectively leverages repeated observations of the same system to learn more generalizable representations of the underlying dynamics. We investigate different instance aggregation strategies and show that even simple mean aggregation can substantially boost performance. MIO is evaluated on systems ranging from one to four dimensions and under varying noise levels, consistently outperforming existing baselines.
Authors: Francesco Innocenti
Abstract: Backpropagation (BP) is the standard algorithm for training the deep neural networks that power modern artificial intelligence including large language models. However, BP is energy inefficient and unlikely to be implemented by the brain. This thesis studies an alternative, potentially more efficient brain-inspired algorithm called predictive coding (PC). Unlike BP, PC networks (PCNs) perform inference by iterative equilibration of neuron activities before learning or weight updates. Recent work has suggested that this iterative inference procedure provides a range of benefits over BP, such as faster training. However, these advantages have not been consistently observed, the inference and learning dynamics of PCNs are still poorly understood, and deep PCNs remain practically untrainable. Here, we make significant progress towards scaling PCNs by taking a theoretical approach grounded in optimisation theory. First, we show that the learning dynamics of PC can be understood as an approximate trust-region method using second-order information, despite explicitly using only first-order local updates. Second, going beyond this approximation, we show that PC can in principle make use of arbitrarily higher-order information, such that for feedforward networks the effective landscape on which PC learns is far more benign and robust to vanishing gradients than the (mean squared error) loss landscape. Third, motivated by a study of the inference dynamics of PCNs, we propose a new parameterisation called ``$\mu$PC'', which for the first time allows stable training of 100+ layer networks with little tuning and competitive performance on simple tasks. Overall, this thesis significantly advances our fundamental understanding of the inference and learning dynamics of PCNs, while highlighting the need for future research to focus on hardware co-design if PC is to compete with BP at scale.
Authors: Mohammad Hossein Jafari Naeimi, Ali Norouzi, Athena Abdi
Abstract: This paper introduces GRAD, a real-time anomaly detection method for autonomous vehicle sensors that integrates statistical analysis and deep learning to ensure the reliability of sensor data. The proposed approach combines the Reinforced Exponential Moving Average (REMA), which adapts smoothing factors and thresholding for outlier detection, with the Multi-Stage Sliding Window (MS-SW) technique for capturing both short- and long-term patterns. These features are processed using a lightweight Gated Recurrent Unit (GRU) model, which detects and classifies anomalies based on bias types, while a recovery module restores damaged sensor data to ensure continuous system operation. GRAD has a lightweight architecture consisting of two layers of GRU with a limited number of neurons that make it appropriate for real-time applications while maintaining high detection accuracy. The GRAD framework achieved remarkable performance in anomaly detection and classification. The model demonstrated an overall F1-score of 97.6% for abnormal data and 99.4% for normal data, signifying its high accuracy in distinguishing between normal and anomalous sensor data. Regarding the anomaly classification, GRAD successfully categorized different anomaly types with high precision, enabling the recovery module to accurately restore damaged sensor data. Relative to analogous studies, GRAD surpasses current models by attaining a balance between elevated detection accuracy and diminished computational expense. These results demonstrate GRAD's potential as a reliable and efficient solution for real-time anomaly detection in autonomous vehicle systems, guaranteeing safe vehicle operation with minimal computational overhead.
Authors: Xinyu Wang, Jonas M. K\"ubler, Kailash Budhathoki, Yida Wang, Matth\"aus Kleindessner
Abstract: When serving a single base LLM with several different LoRA adapters simultaneously, the adapters cannot simply be merged with the base model's weights as the adapter swapping would create overhead and requests using different adapters could not be batched. Rather, the LoRA computations have to be separated from the base LLM computations, and in a multi-device setup the LoRA adapters can be sharded in a way that is well aligned with the base model's tensor parallel execution, as proposed in S-LoRA. However, the S-LoRA sharding strategy encounters some communication overhead, which may be small in theory, but can be large in practice. In this paper, we propose to constrain certain LoRA factors to be block-diagonal, which allows for an alternative way of sharding LoRA adapters that does not require any additional communication for the LoRA computations. We demonstrate in extensive experiments that our block-diagonal LoRA approach is similarly parameter efficient as standard LoRA (i.e., for a similar number of parameters it achieves similar downstream performance) and that it leads to significant end-to-end speed-up over S-LoRA. For example, when serving on eight A100 GPUs, we observe up to 1.79x (1.23x) end-to-end speed-up with 0.87x (1.74x) the number of adapter parameters for Llama-3.1-70B, and up to 1.63x (1.3x) end-to-end speed-up with 0.86x (1.73x) the number of adapter parameters for Llama-3.1-8B.
Authors: Hanzhang Wang, Zonglin Liu, Jingyi Xu, Chenyang Wang, Zhiwei Zhong, Qiangqiang Shen
Abstract: Proximal gradient algorithms (PGA), while foundational for inverse problems like image reconstruction, often yield unstable convergence and suboptimal solutions by violating the critical non-negativity constraint. We identify the gradient descent step as the root cause of this issue, which introduces negative values and induces high sensitivity to hyperparameters. To overcome these limitations, we propose a novel multiplicative update proximal gradient algorithm (SSO-PGA) with convergence guarantees, which is designed for robustness in non-negative inverse problems. Our key innovation lies in superseding the gradient descent step with a learnable sigmoid-based operator, which inherently enforces non-negativity and boundedness by transforming traditional subtractive updates into multiplicative ones. This design, augmented by a sliding parameter for enhanced stability and convergence, not only improves robustness but also boosts expressive capacity and noise immunity. We further formulate a degradation model for multi-modal restoration and derive its SSO-PGA-based optimization algorithm, which is then unfolded into a deep network to marry the interpretability of optimization with the power of deep learning. Extensive numerical and real-world experiments demonstrate that our method significantly surpasses traditional PGA and other state-of-the-art algorithms, ensuring superior performance and stability.
Authors: Hyeongkyun Kim, Orestis Oikonomou
Abstract: Flood susceptibility mapping (FSM) is vital for disaster prevention but remains challenging in data-scarce regions where hydrodynamic models require dense geophysical inputs. This work introduces ZeroFlood, a geospatial foundation model framework for data-efficient FSM. The approach fine-tunes Geospatial Foundation Models (GFMs) with Thinking-in-Modality (TiM) reasoning, enabling flood prediction from basic Earth observation data such as Sentinel-1 or Sentinel-2 imagery. Using paired EO and simulated flood maps from data-rich regions, ZeroFlood bridges data availability gaps through cross-modal representation learning. Experiments with TerraMind and Prithvi GFMs show that TiM enhances model robustness, with the TerraMind-Large configuration achieving an F1 score of 67.21. The results demonstrate the feasibility of foundation-model-based FSM as a scalable and data-efficient solution for flood risk management.
Authors: Hyunseung Kim, Dae-Woong Jeong, Changyoung Park, Won-Ji Lee, Ha-Eun Lee, Ji-Hye Lee, Rodrigo Hormazabal, Sung Moon Ko, Sumin Lee, Soorin Yim, Chanhui Lee, Sehui Han, Sang-Ho Cha, Woohyung Lim
Abstract: Artificial intelligence (AI) has emerged as a powerful accelerator of materials discovery, yet most existing models remain problem-specific, requiring additional data collection and retraining for each new property. Here we introduce and validate GATE (Geometrically Aligned Transfer Encoder) -- a generalizable AI framework that jointly learns 34 physicochemical properties spanning thermal, electrical, mechanical, and optical domains. By aligning these properties within a shared geometric space, GATE captures cross-property correlations that reduce disjoint-property bias -- a key factor causing false negatives in multi-criteria screening. To demonstrate its generalizability, GATE -- without any problem-specific reconfiguration -- was directly applied to the discovery of immersion cooling fluids for data centers, a stringent real-world challenge defined by the Open Compute Project (OCP). Screening billions of candidates, GATE identified 92,861 molecules as promising for practical deployment. Four were experimentally or literarily validated, showing strong agreement with wet-lab measurements and performance comparable to or exceeding a commercial coolant. These results establish GATE as a ready-to-use, generalizable AI platform readily applicable across diverse materials discovery tasks.
Authors: Ashwin Srinivasan, A Baskar, Tirtharaj Dash, Michael Bain, Sanjay Kumar Dey, Mainak Banerjee
Abstract: We investigate a relatively underexplored class of hybrid neurosymbolic models integrating symbolic learning with neural reasoning to construct data generators meeting formal correctness criteria. In \textit{Symbolic Neural Generators} (SNGs), symbolic learners examine logical specifications of feasible data from a small set of instances -- sometimes just one. Each specification in turn constrains the conditional information supplied to a neural-based generator, which rejects any instance violating the symbolic specification. Like other neurosymbolic approaches, SNG exploits the complementary strengths of symbolic and neural methods. The outcome of an SNG is a triple $(H, X, W)$, where $H$ is a symbolic description of feasible instances constructed from data, $X$ a set of generated new instances that satisfy the description, and $W$ an associated weight. We introduce a semantics for such systems, based on the construction of appropriate \textit{base} and \textit{fibre} partially-ordered sets combined into an overall partial order, and outline a probabilistic extension relevant to practical applications. In this extension, SNGs result from searching over a weighted partial ordering. We implement an SNG combining a restricted form of Inductive Logic Programming (ILP) with a large language model (LLM) and evaluate it on early-stage drug design. Our main interest is the description and the set of potential inhibitor molecules generated by the SNG. On benchmark problems -- where drug targets are well understood -- SNG performance is statistically comparable to state-of-the-art methods. On exploratory problems with poorly understood targets, generated molecules exhibit binding affinities on par with leading clinical candidates. Experts further find the symbolic specifications useful as preliminary filters, with several generated molecules identified as viable for synthesis and wet-lab testing.
Authors: Farid Bagirov, Mikhail Arkhipov, Ksenia Sycheva, Evgeniy Glukhov, Egor Bogomolov
Abstract: The application of Reinforcement Learning with Verifiable Rewards (RLVR) to mathematical and coding domains has demonstrated significant improvements in the reasoning and problem-solving abilities of Large Language Models. Despite its success in single generation problem solving, the reinforcement learning fine-tuning process may harm the model's exploration ability, as reflected in decreased diversity of generations and a resulting degradation of performance during Best-of-N sampling for large N values. In this work, we focus on optimizing the max@k metric, a continuous generalization of pass@k. We derive an unbiased on-policy gradient estimate for direct optimization of this metric. Furthermore, we extend our derivations to the off-policy updates, a common element in modern RLVR algorithms, that allows better sample efficiency. Empirically, we show that our objective effectively optimizes max@k metric in off-policy scenarios, aligning the model with the Best-of-N inference strategy.
Authors: Youngjun Choi, Joonseong Kang, Sungjun Lim, Kyungwoo Song
Abstract: Data valuation has become central in the era of data-centric AI. It drives efficient training pipelines and enables objective pricing in data markets by assigning a numeric value to each data point. Most existing data valuation methods estimate the effect of removing individual data points by evaluating changes in model validation performance under in-distribution (ID) settings, as opposed to out-of-distribution (OOD) scenarios where data follow different patterns. Since ID and OOD data behave differently, data valuation methods based on ID loss often fail to generalize to OOD settings, particularly when the validation set contains no OOD data. Furthermore, although OOD-aware methods exist, they involve heavy computational costs, which hinder practical deployment. To address these challenges, we introduce \emph{Eigen-Value} (EV), a plug-and-play data valuation framework for OOD robustness that uses only an ID data subset, including during validation. EV provides a new spectral approximation of domain discrepancy, which is the gap of loss between ID and OOD using ratios of eigenvalues of ID data's covariance matrix. EV then estimates the marginal contribution of each data point to this discrepancy via perturbation theory, alleviating the computational burden. Subsequently, EV plugs into ID loss-based methods by adding an EV term without any additional training loop. We demonstrate that EV achieves improved OOD robustness and stable value rankings across real-world datasets, while remaining computationally lightweight. These results indicate that EV is practical for large-scale settings with domain shift, offering an efficient path to OOD-robust data valuation.
Authors: Luca Melis, Matthew Grange, Iden Kalemaj, Karan Chadha, Shengyuan Hu, Elena Kashtelyan, Will Bullock
Abstract: The increasing deployment of Machine Learning (ML) models in sensitive domains motivates the need for robust, practical privacy assessment tools. PrivacyGuard is a comprehensive tool for empirical differential privacy (DP) analysis, designed to evaluate privacy risks in ML models through state-of-the-art inference attacks and advanced privacy measurement techniques. To this end, PrivacyGuard implements a diverse suite of privacy attack-- including membership inference , extraction, and reconstruction attacks -- enabling both off-the-shelf and highly configurable privacy analyses. Its modular architecture allows for the seamless integration of new attacks, and privacy metrics, supporting rapid adaptation to emerging research advances. We make PrivacyGuard available at https://github.com/facebookresearch/PrivacyGuard.
Authors: Michael L. Parker, Samar Mahmoud, Bailey Montefiore, Mario \"Oeren, Himani Tandon, Charlotte Wharrick, Matthew D. Segall
Abstract: We explore a "best-of-both" approach to modelling molecular properties by combining learned molecular descriptors from a graph neural network (GNN) with general-purpose descriptors and a mixed ensemble of machine learning (ML) models. We introduce a MetaModel framework to aggregate predictions from a diverse set of leading ML models. We present a featurisation scheme for combining task-specific GNN-derived features with conventional molecular descriptors. We demonstrate that our framework outperforms the cutting-edge ChemProp model on all regression datasets tested and 6 of 9 classification datasets. We further show that including the GNN features derived from ChemProp boosts the ensemble model's performance on several datasets where it otherwise would have underperformed. We conclude that to achieve optimal performance across a wide set of problems, it is vital to combine general-purpose descriptors with task-specific learned features and use a diverse set of ML models to make the predictions.
Authors: Lingxiao Huang, Zhize Li, Nisheeth K. Vishnoi, Runkai Yang, Haoyu Zhao
Abstract: We study the problem of constructing coresets for $(k, z)$-clustering when the input dataset is corrupted by stochastic noise drawn from a known distribution. In this setting, evaluating the quality of a coreset is inherently challenging, as the true underlying dataset is unobserved. To address this, we investigate coreset construction using surrogate error metrics that are tractable and provably related to the true clustering cost. We analyze a traditional metric from prior work and introduce a new error metric that more closely aligns with the true cost. Although our metric is defined independently of the noise distribution, it enables approximation guarantees that scale with the noise level. We design a coreset construction algorithm based on this metric and show that, under mild assumptions on the data and noise, enforcing an $\varepsilon$-bound under our metric yields smaller coresets and tighter guarantees on the true clustering cost than those obtained via classical metrics. In particular, we prove that the coreset size can improve by a factor of up to $\mathrm{poly}(k)$, where $n$ is the dataset size. Experiments on real-world datasets support our theoretical findings and demonstrate the practical advantages of our approach.
Authors: Xingtu Liu
Abstract: In this work, we study out-of-distribution generalization in meta-learning from an information-theoretic perspective. We focus on two scenarios: (i) when the testing environment mismatches the training environment, and (ii) when the training environment is broader than the testing environment. The first corresponds to the standard distribution mismatch setting, while the second reflects a broad-to-narrow training scenario. We further formalize the generalization problem in meta-reinforcement learning and establish corresponding generalization bounds. Finally, we analyze the generalization performance of a gradient-based meta-reinforcement learning algorithm.
Authors: M. M. Hammad
Abstract: We introduce the Schrodinger Neural Network (SNN), a principled architecture for conditional density estimation and uncertainty quantification inspired by quantum mechanics. The SNN maps each input to a normalized wave function on the output domain and computes predictive probabilities via the Born rule. The SNN departs from standard parametric likelihood heads by learning complex coefficients of a spectral expansion (e . g ., Chebyshev polynomials) whose squared modulus yields the conditional density $p(y|x)=\left| \psi _x(y)\right| {}^2$ with analytic normalization. This representation confers three practical advantages: positivity and exact normalization by construction, native multimodality through interference among basis modes without explicit mixture bookkeeping, and yields closed-form (or efficiently computable) functionals$-$such as moments and several calibration diagnostics$-$as quadratic forms in coefficient space. We develop the statistical and computational foundations of the SNN, including (i) training by exact maximum-likelihood with unit-sphere coefficient parameterization, (ii) physics-inspired quadratic regularizers (kinetic and potential energies) motivated by uncertainty relations between localization and spectral complexity, (iii) scalable low-rank and separable extensions for multivariate outputs, (iv) operator-based extensions that represent observables, constraints, and weak labels as self-adjoint matrices acting on the amplitude space, and (v) a comprehensive framework for evaluating multimodal predictions. The SNN provides a coherent, tractable framework to elevate probabilistic prediction from point estimates to physically inspired amplitude-based distributions.
Authors: Khoa Nguyen, Khang Tran, NhatHai Phan, Cristian Borcea, Rouming Jin, Issa Khalil
Abstract: This paper proposes Stochastic Geographic Gradient Fusion (SGFusion), a novel training algorithm to leverage the geographic information of mobile users in Federated Learning (FL). SGFusion maps the data collected by mobile devices onto geographical zones and trains one FL model per zone, which adapts well to the data and behaviors of users in that zone. SGFusion models the local data-based correlation among geographical zones as a hierarchical random graph (HRG) optimized by Markov Chain Monte Carlo sampling. At each training step, every zone fuses its local gradient with gradients derived from a small set of other zones sampled from the HRG. This approach enables knowledge fusion and sharing among geographical zones in a probabilistic and stochastic gradient fusion process with self-attention weights, such that "more similar" zones have "higher probabilities" of sharing gradients with "larger attention weights." SGFusion remarkably improves model utility without introducing undue computational cost. Extensive theoretical and empirical results using a heart-rate prediction dataset collected across 6 countries show that models trained with SGFusion converge with upper-bounded expected errors and significantly improve utility in all countries compared to existing approaches without notable cost in system scalability.
Authors: Hao Liang, Haifeng Wen, Kaishun Wu, Hong Xing
Abstract: Federated Learning (FL) is a distributed learning paradigm that preserves privacy by eliminating the need to exchange raw data during training. In its prototypical edge instantiation with underlying wireless transmissions enabled by analog over-the-air computing (AirComp), referred to as \emph{over-the-air FL (AirFL)}, the inherent channel noise plays a unique role of \emph{frenemy} in the sense that it degrades training due to noisy global aggregation while providing a natural source of randomness for privacy-preserving mechanisms, formally quantified by \emph{differential privacy (DP)}. It remains, nevertheless, challenging to effectively harness such channel impairments, as prior arts, under assumptions of either simple channel models or restricted types of loss functions, mostly considering (local) DP enhancement with a single-round or non-convergent bound on privacy loss. In this paper, we study AirFL over multiple-access fading channels with a multi-antenna base station (BS) subject to user-level DP requirements. Despite a recent study, which claimed in similar settings that artificial noise (AN) must be injected to ensure DP in general, we demonstrate, on the contrary, that DP can be gained as a \emph{perk} even \emph{without} employing any AN. Specifically, we derive a novel bound on DP that converges under general bounded-domain assumptions on model parameters, along with a convergence bound with general smooth and non-convex loss functions. Next, we optimize over receive beamforming and power allocations to characterize the optimal convergence-privacy trade-offs, which also reveal explicit conditions in which DP is achievable without compromising training. Finally, our theoretical findings are validated by extensive numerical results.
Authors: Yuhan Yang, Xingbo Fu, Jundong Li
Abstract: In recent years, pre-training Graph Neural Networks (GNNs) through self-supervised learning on unlabeled graph data has emerged as a widely adopted paradigm in graph learning. Although the paradigm is effective for pre-training powerful GNN models, the objective gap often exists between pre-training and downstream tasks. To bridge this gap, graph prompting adapts pre-trained GNN models to specific downstream tasks with extra learnable prompts while keeping the pre-trained GNN models frozen. As recent graph prompting methods largely focus on enhancing model utility on downstream tasks, they often overlook fairness concerns when designing prompts for adaptation. In fact, pre-trained GNN models will produce discriminative node representations across demographic subgroups, as downstream graph data inherently contains biases in both node attributes and graph structures. To address this issue, we propose an Adaptive Dual Prompting (ADPrompt) framework that enhances fairness for adapting pre-trained GNN models to downstream tasks. To mitigate attribute bias, we design an Adaptive Feature Rectification module that learns customized attribute prompts to suppress sensitive information at the input layer, reducing bias at the source. Afterward, we propose an Adaptive Message Calibration module that generates structure prompts at each layer, which adjust the message from neighboring nodes to enable dynamic and soft calibration of the information flow. Finally, ADPrompt jointly optimizes the two prompting modules to adapt the pre-trained GNN while enhancing fairness. We conduct extensive experiments on four datasets with four pre-training strategies to evaluate the performance of ADPrompt. The results demonstrate that our proposed ADPrompt outperforms seven baseline methods on node classification tasks.
Authors: Ke Xue, Ruo-Tong Chen, Rong-Xi Tan, Xi Lin, Yunqi Shi, Siyuan Xu, Mingxuan Yuan, Chao Qian
Abstract: Chip placement is a vital stage in modern chip design as it has a substantial impact on the subsequent processes and the overall quality of the final chip. The use of black-box optimization (BBO) for chip placement has a history of several decades. However, early efforts were limited by immature problem formulations and inefficient algorithm designs. Recent progress has shown the effectiveness and efficiency of BBO for chip placement, proving its potential to achieve state-of-the-art results. Despite these advancements, the field lacks a unified, BBO-specific benchmark for thoroughly assessing various problem formulations and BBO algorithms. To fill this gap, we propose BBOPlace-Bench, the first benchmark designed specifically for evaluating and developing BBO algorithms for chip placement tasks. It integrates three problem formulations of BBO for chip placement, and offers a modular, decoupled, and flexible framework that enables users to seamlessly implement, test, and compare their own algorithms. BBOPlace-Bench integrates a wide variety of existing BBO algorithms, including simulated annealing (SA), evolutionary algorithms (EAs), and Bayesian optimization (BO). Experimental results show that the problem formulations of mask-guided optimization and hyperparameter optimization exhibit superior performance than the sequence pair problem formulation, while EAs demonstrate better overall performance than SA and BO, especially in high-dimensional search spaces, and also achieve state-of-the-art performance compared to the mainstream chip placement methods. BBOPlace-Bench not only facilitates the development of efficient BBO-driven solutions for chip placement but also broadens the practical application scenarios (which are urgently needed) for the BBO community. The code of BBOPlace-Bench is available at https://github.com/lamda-bbo/BBOPlace-Bench.
Authors: Julie Mordacq, David Loiseaux, Vicky Kalogeiton, Steve Oudot
Abstract: Self-supervised learning (SSL) has emerged as a powerful paradigm for learning representations without labeled data, often by enforcing invariance to input transformations such as rotations or blurring. Recent studies have highlighted two pivotal properties for effective representations: (i) avoiding dimensional collapse-where the learned features occupy only a low-dimensional subspace, and (ii) enhancing uniformity of the induced distribution. In this work, we introduce T-REGS, a simple regularization framework for SSL based on the length of the Minimum Spanning Tree (MST) over the learned representation. We provide theoretical analysis demonstrating that T-REGS simultaneously mitigates dimensional collapse and promotes distribution uniformity on arbitrary compact Riemannian manifolds. Several experiments on synthetic data and on classical SSL benchmarks validate the effectiveness of our approach at enhancing representation quality.
Authors: Alex Ayoub, Kavosh Asadi, Dale Schuurmans, Csaba Szepesv\'ari, Karim Bouyarmane
Abstract: Large reasoning models (LRMs) often consume excessive tokens, inflating computational cost and latency. We challenge the assumption that longer responses improve accuracy. By penalizing reasoning tokens using a discounted reinforcement learning setup (interpretable as a small token cost) and analyzing Blackwell optimality in restricted policy classes, we encourage concise yet accurate reasoning. Experiments confirm our theoretical results that this approach shortens chains of thought while preserving accuracy.
Authors: Elena Celledoni, Brynjulf Owren, Lars Ruthotto, Tianjiao Nicole Yang
Abstract: Exploiting low-precision computations has become a standard strategy in deep learning to address the growing computational costs imposed by ever larger models and datasets. However, naively performing all computations in low precision can lead to roundoff errors and instabilities. Therefore, mixed precision training schemes usually store the weights in high precision and use low-precision computations only for whitelisted operations. Despite their success, these principles are currently not reliable for training continuous-time architectures such as neural ordinary differential equations (Neural ODEs). This paper presents a mixed precision training framework for neural ODEs, combining explicit ODE solvers with a custom backpropagation scheme, and demonstrates its effectiveness across a range of learning tasks. Our scheme uses low-precision computations for evaluating the velocity, parameterized by the neural network, and for storing intermediate states, while stability is provided by a custom dynamic adjoint scaling and by accumulating the solution and gradients in higher precision. These contributions address two key challenges in training neural ODE: the computational cost of repeated network evaluations and the growth of memory requirements with the number of time steps or layers. Along with the paper, we publish our extendable, open-source PyTorch package rampde, whose syntax resembles that of leading packages to provide a drop-in replacement in existing codes. We demonstrate the reliability and effectiveness of our scheme using challenging test cases and on neural ODE applications in image classification and generative models, achieving approximately 50% memory reduction and up to 2x speedup while maintaining accuracy comparable to single-precision training.
Authors: Spyros Rigas, Fotios Anagnostopoulos, Michalis Papachristou, Georgios Alexandridis
Abstract: Since their introduction, Kolmogorov-Arnold Networks (KANs) have been successfully applied across several domains, with physics-informed machine learning (PIML) emerging as one of the areas where they have thrived. In the PIML setting, Chebyshev-based physics-informed KANs (cPIKANs) have become the standard due to their computational efficiency. However, like their multilayer perceptron-based counterparts, cPIKANs face significant challenges when scaled to depth, leading to training instabilities that limit their applicability to several PDE problems. To address this, we propose a basis-agnostic, Glorot-like initialization scheme that preserves activation variance and yields substantial improvements in stability and accuracy over the default initialization of cPIKANs. Inspired by the PirateNet architecture, we further introduce Residual-Gated Adaptive KANs (RGA KANs), designed to mitigate divergence in deep cPIKANs where initialization alone is not sufficient. Through empirical tests and information bottleneck analysis, we show that RGA KANs successfully traverse all training phases, unlike baseline cPIKANs, which stagnate in the diffusion phase in specific PDE settings. Evaluations on seven standard forward PDE benchmarks under a fixed training pipeline with adaptive components demonstrate that RGA KANs consistently outperform parameter-matched cPIKANs and PirateNets - often by several orders of magnitude - while remaining stable in settings where the others diverge.
Authors: Siamak Ghodsi, Amjad Seyedi, Tai Le Quy, Fariba Karimi, Eirini Ntoutsi
Abstract: Fair graph clustering seeks partitions that respect network structure while maintaining proportional representation across sensitive groups, with applications spanning community detection, team formation, resource allocation, and social network analysis. Many existing approaches enforce rigid constraints or rely on multi-stage pipelines (e.g., spectral embedding followed by $k$-means), limiting trade-off control, interpretability, and scalability. We introduce \emph{DFNMF}, an end-to-end deep nonnegative tri-factorization tailored to graphs that directly optimizes cluster assignments with a soft statistical-parity regularizer. A single parameter $\lambda$ tunes the fairness--utility balance, while nonnegativity yields parts-based factors and transparent soft memberships. The optimization uses sparse-friendly alternating updates and scales near-linearly with the number of edges. Across synthetic and real networks, DFNMF achieves substantially higher group balance at comparable modularity, often dominating state-of-the-art baselines on the Pareto front. The code is available at https://github.com/SiamakGhodsi/DFNMF.git.
Authors: Chen Lu, Ke Xue, Lei Yuan, Yao Wang, Yaoyuan Wang, Sheng Fu, Chao Qian
Abstract: Dynamic algorithm configuration (DAC) is a recent trend in automated machine learning, which can dynamically adjust the algorithm's configuration during the execution process and relieve users from tedious trial-and-error tuning tasks. Recently, multi-agent reinforcement learning (MARL) approaches have improved the configuration of multiple heterogeneous hyperparameters, making various parameter configurations for complex algorithms possible. However, many complex algorithms have inherent inter-dependencies among multiple parameters (e.g., determining the operator type first and then the operator's parameter), which are, however, not considered in previous approaches, thus leading to sub-optimal results. In this paper, we propose the sequential multi-agent DAC (Seq-MADAC) framework to address this issue by considering the inherent inter-dependencies of multiple parameters. Specifically, we propose a sequential advantage decomposition network, which can leverage action-order information through sequential advantage decomposition. Experiments from synthetic functions to the configuration of multi-objective optimization algorithms demonstrate Seq-MADAC's superior performance over state-of-the-art MARL methods and show strong generalization across problem classes. Seq-MADAC establishes a new paradigm for the widespread dependency-aware automated algorithm configuration. Our code is available at https://github.com/lamda-bbo/seq-madac.
Authors: Siddharth Sahay, Radhika Agarwal
Abstract: This paper presents an end-to-end multilingual translation pipeline that integrates a custom U-Net for text detection, the Tesseract engine for text recognition, and a from-scratch sequence-to-sequence (Seq2Seq) Transformer for Neural Machine Translation (NMT). Our approach first utilizes a U-Net model, trained on a synthetic dataset , to accurately segment and detect text regions from an image. These detected regions are then processed by Tesseract to extract the source text. This extracted text is fed into a custom Transformer model trained from scratch on a multilingual parallel corpus spanning 5 languages. Unlike systems reliant on monolithic pre-trained models, our architecture emphasizes full customization and adaptability. The system is evaluated on its text detection accuracy, text recognition quality, and translation performance via BLEU scores. The complete pipeline demonstrates promising results, validating the viability of a custom-built system for translating text directly from images.
Authors: Zhongyi Yu, Jianqiu Wu, Zhenghao Wu, Shuhan Zhong, Weifeng Su, Chul-Ho Lee, Weipeng Zhuo
Abstract: Temporal graph link prediction aims to predict future interactions between nodes in a graph based on their historical interactions, which are encoded in node embeddings. We observe that heterogeneity naturally appears in temporal interactions, e.g., a few node pairs can make most interaction events, and interaction events happen at varying intervals. This leads to the problems of ineffective temporal information encoding and forgetting of past interactions for a pair of nodes that interact intermittently for their link prediction. Existing methods, however, do not consider such heterogeneity in their learning process, and thus their learned temporal node embeddings are less effective, especially when predicting the links for infrequently interacting node pairs. To cope with the heterogeneity, we propose a novel framework called TAMI, which contains two effective components, namely log time encoding function (LTE) and link history aggregation (LHA). LTE better encodes the temporal information through transforming interaction intervals into more balanced ones, and LHA prevents the historical interactions for each target node pair from being forgotten. State-of-the-art temporal graph neural networks can be seamlessly and readily integrated into TAMI to improve their effectiveness. Experiment results on 13 classic datasets and three newest temporal graph benchmark (TGB) datasets show that TAMI consistently improves the link prediction performance of the underlying models in both transductive and inductive settings. Our code is available at https://github.com/Alleinx/TAMI_temporal_graph.
Authors: Cheol Woo Kim, Shresth Verma, Mauricio Tec, Milind Tambe
Abstract: Direct Preference Optimization (DPO) has become a popular method for fine-tuning large language models (LLMs) due to its stability and simplicity. However, it is also known to be sensitive to noise in the data and prone to overfitting. Recent works have proposed using distributionally robust optimization (DRO) to address potential noise and distributional shift in the data. However, these methods often suffer from excessive conservatism and high computational cost. We propose DPO-PRO (DPO with Preference Robustness), a robust fine-tuning algorithm based on DPO which accounts for uncertainty in the preference distribution through a lightweight DRO formulation. Unlike prior DRO-based variants, DPO-PRO focuses solely on uncertainty in preferences, avoiding unnecessary conservatism and incurring negligible computational overhead. We further show that DPO-PRO is equivalent to a regularized DPO objective that penalizes model overconfidence under weak preference signals. We evaluate DPO-PRO on standard alignment benchmarks and a real-world public health task. Experimental results show that our method consistently improves robustness to noisy preference signals compared to existing DPO variants.
Authors: Yichi Zhang, Alex Schwing, Zhizhen Zhao
Abstract: Masked diffusion models have recently emerged as a flexible framework for discrete generative modeling. However, a key limitation of standard masked diffusion is its inability to effectively capture dependencies among tokens that are predicted concurrently, leading to degraded generation quality when dependencies among tokens are important. To explicitly model dependencies among tokens, we propose Variational Masked Diffusion (VMD), a framework that introduces latent variables into the masked diffusion process. Through controlled experiments on synthetic datasets, we demonstrate that VMD successfully learns dependencies that conventional masked diffusion fails to capture. We further validate the effectiveness of our approach on Sudoku puzzles and text datasets, where learning of dependencies among tokens improves global consistency. Across these domains, VMD enhances both generation quality and dependency awareness, highlighting the value of integrating variational inference into masked diffusion. Our code is available at: https://riccizz.github.io/VMD.
Authors: Eric Ramos, Sunny Sun
Abstract: Given a graph $G$, its independence sequence is the integral sequence $a_1,a_2,...,a_n$, where $a_i$ is the number of independent sets of vertices of size i. In the late 80's Alavi, Erdos, Malde, Schwenk showed that this sequence need not be unimodal for general graphs, but conjectured that it is always unimodal whenever $G$ is a tree. This conjecture was then naturally generalized to claim that the independence sequence of trees should be log concave, in the sense that $a_i^2$ is always above $a_{i-1}a_{i+1}$. This conjecture stood for many years, until in 2023, Kadrawi, Levit, Yosef, and Mizrachi proved that there were exactly two trees on 26 vertices whose independence sequence was not log concave. In this paper, we use the AI architecture PatternBoost, developed by Charton, Ellenberg, Wagner, and Williamson to train a machine to find counter-examples to the log-concavity conjecture. We will discuss the successes of this approach - finding tens of thousands of new counter-examples to log-concavity with vertex set sizes varying from 27 to 101 - and some of its fascinating failures.
Authors: Yin Sun, Yiwen Liu, Junjie Song, Chenyu Zhang, Xinyuan Zhang, Lingjie Liu, Siqi Chen, Yuji Cao
Abstract: In the area of commercial auto sales system, high-quality lead score sequencing determines the priority of a sale's work and is essential for optimizing the efficiency of the sales system. Since CRM (Customer Relationship Management) system contains plenty of textual interaction features between sales and customers, traditional techniques such as Click Through Rate (CTR) prediction struggle with processing the complex information inherent in natural language features, which limits their effectiveness in sales lead ranking. Bridging this gap is critical for enhancing business intelligence and decision-making. Recently, the emergence of large language models (LLMs) has opened new avenues for improving recommendation systems, this study introduces asLLR (LLM-based Leads Ranking in Auto Sales), which integrates CTR loss and Question Answering (QA) loss within a decoder-only large language model architecture. This integration enables the simultaneous modeling of both tabular and natural language features. To verify the efficacy of asLLR, we constructed an innovative dataset derived from the customer lead pool of a prominent new energy vehicle brand, with 300,000 training samples and 40,000 testing samples. Our experimental results demonstrate that asLLR effectively models intricate patterns in commercial datasets, achieving the AUC of 0.8127, surpassing traditional CTR estimation methods by 0.0231. Moreover, asLLR enhances CTR models when used for extracting text features by 0.0058. In real-world sales scenarios, after rigorous online A/B testing, asLLR increased the sales volume by about 9.5% compared to the traditional method, providing a valuable tool for business intelligence and operational decision-making.
Authors: Anant Pareek
Abstract: The confluence of Artificial Intelligence and Computational Psychology presents an opportunity to model, understand, and interact with complex human psychological states through computational means. This paper presents a comprehensive, multi-faceted framework designed to bridge the gap between isolated predictive modeling and an interactive system for psychological analysis. The methodology encompasses a rigorous, end-to-end development lifecycle. First, foundational performance benchmarks were established on four diverse psychological datasets using classical machine learning techniques. Second, state-of-the-art transformer models were fine-tuned, a process that necessitated the development of effective solutions to overcome critical engineering challenges, including the resolution of numerical instability in regression tasks and the creation of a systematic workflow for conducting large-scale training under severe resource constraints. Third, a generative large language model (LLM) was fine-tuned using parameter-efficient techniques to function as an interactive "Personality Brain." Finally, the entire suite of predictive and generative models was architected and deployed as a robust, scalable microservices ecosystem. Key findings include the successful stabilization of transformer-based regression models for affective computing, showing meaningful predictive performance where standard approaches failed, and the development of a replicable methodology for democratizing large-scale AI research. The significance of this work lies in its holistic approach, demonstrating a complete research-to-deployment pipeline that integrates predictive analysis with generative dialogue, thereby providing a practical model for future research in computational psychology and human-AI interaction.
Authors: Apoorva Chavali, Reeve Menezes
Abstract: Current recommendation systems often tend to overlook emotional context and rely on historical listening patterns or static mood tags. This paper introduces a novel music recommendation framework employing a variant of Wide and Deep Learning architecture that takes in real-time emotional states inferred directly from natural language as inputs and recommends songs that closely portray the mood. The system captures emotional contexts from user-provided textual descriptions by using transformer-based embeddings, which were finetuned to predict the emotional dimensions of valence-arousal. The deep component of the architecture utilizes these embeddings to generalize unseen emotional patterns, while the wide component effectively memorizes user-emotion and emotion-genre associations through cross-product features. Experimental results show that personalized music selections positively influence the user's emotions and lead to a significant improvement in emotional relevance.
Authors: Weichen Wang, Chengchun Shi
Abstract: This paper is a discussion of the 2025 JASA discussion paper by Su et al. (2025). We would like to congratulate the authors on conducting a comprehensive and insightful empirical investigation of the 2023 ICML ranking data. The review quality of machine learning (ML) conferences has become a big concern in recent years, due to the rapidly growing number of submitted manuscripts. In this discussion, we propose an approach alternative to Su et al. (2025) that leverages ranking information from reviewers rather than authors. We simulate review data that closely mimics the 2023 ICML conference submissions. Our results show that (i) incorporating ranking information from reviewers can significantly improve the evaluation of each paper's quality, often outperforming the use of ranking information from authors alone; and (ii) combining ranking information from both reviewers and authors yields the most accurate evaluation of submitted papers in most scenarios.
Authors: Yichi Zhang, Jun Bai, Zhixin Cai, Shuhan Qin, Zhuofan Chen, Jinghua Guan, Wenge Rong
Abstract: Dense retrievers enhance retrieval by encoding queries and documents into continuous vectors, but they often struggle with reasoning-intensive queries. Although Large Language Models (LLMs) can reformulate queries to capture complex reasoning, applying them universally incurs significant computational cost. In this work, we propose Adaptive Query Reasoning (AdaQR), a hybrid query rewriting framework. Within this framework, a Reasoner Router dynamically directs each query to either fast dense reasoning or deep LLM reasoning. The dense reasoning is achieved by the Dense Reasoner, which performs LLM-style reasoning directly in the embedding space, enabling a controllable trade-off between efficiency and accuracy. Experiments on large-scale retrieval benchmarks BRIGHT show that AdaQR reduces reasoning cost by 28% while preserving-or even improving-retrieval performance by 7%.
Authors: Yuhui Liu, Samannita Halder, Shian Wang, Tianyi Li
Abstract: This study introduces a novel control framework for adaptive cruise control (ACC) in automated driving, leveraging Long Short-Term Memory (LSTM) networks and physics-informed constraints. As automated vehicles (AVs) adopt advanced features like ACC, transportation systems are becoming increasingly intelligent and efficient. However, existing AV control strategies primarily focus on optimizing the performance of individual vehicles or platoons, often neglecting their interactions with human-driven vehicles (HVs) and the broader impact on traffic flow. This oversight can exacerbate congestion and reduce overall system efficiency. To address this critical research gap, we propose a neural network-based, socially compliant AV control framework that incorporates social value orientation (SVO). This framework enables AVs to account for their influence on HVs and traffic dynamics. By leveraging AVs as mobile traffic regulators, the proposed approach promotes adaptive driving behaviors that reduce congestion, improve traffic efficiency, and lower energy consumption. Within this framework, we define utility functions for both AVs and HVs, which are optimized based on the SVO of each AV to balance its own control objectives with broader traffic flow considerations. Numerical results demonstrate the effectiveness of the proposed method in adapting to varying traffic conditions, thereby enhancing system-wide efficiency. Specifically, when the AV's control mode shifts from prioritizing energy consumption to optimizing traffic flow efficiency, vehicles in the following platoon experience at least a 58.99% increase in individual energy consumption alongside at least a 38.39% improvement in individual average speed, indicating significant enhancements in traffic dynamics.
Authors: Kumater Ter, RexCharles Donatus, Ore-Ofe Ajayi, Daniel Udekwe
Abstract: Reinforcement learning (RL) has become a foundational approach for enabling intelligent robotic behavior in dynamic and uncertain environments. This work presents an in-depth review of RL principles, advanced deep reinforcement learning (DRL) algorithms, and their integration into robotic and control systems. Beginning with the formalism of Markov Decision Processes (MDPs), the study outlines essential elements of the agent-environment interaction and explores core algorithmic strategies including actor-critic methods, value-based learning, and policy gradients. Emphasis is placed on modern DRL techniques such as DDPG, TD3, PPO, and SAC, which have shown promise in solving high-dimensional, continuous control tasks. A structured taxonomy is introduced to categorize RL applications across domains such as locomotion, manipulation, multi-agent coordination, and human-robot interaction, along with training methodologies and deployment readiness levels. The review synthesizes recent research efforts, highlighting technical trends, design patterns, and the growing maturity of RL in real-world robotics. Overall, this work aims to bridge theoretical advances with practical implementations, providing a consolidated perspective on the evolving role of RL in autonomous robotic systems.
Authors: Arun Muthukkumar
Abstract: Pose estimation is essential for many applications within computer vision and robotics. Despite its uses, few works provide rigorous uncertainty quantification for poses under dense or learned models. We derive a closed-form lower bound on the covariance of camera pose estimates by treating a differentiable renderer as a measurement function. Linearizing image formation with respect to a small pose perturbation on the manifold yields a render-aware Cram\'er-Rao bound. Our approach reduces to classical bundle-adjustment uncertainty, ensuring continuity with vision theory. It also naturally extends to multi-agent settings by fusing Fisher information across cameras. Our statistical formulation has downstream applications for tasks such as cooperative perception and novel view synthesis without requiring explicit keypoint correspondences.
Authors: Juan Guti\'errez, \'Angel Mora, Pablo Regod\'on, Silvia Rodriguez, Jos\'e Luis Blanco
Abstract: We explore the enhancement of Human-in-the-Loop video annotation by integrating automatic capabilities to ease the task for annotators and assess their performance. The research delves into the practical implications of the annotation processes, the integration of AI components, and the evaluation of its outcomes. We analyze their impact on efficiency, accuracy, and overall annotation quality. Focusing on the Human-in-the-Loop for video annotation tasks, we implemented a single-iteration scheme using Label Studio and AI-powered zero-shot pre-annotations. Using this framework, we designed a test based on the annotation of the UCF-Crime dataset to discriminate between normal and abnormal activities in video footage. Our results evidence how automatic AI-based pre-annotation can streamline the video annotation workflow, empowering human annotators and optimizing the overall pipeline. Using the pre-annotated data, we observed a 35% reduction in the annotation time for 70% of the annotators with similar quality annotations, compared to the traditional manual annotation task. Results are consistent with asset duration and complexity. We also observed that while annotators rapidly learned to use the tool, the produced annotations are more coherent among annotators and better match the natural clustering of the video frames.
Authors: Marouane Tliba, Mohamed Amine Kerkouri, Yassine Nasser, Nour Aburaed, Aladine Chetouani, Ulas Bagci, Rachid Jennane
Abstract: Knee osteoarthritis (KOA) diagnosis from radiographs remains challenging due to the subtle morphological details that standard deep learning models struggle to capture effectively. We propose a novel multimodal framework that combines anatomical structure with radiographic features by integrating a morphological graph representation - derived from Segment Anything Model (SAM) segmentations - with a vision encoder. Our approach enforces alignment between geometry-informed graph embeddings and radiographic features through mutual information maximization, significantly improving KOA classification accuracy. By constructing graphs from anatomical features, we introduce explicit morphological priors that mirror clinical assessment criteria, enriching the feature space and enhancing the model's inductive bias. Experiments on the Osteoarthritis Initiative dataset demonstrate that our approach surpasses single-modality baselines by up to 10\% in accuracy (reaching nearly 80\%), while outperforming existing state-of-the-art methods by 8\% in accuracy and 11\% in F1 score. These results underscore the critical importance of incorporating anatomical structure into radiographic analysis for accurate KOA severity grading.
Authors: Pedro Cisneros-Velarde
Abstract: We consider the situation where we have a limited number of denoising steps, i.e., of evaluations of a diffusion model. We show that two parallel processors or samplers under such limitation can improve the quality of the sampled image. Particularly, the two samplers make denoising steps at successive times, and their information is appropriately integrated in the latent image. Remarkably, our method is simple both conceptually and to implement: it is plug-&-play, model agnostic, and does not require any additional fine-tuning or external models. We test our method with both automated and human evaluations for different diffusion models. We also show that a naive integration of the information from the two samplers lowers sample quality. Finally, we find that adding more parallel samplers does not necessarily improve sample quality.
Authors: Zhao Liu, Yichen Zhu, Yiqing Yang, Guoping Tang, Rui Huang, Qiang Luo, Xiao Lv, Ruiming Tang, Kun Gai, Guorui Zhou
Abstract: Generative recommendation (GR) is an emerging paradigm that represents each item via a tokenizer as an n-digit semantic ID (SID) and predicts the next item by autoregressively generating its SID conditioned on the user's history. However, two structural properties of SIDs make ARMs ill-suited. First, intra-item consistency: the n digits jointly specify one item, yet the left-to-right causality trains each digit only under its prefix and blocks bidirectional cross-digit evidence, collapsing supervision to a single causal path. Second, inter-digit heterogeneity: digits differ in semantic granularity and predictability, while the uniform next-token objective assigns equal weight to all digits, overtraining easy digits and undertraining hard digits. To address these two issues, we propose DiffGRM, a diffusion-based GR model that replaces the autoregressive decoder with a masked discrete diffusion model (MDM), thereby enabling bidirectional context and any-order parallel generation of SID digits for recommendation. Specifically, we tailor DiffGRM in three aspects: (1) tokenization with Parallel Semantic Encoding (PSE) to decouple digits and balance per-digit information; (2) training with On-policy Coherent Noising (OCN) that prioritizes uncertain digits via coherent masking to concentrate supervision on high-value signals; and (3) inference with Confidence-guided Parallel Denoising (CPD) that fills higher-confidence digits first and generates diverse Top-K candidates. Experiments show consistent gains over strong generative and discriminative recommendation baselines on multiple datasets, improving NDCG@10 by 6.9%-15.5%. Code is available at https://github.com/liuzhao09/DiffGRM.
Authors: Chanyoung Chung, Kyeongryul Lee, Sunbin Park, Joyce Jiyoung Whang
Abstract: Recommender systems have long been built upon the modeling of interactions between users and items, while recent studies have sought to broaden this paradigm by generalizing to new users and items, incorporating diverse information sources, and transferring knowledge across domains. Nevertheless, these efforts have largely focused on individual aspects, hindering their ability to tackle the complex recommendation scenarios that arise in daily consumptions across diverse domains. In this paper, we present MICRec, a unified framework that fuses inductive modeling, multimodal guidance, and cross-domain transfer to capture user contexts and latent preferences in heterogeneous and incomplete real-world data. Moving beyond the inductive backbone of INMO, our model refines expressive representations through modality-based aggregation and alleviates data sparsity by leveraging overlapping users as anchors across domains, thereby enabling robust and generalizable recommendation. Experiments show that MICRec outperforms 12 baselines, with notable gains in domains with limited training data.
Authors: Samuel J. Barrett, Docko Sow
Abstract: Earth Observation (EO) Foundation Modelling (FM) holds great promise for simplifying and improving the use of EO data for diverse real-world tasks. However, most existing models require additional adaptation before they can be used and are structured rigidly around particular data sources or training approaches. To address this, we take inspiration from large language models, where diverse tasks, both pre-training and downstream, are implicitly captured through next-token prediction over unified token sequences, leveraging the structure and diversity of the training data. We introduce SITS-DECO (Satellite Image Time Series-DECoder Only), a proof-of-concept generative model that applies this unified-sequence framing to EO data. Using a simple GPT-style decoder-only architecture, and demonstrate its ability to perform useful EO tasks (pixel-wise, multi-temporal, multi-modal crop-type classification) in a purely generative framework. Through symbolic prompting, we show that the model can perform multiple supervised and self-supervised tasks within a single unified architecture, without task- or modality-specific adaptation. Despite its simplicity and lack of spatial context, SITS-DECO outperforms much larger EO foundation models on crop-type classification (PASTIS-R) demonstrating that dense temporal sequence modelling is a critical missing ingredient in the current paradigm. This work exemplifies a data-centric modelling paradigm in which capability arises from the diversity and structure of the training data rather than from architectural complexity. SITS-DECO provides a lightweight, practical route to multi-modal, multi-task EO modelling, and a conceptual bridge toward future generative EO foundation models.
Authors: Xiaoyu Liu, Chaoyou Fu, Chi Yan, Chu Wu, Haihan Gao, Yi-Fan Zhang, Shaoqi Dong, Cheng Qian, Bin Luo, Xiuyong Yang, Guanwu Li, Yusheng Cai, Yunhang Shen, Deqiang Jiang, Haoyu Cao, Xing Sun, Caifeng Shan, Ran He
Abstract: Current Vision-Language-Action (VLA) models are often constrained by a rigid, static interaction paradigm, which lacks the ability to see, hear, speak, and act concurrently as well as handle real-time user interruptions dynamically. This hinders seamless embodied collaboration, resulting in an inflexible and unresponsive user experience. To address these limitations, we introduce VITA-E, a novel embodied interaction framework designed for both behavioral concurrency and nearly real-time interruption. The core of our approach is a dual-model architecture where two parallel VLA instances operate as an ``Active Model'' and a ``Standby Model'', allowing the embodied agent to observe its environment, listen to user speech, provide verbal responses, and execute actions, all concurrently and interruptibly, mimicking human-like multitasking capabilities. We further propose a ``model-as-controller'' paradigm, where we fine-tune the VLM to generate special tokens that serve as direct system-level commands, coupling the model's reasoning with the system's behavior. Experiments conducted on a physical humanoid platform demonstrate that VITA-E can reliably handle complex interactive scenarios. Our framework is compatible with various dual-system VLA models, achieving an extremely high success rate on emergency stops and speech interruptions while also successfully performing concurrent speech and action. This represents a significant step towards more natural and capable embodied assistants.
Authors: Rohan Senthil, Swee Liang Wong
Abstract: Anomaly detection in cybersecurity is a challenging task, where normal events far outnumber anomalous ones with new anomalies occurring frequently. Classical autoencoders have been used for anomaly detection, but struggles in data-limited settings which quantum counterparts can potentially overcome. In this work, we apply Quantum Autoencoders (QAEs) for anomaly detection in cybersecurity, specifically on the BPF-extended tracking honeypot (BETH) dataset. QAEs are evaluated across multiple encoding techniques, ansatz types, repetitions, and feature selection strategies. Our results demonstrate that an 8-feature QAE using Dense-Angle encoding with a RealAmplitude ansatz can outperform Classical Autoencoders (CAEs), even when trained on substantially fewer samples. The effects of quantum encoding and feature selection for developing quantum models are demonstrated and discussed. In a data-limited setting, the best performing QAE model has a F1 score of 0.87, better than that of CAE (0.77). These findings suggest that QAEs may offer practical advantages for anomaly detection in data-limited scenarios.
Authors: Marco Siino, Giuseppe Bonomo, Rosario Sorbello, Ilenia Tinnirello
Abstract: Brain-computer interfaces (BCIs) based on motor imagery (MI) translate covert movement intentions into actionable commands, yet reliable decoding from non-invasive EEG remains challenging due to nonstationarity, low SNR, and subject variability. We present RatioWaveNet, which augments a strong temporal CNN-Transformer backbone (TCFormer) with a trainable, Rationally-Dilated Wavelet Transform (RDWT) front end. The RDWT performs an undecimated, multi-resolution subband decomposition that preserves temporal length and shift-invariance, enhancing sensorimotor rhythms while mitigating jitter and mild artifacts; subbands are fused via lightweight grouped 1-D convolutions and passed to a multi-kernel CNN for local temporal-spatial feature extraction, a grouped-query attention encoder for long-range context, and a compact TCN head for causal temporal integration. Our goal is to test whether this principled wavelet front end improves robustness precisely where BCIs typically fail - on the hardest subjects - and whether such gains persist on average across seeds under both intra- and inter-subject protocols. On BCI-IV-2a and BCI-IV-2b, across five seeds, RatioWaveNet improves worst-subject accuracy over the Transformer backbone by +0.17 / +0.42 percentage points (Sub-Dependent / LOSO) on 2a and by +1.07 / +2.54 percentage points on 2b, with consistent average-case gains and modest computational overhead. These results indicate that a simple, trainable wavelet front end is an effective plug-in to strengthen Transformer-based BCIs, improving worst-case reliability without sacrificing efficiency.
Authors: Babaniyi Olaniyi, Ina Kalisa, Ana Fern\'andez del R\'io, Jean Marie Vianney Hakizayezu, Enric Jan\'e, Eniola Olaleye, Juan Francisco Garamendi, Ivan Nazarov, Aditya Rastogi, Mateo Diaz-Quiroz, \'Africa Peri\'a\~nez, Regis Hitimana
Abstract: As part of Rwanda's transition toward universal health coverage, the national Community-Based Health Insurance (CBHI) scheme is moving from retrospective fee-for-service reimbursements to prospective capitation payments for public primary healthcare providers. This report outlines a data-driven approach to designing, calibrating, and monitoring the capitation model using individual-level claims data from the Intelligent Health Benefits System (IHBS). We introduce a transparent, interpretable formula for allocating payments to Health Centers and their affiliated Health Posts. The formula is based on catchment population, service utilization patterns, and patient inflows, with parameters estimated via regression models calibrated on national claims data. Repeated validation exercises show the payment scheme closely aligns with historical spending while promoting fairness and adaptability across diverse facilities. In addition to payment design, the same dataset enables actionable behavioral insights. We highlight the use case of monitoring antibiotic prescribing patterns, particularly in pediatric care, to flag potential overuse and guideline deviations. Together, these capabilities lay the groundwork for a learning health financing system: one that connects digital infrastructure, resource allocation, and service quality to support continuous improvement and evidence-informed policy reform.
Authors: Ryan Zhang, Herbert Woisetscl\"ager
Abstract: Real-world AI systems are tackling increasingly complex problems, often through interactions among large language model (LLM) agents. When these agents develop inconsistent conventions, coordination can break down. Applications such as collaborative coding and distributed planning therefore require reliable, consistent communication, and scalability is a central concern as systems grow. We introduce Schema-Induced Games for Naming (SIGN), a naming game that examines how lightweight structure can steer convention formation. We compare schema-induced communication to unconstrained natural language and find faster convergence with up to 5.8x higher agreement. These results suggest that minimal structure can act as a simple control knob for efficient multi-agent coordination, pointing toward broader applications beyond the naming game.
Authors: F. I. Qowy
Abstract: Caching and prefetching techniques are fundamental to modern computing, serving to bridge the growing performance gap between processors and memory. Traditional prefetching strategies are often limited by their reliance on predefined heuristics or simplified statistical models, which fail to capture the complex, non-linear dependencies in modern data access patterns. This paper introduces a modular framework leveraging Graph Neural Networks (GNNs) to model and predict access patterns within graph-structured data, focusing on web navigation and hierarchical file systems. The toolchain consists of: a route mapper for extracting structural information, a graph constructor for creating graph representations, a walk session generator for simulating user behaviors, and a gnn prefetch module for training and inference. We provide a detailed conceptual analysis showing how GNN-based approaches can outperform conventional methods by learning intricate dependencies. This work offers both theoretical foundations and a practical, replicable pipeline for future research in graph-driven systems optimization.
Authors: Shamim Yazdani, Akansha Singh, Nripsuta Saxena, Zichong Wang, Avash Palikhe, Deng Pan, Umapada Pal, Jie Yang, Wenbin Zhang
Abstract: In recent years, deep learning based generative models, particularly Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Diffusion Models (DMs), have been instrumental in in generating diverse, high-quality content across various domains, such as image and video synthesis. This capability has led to widespread adoption of these models and has captured strong public interest. As they continue to advance at a rapid pace, the growing volume of research, expanding application areas, and unresolved technical challenges make it increasingly difficult to stay current. To address this need, this survey introduces a comprehensive taxonomy that organizes the literature and provides a cohesive framework for understanding the development of GANs, VAEs, and DMs, including their many variants and combined approaches. We highlight key innovations that have improved the quality, diversity, and controllability of generated outputs, reflecting the expanding potential of generative artificial intelligence. In addition to summarizing technical progress, we examine rising ethical concerns, including the risks of misuse and the broader societal impact of synthetic media. Finally, we outline persistent challenges and propose future research directions, offering a structured and forward looking perspective for researchers in this fast evolving field.
Authors: Shayan Karimi, Xiaoqi Tan
Abstract: This paper investigates the computational complexity of reinforcement learning in a novel linear function approximation regime, termed partial $q^{\pi}$-realizability. In this framework, the objective is to learn an $\epsilon$-optimal policy with respect to a predefined policy set $\Pi$, under the assumption that all value functions for policies in $\Pi$ are linearly realizable. The assumptions of this framework are weaker than those in $q^{\pi}$-realizability but stronger than those in $q^*$-realizability, providing a practical model where function approximation naturally arises. We prove that learning an $\epsilon$-optimal policy in this setting is computationally hard. Specifically, we establish NP-hardness under a parameterized greedy policy set (argmax) and show that - unless NP = RP - an exponential lower bound (in feature vector dimension) holds when the policy set contains softmax policies, under the Randomized Exponential Time Hypothesis. Our hardness results mirror those in $q^*$-realizability and suggest computational difficulty persists even when $\Pi$ is expanded beyond the optimal policy. To establish this, we reduce from two complexity problems, $\delta$-Max-3SAT and $\delta$-Max-3SAT(b), to instances of GLinear-$\kappa$-RL (greedy policy) and SLinear-$\kappa$-RL (softmax policy). Our findings indicate that positive computational results are generally unattainable in partial $q^{\pi}$-realizability, in contrast to $q^{\pi}$-realizability under a generative access model.
Authors: Marios Andreou, Nan Chen
Abstract: Causal inference identifies cause-and-effect relationships between variables. While traditional approaches rely on data to reveal causal links, a recently developed method, assimilative causal inference (ACI), integrates observations with dynamical models. It utilizes Bayesian data assimilation to trace causes back from observed effects by quantifying the reduction in uncertainty. ACI advances the detection of instantaneous causal relationships and the intermittent reversal of causal roles over time. Beyond identifying causal connections, an equally important challenge is determining the associated causal influence range (CIR), indicating when causal influences emerged and for how long they persist. In this paper, ACI is employed to develop mathematically rigorous formulations of both forward and backward CIRs at each time. The forward CIR quantifies the temporal impact of a cause, while the backward CIR traces the onset of triggers for an observed effect, thus characterizing causal predictability and attribution of outcomes at each transient phase, respectively. Objective and robust metrics for both CIRs are introduced, eliminating the need for empirical thresholds. Computationally efficient approximation algorithms to compute CIRs are developed, which facilitate the use of closed-form expressions for a broad class of nonlinear dynamical systems. Numerical simulations demonstrate how this forward and backward CIR framework provides new possibilities for probing complex dynamical systems. It advances the study of bifurcation-driven and noise-induced tipping points in Earth systems, investigates the impact from resolving the interfering variables when determining the influence ranges, and elucidates atmospheric blocking mechanisms in the equatorial region. These results have direct implications for science, policy, and decision-making.
Authors: Dhrupad Bhardwaj, Julia Kempe, Tim G. J. Rudner
Abstract: To deploy large language models (LLMs) in high-stakes application domains that require substantively accurate responses to open-ended prompts, we need reliable, computationally inexpensive methods that assess the trustworthiness of long-form responses generated by LLMs. However, existing approaches often rely on claim-by-claim fact-checking, which is computationally expensive and brittle in long-form responses to open-ended prompts. In this work, we introduce semantic isotropy -- the degree of uniformity across normalized text embeddings on the unit sphere -- and use it to assess the trustworthiness of long-form responses generated by LLMs. To do so, we generate several long-form responses, embed them, and estimate the level of semantic isotropy of these responses as the angular dispersion of the embeddings on the unit sphere. We find that higher semantic isotropy -- that is, greater embedding dispersion -- reliably signals lower factual consistency across samples. Our approach requires no labeled data, no fine-tuning, and no hyperparameter selection, and can be used with open- or closed-weight embedding models. Across multiple domains, our method consistently outperforms existing approaches in predicting nonfactuality in long-form responses using only a handful of samples -- offering a practical, low-cost approach for integrating trust assessment into real-world LLM workflows.
Authors: Siddharth Chaudhary
Abstract: Large language models display in-context learning as an emergent effect of scale, but they rely on static weights during inference. In contrast, biological systems continually adapt via synaptic plasticity. We investigate whether explicit, biologically inspired plasticity can endow Transformers with faster in-sequence adaptation. To this end, we augment decoder-only Transformers with fast-weight modules updated either by (i) a neuromodulated Hebbian rule or (ii) the gradient-based plasticity mechanism of Duan et al. (2023). Across copying, regression, and few-shot classification tasks (CIFAR-FS, Omniglot), Hebbian plasticity consistently achieves lower loss and stronger few-shot generalization, while gradient-based updates perform best on long-horizon credit assignment. When associations are short and linearly separable, static weights suffice, defining a clear boundary condition for when plasticity helps. Analysis of learned modulatory signals reveals that gradient-based rules maintain large, persistent updates, whereas Hebbian plasticity is sharply gated around salient events. Together, these results show that explicit plasticity complements attention by enabling rapid, task-specific adaptation, and clarify when different plasticity mechanisms are most effective.
Authors: Weiyu Chen, Arnaud Delorme
Abstract: Detecting single-trial P300 from EEG is difficult when only a few labeled trials are available. When attempting to boost a small target set with a large source dataset through transfer learning, cross-dataset shift arises. To address this challenge, we study transfer between two public visual-oddball ERP datasets using five shared electrodes (Fz, Pz, P3, P4, Oz) under a strict small-sample regime (target: 10 trials/subject; source: 80 trials/subject). We introduce Adaptive Split Maximum Mean Discrepancy Training (AS-MMD), which combines (i) a target-weighted loss with warm-up tied to the square root of the source/target size ratio, (ii) Split Batch Normalization (Split-BN) with shared affine parameters and per-domain running statistics, and (iii) a parameter-free logit-level Radial Basis Function kernel Maximum Mean Discrepancy (RBF-MMD) term using the median-bandwidth heuristic. Implemented on an EEG Conformer, AS-MMD is backbone-agnostic and leaves the inference-time model unchanged. Across both transfer directions, it outperforms target-only and pooled training (Active Visual Oddball: accuracy/AUC 0.66/0.74; ERP CORE P3: 0.61/0.65), with gains over pooling significant under corrected paired t-tests. Ablations attribute improvements to all three components.
Authors: Matthew J. Colbrook, Zlatko Drma\v{c}, Andrew Horning
Abstract: Koopman operators provide a linear framework for data-driven analyses of nonlinear dynamical systems, but their infinite-dimensional nature presents major computational challenges. In this article, we offer an introductory guide to Koopman learning, emphasizing rigorously convergent data-driven methods for forecasting and spectral analysis. We provide a unified account of error control via residuals in both finite- and infinite-dimensional settings, an elementary proof of convergence for generalized Laplace analysis -- a variant of filtered power iteration that works for operators with continuous spectra and no spectral gaps -- and review state-of-the-art approaches for computing continuous spectra and spectral measures. The goal is to provide both newcomers and experts with a clear, structured overview of reliable data-driven techniques for Koopman spectral analysis.
Authors: Or Ronai, Vladimir Kulikov, Tomer Michaeli
Abstract: The remarkable success of diffusion and flow-matching models has ignited a surge of works on adapting them at test time for controlled generation tasks. Examples range from image editing to restoration, compression and personalization. However, due to the iterative nature of the sampling process in those models, it is computationally impractical to use gradient-based optimization to directly control the image generated at the end of the process. As a result, existing methods typically resort to manipulating each timestep separately. Here we introduce FlowOpt - a zero-order (gradient-free) optimization framework that treats the entire flow process as a black box, enabling optimization through the whole sampling path without backpropagation through the model. Our method is both highly efficient and allows users to monitor the intermediate optimization results and perform early stopping if desired. We prove a sufficient condition on FlowOpt's step-size, under which convergence to the global optimum is guaranteed. We further show how to empirically estimate this upper bound so as to choose an appropriate step-size. We demonstrate how FlowOpt can be used for image editing, showcasing two options: (i) inversion (determining the initial noise that generates a given image), and (ii) directly steering the edited image to be similar to the source image while conforming to a target text prompt. In both cases, FlowOpt achieves state-of-the-art results while using roughly the same number of neural function evaluations (NFEs) as existing methods. Code and examples are available on the project's webpage.
Authors: Rick Chen, Joseph Ternasky, Aaron Ontoyin Yin, Xianling Mu, Fuat Alican, Yigit Ihlamur
Abstract: Large language models (LLMs) can already identify patterns and reason effectively, yet their variable accuracy hampers adoption in high-stakes decision-making applications. In this paper, we study this issue from a venture capital perspective by predicting idea-stage startup success based on founder traits. (i) To build a reliable prediction model, we introduce LLM-AR, a pipeline inspired by neural-symbolic systems that distils LLM-generated heuristics into probabilistic rules executed by the ProbLog automated-reasoning engine. (ii) An iterative policy-evolution loop incorporates association-rule mining to progressively refine the prediction rules. On unseen folds, LLM-AR achieves 59.5% precision and 8.7% recall, 5.9x the random baseline precision, while exposing every decision path for human inspection. The framework is interpretable and tunable via hyperparameters, showing promise to extend into other domains.
Authors: Shayne Longpre, Sneha Kudugunta, Niklas Muennighoff, I-Hung Hsu, Isaac Caswell, Alex Pentland, Sercan Arik, Chen-Yu Lee, Sayna Ebrahimi
Abstract: Scaling laws research has focused overwhelmingly on English -- yet the most prominent AI models explicitly serve billions of international users. In this work, we undertake the largest multilingual scaling laws study to date, totaling 774 multilingual training experiments, spanning 10M-8B model parameters, 400+ training languages and 48 evaluation languages. We introduce the Adaptive Transfer Scaling Law (ATLAS) for both monolingual and multilingual pretraining, which outperforms existing scaling laws' out-of-sample generalization often by more than 0.3 R^2. Our analyses of the experiments shed light on multilingual learning dynamics, transfer properties between languages, and the curse of multilinguality. First, we derive a cross-lingual transfer matrix, empirically measuring mutual benefit scores between 38 x 38=1444 language pairs. Second, we derive a language-agnostic scaling law that reveals how to optimally scale model size and data when adding languages without sacrificing performance. Third, we identify the computational crossover points for when to pretrain from scratch versus finetune from multilingual checkpoints. We hope these findings provide the scientific foundation for democratizing scaling laws across languages, and enable practitioners to efficiently scale models -- beyond English-first AI.
Authors: Zhimin Chen, Chenyu Zhao, Ka Chun Mo, Yunjiang Jiang, Jane H. Lee, Shouwei Chen, Khushhall Chandra Mahajan, Ning Jiang, Kai Ren, Jinhui Li, Wen-Yun Yang
Abstract: Modern large-scale recommendation systems rely heavily on user interaction history sequences to enhance the model performance. The advent of large language models and sequential modeling techniques, particularly transformer-like architectures, has led to significant advancements recently (e.g., HSTU, SIM, and TWIN models). While scaling to ultra-long user histories (10k to 100k items) generally improves model performance, it also creates significant challenges on latency, queries per second (QPS) and GPU cost in industry-scale recommendation systems. Existing models do not adequately address these industrial scalability issues. In this paper, we propose a novel two-stage modeling framework, namely VIrtual Sequential Target Attention (VISTA), which decomposes traditional target attention from a candidate item to user history items into two distinct stages: (1) user history summarization into a few hundred tokens; followed by (2) candidate item attention to those tokens. These summarization token embeddings are then cached in storage system and then utilized as sequence features for downstream model training and inference. This novel design for scalability enables VISTA to scale to lifelong user histories (up to one million items) while keeping downstream training and inference costs fixed, which is essential in industry. Our approach achieves significant improvements in offline and online metrics and has been successfully deployed on an industry leading recommendation platform serving billions of users.
Authors: Marcus Thomas
Abstract: Contemporary machine learning optimizes for predictive accuracy, yet systems that achieve state of the art performance remain causally opaque: their internal representations provide no principled handle for intervention. We can retrain such models, but we cannot surgically edit specific mechanisms while holding others fixed, because learned latent variables lack causal semantics. We argue for a conceptual reorientation: intelligence is the ability to build and refine explanations, falsifiable claims about manipulable structure that specify what changes and what remains invariant under intervention. Explanations subsume prediction but demand more: causal commitments that can be independently tested and corrected at the level of mechanisms. We introduce computational explanations, mappings from observations to intervention ready causal accounts. We instantiate these explanations with Energy Structured Causal Models (ESCMs), in which mechanisms are expressed as constraints (energy functions or vector fields) rather than explicit input output maps, and interventions act by local surgery on those constraints. This shift makes internal structure manipulable at the level where explanations live: which relations must hold, which can change, and what follows when they do. We provide concrete instantiations of the structural-causal principles LAP and ICM in the ESCM context, and also argue that empirical risk minimization systematically produces fractured, entangled representations, a failure we analyze as gauge ambiguity in encoder energy pairs. Finally, we show that under mild conditions, ESCMs recover standard SCM semantics. Building on Part I's principles (LAP, ICM, CAP) and its definition of intelligence as explanation-building under criticism, this paper offers a formal language for causal reasoning in systems that aspire to understand, not merely to predict.
Authors: Abhijit Chatterjee, Niraj K. Jha, Jonathan D. Cohen, Thomas L. Griffiths, Hongjing Lu, Diana Marculescu, Ashiqur Rasul, Keshab K. Parhi
Abstract: The field of artificial intelligence (AI) has taken a tight hold on broad aspects of society, industry, business, and governance in ways that dictate the prosperity and might of the world's economies. The AI market size is projected to grow from 189 billion USD in 2023 to 4.8 trillion USD by 2033. Currently, AI is dominated by large language models that exhibit linguistic and visual intelligence. However, training these models requires a massive amount of data scraped from the web as well as large amounts of energy (50--60 GWh to train GPT-4). Despite these costs, these models often hallucinate, a characteristic that prevents them from being deployed in critical application domains. In contrast, the human brain consumes only 20~W of power. What is needed is the next level of AI evolution in which lightweight domain-specific multimodal models with higher levels of intelligence can reason, plan, and make decisions in dynamic environments with real-time data and prior knowledge, while learning continuously and evolving in ways that enhance future decision-making capability. This will define the next wave of AI, progressing from today's large models, trained with vast amounts of data, to nimble energy-efficient domain-specific agents that can reason and think in a world full of uncertainty. To support such agents, hardware will need to be reimagined to allow energy efficiencies greater than 1000x over the state of the art. Such a vision of future AI systems is developed in this work.
Authors: Yuli Slavutsky, Sebastian Salazar, David M. Blei
Abstract: This paper studies prediction with multiple candidate models, where the goal is to combine their outputs. This task is especially challenging in heterogeneous settings, where different models may be better suited to different inputs. We propose input adaptive Bayesian Model Averaging (IA-BMA), a Bayesian method that assigns model weights conditional on the input. IA-BMA employs an input adaptive prior, and yields a posterior distribution that adapts to each prediction, which we estimate with amortized variational inference. We derive formal guarantees for its performance, relative to any single predictor selected per input. We evaluate IABMA across regression and classification tasks, studying data from personalized cancer treatment, credit-card fraud detection, and UCI datasets. IA-BMA consistently delivers more accurate and better-calibrated predictions than both non-adaptive baselines and existing adaptive methods.
Authors: Petros Prastakos, Kayhan Behdin, Rahul Mazumder
Abstract: Sparse variable selection improves interpretability and generalization in high-dimensional learning by selecting a small subset of informative features. Recent advances in Mixed Integer Programming (MIP) have enabled solving large-scale non-private sparse regression - known as Best Subset Selection (BSS) - with millions of variables in minutes. However, extending these algorithmic advances to the setting of Differential Privacy (DP) has remained largely unexplored. In this paper, we introduce two new pure differentially private estimators for sparse variable selection, levering modern MIP techniques. Our framework is general and applies broadly to problems like sparse regression or classification, and we provide theoretical support recovery guarantees in the case of BSS. Inspired by the exponential mechanism, we develop structured sampling procedures that efficiently explore the non-convex objective landscape, avoiding the exhaustive combinatorial search in the exponential mechanism. We complement our theoretical findings with extensive numerical experiments, using both least squares and hinge loss for our objective function, and demonstrate that our methods achieve state-of-the-art empirical support recovery, outperforming competing algorithms in settings with up to $p=10^4$.
Authors: Anchit Jain, Stephen Bates
Abstract: Decomposing prediction uncertainty into its aleatoric (irreducible) and epistemic (reducible) components is critical for the development and deployment of machine learning systems. A popular, principled measure for epistemic uncertainty is the mutual information between the response variable and model parameters. However, evaluating this measure requires access to the posterior distribution of the model parameters, which is challenging to compute. In view of this, we introduce a frequentist measure of epistemic uncertainty based on the bootstrap. Our main theoretical contribution is a novel asymptotic expansion that reveals that our proposed (frequentist) measure and the (Bayesian) mutual information are asymptotically equivalent. This provides frequentist interpretations to mutual information and new computational strategies for approximating it. Moreover, we link our proposed approach to the widely-used heuristic approach of deep ensembles, giving added perspective on their practical success.
Authors: Pavlos Ntais
Abstract: Large language models (LLMs) remain vulnerable to sophisticated prompt engineering attacks that exploit contextual framing to bypass safety mechanisms, posing significant risks in cybersecurity applications. We introduce Jailbreak Mimicry, a systematic methodology for training compact attacker models to automatically generate narrative-based jailbreak prompts in a one-shot manner. Our approach transforms adversarial prompt discovery from manual craftsmanship into a reproducible scientific process, enabling proactive vulnerability assessment in AI-driven security systems. Developed for the OpenAI GPT-OSS-20B Red-Teaming Challenge, we use parameter-efficient fine-tuning (LoRA) on Mistral-7B with a curated dataset derived from AdvBench, achieving an 81.0% Attack Success Rate (ASR) against GPT-OSS-20B on a held-out test set of 200 items. Cross-model evaluation reveals significant variation in vulnerability patterns: our attacks achieve 66.5% ASR against GPT-4, 79.5% on Llama-3 and 33.0% against Gemini 2.5 Flash, demonstrating both broad applicability and model-specific defensive strengths in cybersecurity contexts. This represents a 54x improvement over direct prompting (1.5% ASR) and demonstrates systematic vulnerabilities in current safety alignment approaches. Our analysis reveals that technical domains (Cybersecurity: 93% ASR) and deception-based attacks (Fraud: 87.8% ASR) are particularly vulnerable, highlighting threats to AI-integrated threat detection, malware analysis, and secure systems, while physical harm categories show greater resistance (55.6% ASR). We employ automated harmfulness evaluation using Claude Sonnet 4, cross-validated with human expert assessment, ensuring reliable and scalable evaluation for cybersecurity red-teaming. Finally, we analyze failure mechanisms and discuss defensive strategies to mitigate these vulnerabilities in AI for cybersecurity.
Authors: Shvetank Prakash, Andrew Cheng, Arya Tschand, Mark Mazumder, Varun Gohil, Jeffrey Ma, Jason Yik, Zishen Wan, Jessica Quaye, Elisavet Lydia Alvanaki, Avinash Kumar, Chandrashis Mazumdar, Tuhin Khare, Alexander Ingare, Ikechukwu Uchendu, Radhika Ghosal, Abhishek Tyagi, Chenyu Wang, Andrea Mattia Garavagno, Sarah Gu, Alice Guo, Grace Hur, Luca Carloni, Tushar Krishna, Ankita Nayak, Amir Yazdanbakhsh, Vijay Janapa Reddi
Abstract: The field of computer architecture, which bridges high-level software abstractions and low-level hardware implementations, remains absent from current large language model (LLM) evaluations. To this end, we present QuArch (pronounced 'quark'), the first benchmark designed to facilitate the development and evaluation of LLM knowledge and reasoning capabilities specifically in computer architecture. QuArch provides a comprehensive collection of 2,671 expert-validated question-answer (QA) pairs covering various aspects of computer architecture, including processor design, memory systems, and interconnection networks. Our evaluation reveals that while frontier models possess domain-specific knowledge, they struggle with skills that require higher-order thinking in computer architecture. Frontier model accuracies vary widely (from 34% to 72%) on these advanced questions, highlighting persistent gaps in architectural reasoning across analysis, design, and implementation QAs. By holistically assessing fundamental skills, QuArch provides a foundation for building and measuring LLM capabilities that can accelerate innovation in computing systems. With over 140 contributors from 40 institutions, this benchmark represents a community effort to set the standard for architectural reasoning in LLM evaluation.
Authors: Kayhan Behdin, Qingquan Song, Sriram Vasudevan, Jian Sheng, Xiaojing Ma, Z Zhou, Chuanrui Zhu, Guoyao Li, Chanh Nguyen, Sayan Ghosh, Hejian Sang, Ata Fatahi Baarzi, Sundara Raman Ramachandran, Xiaoqing Wang, Qing Lan, Vinay Y S, Qi Guo, Caleb Johnson, Zhipeng Wang, Fedor Borisyuk
Abstract: Large Language Models (LLMs) have demonstrated impressive quality when applied to predictive tasks such as relevance ranking and semantic search. However, deployment of such LLMs remains prohibitively expensive for industry applications with strict latency and throughput requirements. In this work, we present lessons and efficiency insights from developing a purely text-based decoder-only Small Language Model (SLM) for a semantic search application at LinkedIn. Particularly, we discuss model compression techniques such as pruning that allow us to reduce the model size by up to $40\%$ while maintaining the accuracy. Additionally, we present context compression techniques that allow us to reduce the input context length by up to $10$x with minimal loss of accuracy. Finally, we present practical lessons from optimizing the serving infrastructure for deploying such a system on GPUs at scale, serving millions of requests per second. Taken together, this allows us to increase our system's throughput by $10$x in a real-world deployment, while meeting our quality bar.
Authors: Wenxuan Bao, Ruxi Deng, Jingrui He
Abstract: Pretrained vision-language models such as CLIP achieve strong zero-shot generalization but remain vulnerable to distribution shifts caused by input corruptions. In this work, we investigate how corruptions affect CLIP's image embeddings and uncover a consistent phenomenon we term as embedding variance collapse, where both intra-class and inter-class variances shrink as corruption severity increases. We find that this collapse is closely tied to performance degradation, with inter-class variance strongly correlated with classification accuracy. To explain this phenomenon, we analyze how corruptions alter the structure of the embedding space. Our theoretical results suggest that the visual encoder tends to encode corruption-related signals, which dilute class-discriminative features and compress the representation geometry. We further show that maximizing inter-class variance, even when estimated from pseudo-labels, can provably enhance embedding quality. Based on this insight, we propose Mint, a simple test-time adaptation method that maximizes pseudo-label-based inter-class variance on the fly using a mean accumulator and a gradient accumulator. Mint operates effectively with small batch sizes and consistently improves performance across multiple corruption benchmarks and CLIP architectures. Our code is available at https://github.com/baowenxuan/Mint .
Authors: Eduardo Fabricio Gomes Trindade, Felipe Silveira de Almeida, Gioliano de Oliveira Braga, Rafael Pimenta de Mattos Paix\~ao, Pedro Henrique dos Santos Rocha, Lourenco Alves Pereira Jr
Abstract: Wi-Fi Channel State Information (CSI) has been extensively studied for sensing activities. However, its practical application in user authentication still needs to be explored. This study presents a novel approach to biometric authentication using Wi-Fi Channel State Information (CSI) data for palm recognition. The research delves into utilizing a Raspberry Pi encased in a custom-built box with antenna power reduced to 1dBm, which was used to capture CSI data from the right hands of 20 participants (10 men and 10 women). The dataset was normalized using MinMax scaling to ensure uniformity and accuracy. By focusing on biophysical aspects such as hand size, shape, angular spread between fingers, and finger phalanx lengths, among other characteristics, the study explores how these features affect electromagnetic signals, which are then reflected in Wi-Fi CSI, allowing for precise user identification. Five classification algorithms were evaluated, with the Random Forest classifier achieving an average F1-Score of 99.82\% using 10-fold cross-validation. Amplitude and Phase data were used, with each capture session recording approximately 1000 packets per second in five 5-second intervals for each User. This high accuracy highlights the potential of Wi-Fi CSI in developing robust and reliable user authentication systems based on palm biometric data.
Authors: Yuhang Gao, Xiang Xiang, Sheng Zhong, Guoyou Wang
Abstract: Vision-Language Models (VLMs) have shown significant progress in open-set challenges. However, the limited availability of 3D datasets hinders their effective application in 3D scene understanding. We propose LOC, a general language-guided framework adaptable to various occupancy networks, supporting both supervised and self-supervised learning paradigms. For self-supervised tasks, we employ a strategy that fuses multi-frame LiDAR points for dynamic/static scenes, using Poisson reconstruction to fill voids, and assigning semantics to voxels via K-Nearest Neighbor (KNN) to obtain comprehensive voxel representations. To mitigate feature over-homogenization caused by direct high-dimensional feature distillation, we introduce Densely Contrastive Learning (DCL). DCL leverages dense voxel semantic information and predefined textual prompts. This efficiently enhances open-set recognition without dense pixel-level supervision, and our framework can also leverage existing ground truth to further improve performance. Our model predicts dense voxel features embedded in the CLIP feature space, integrating textual and image pixel information, and classifies based on text and semantic similarity. Experiments on the nuScenes dataset demonstrate the method's superior performance, achieving high-precision predictions for known classes and distinguishing unknown classes without additional training data.
Authors: Yunhong Tao, Wenbing Tao, Xiang Xiang
Abstract: Low-light image enhancement (LLIE) aims at improving the perception or interpretability of an image captured in an environment with poor illumination. With the advent of deep learning, the LLIE technique has achieved significant breakthroughs. However, existing LLIE methods either ignore the important role of frequency domain information or fail to effectively promote the propagation and flow of information, limiting the LLIE performance. In this paper, we develop a novel frequency-spatial interaction-driven network (FSIDNet) for LLIE based on two-stage architecture. To be specific, the first stage is designed to restore the amplitude of low-light images to improve the lightness, and the second stage devotes to restore phase information to refine fine-grained structures. Considering that Frequency domain and spatial domain information are complementary and both favorable for LLIE, we further develop two frequency-spatial interaction blocks which mutually amalgamate the complementary spatial and frequency information to enhance the capability of the model. In addition, we construct the Information Exchange Module (IEM) to associate two stages by adequately incorporating cross-stage and cross-scale features to effectively promote the propagation and flow of information in the two-stage network structure. Finally, we conduct experiments on several widely used benchmark datasets (i.e., LOL-Real, LSRW-Huawei, etc.), which demonstrate that our method achieves the excellent performance in terms of visual results and quantitative metrics while preserving good model efficiency.
Authors: Austin A. Barr, Brij S. Karmur, Anthony J. Winder, Eddie Guo, John T. Lysack, James N. Scott, William F. Morrish, Muneer Eesa, Morgan Willson, David W. Cadotte, Michael M. H. Yang, Ian Y. M. Chan, Sanju Lama, Garnette R. Sutherland
Abstract: Machine learning in neurosurgery is limited by challenges in assembling large, high-quality imaging datasets. Synthetic data offers a scalable, privacy-preserving solution. We evaluated the feasibility of generating realistic lateral cervical spine radiographs using a denoising diffusion probabilistic model (DDPM) trained on 4,963 images from the Cervical Spine X-ray Atlas. Model performance was monitored via training/validation loss and Frechet inception distance, and synthetic image quality was assessed in a blinded "clinical Turing test" with six neuroradiologists and two spine-fellowship trained neurosurgeons. Experts reviewed 50 quartets containing one real and three synthetic images, identifying the real image and rating realism on a 4-point Likert scale. Experts correctly identified the real image in 29% of trials (Fleiss' kappa=0.061). Mean realism scores were comparable between real (3.323) and synthetic images (3.228, 3.258, and 3.320; p=0.383, 0.471, 1.000). Nearest-neighbor analysis found no evidence of memorization. We also provide a dataset of 20,063 synthetic radiographs. These results demonstrate that DDPM-generated cervical spine X-rays are statistically indistinguishable in realism and quality from real clinical images, offering a novel approach to creating large-scale neuroimaging datasets for ML applications in landmarking, segmentation, and classification.
Authors: Saranraj Nambusubramaniyan, Shervin Safavi, Raja Guru, Andreas Knoblauch
Abstract: Solving the synaptic Credit Assignment Problem(CAP) is central to learning in both biological and artificial neural systems. Finding an optimal solution for synaptic CAP means setting the synaptic weights that assign credit to each neuron for influencing the final output and behavior of neural networks or animals. Gradient-based methods solve this problem in artificial neural networks using back-propagation, however, not in the most efficient way. For instance, back-propagation requires a chain of top-down gradient computations. This leads to an expensive optimization process in terms of computing power and memory linked with well-known weight transport and update locking problems. To address these shortcomings, we take a NeuroAI approach and draw inspiration from neural Reinforcement Learning to develop a derivative-free optimizer for training neural networks, Dopamine. Dopamine is developed for Weight Perturbation (WP) learning that exploits stochastic updating of weights towards optima. It achieves this by minimizing the regret, a form of Reward Prediction Error (RPE) between the expected outcome from the perturbed model and the actual outcome from the unperturbed model. We use this RPE to adjust the learning rate in the network (i.e., creating an adaptive learning rate strategy, similar to the role of dopamine in the brain). We tested the Dopamine optimizer for training multi-layered perceptrons for XOR tasks, and recurrent neural networks for chaotic time series forecasting. Dopamine-trained models demonstrate accelerated convergence and outperform standard WP, and give comparable performance to gradient-based algorithms, while consuming significantly less computation and memory. Overall, the Dopamine optimizer not only finds robust solutions and comparable performance to the state-of-the-art Machine Learning optimizers but is also neurobiologically more plausible.
Authors: M. S. Hossain (Center for Computational and Data Sciences, Independent University, Bangladesh), M. S. H. Shahal (Center for Astronomy, Space Science and Astrophysics, Independent University, Bangladesh, Department of Computer Science and Engineering, Independent University, Bangladesh), A. Khan (Center for Computational and Data Sciences, Independent University, Bangladesh, Center for Astronomy, Space Science and Astrophysics, Independent University, Bangladesh), K. M. B. Asad (Center for Astronomy, Space Science and Astrophysics, Independent University, Bangladesh, Department of Physical Sciences, Independent University, Bangladesh), P. Saikia (Department of Astronomy and Physics, Yale University, USA), F. Akter (Department of Agricultural and Biosystems Engineering, North Dakota State University, USA), A. Ali (Center for Computational and Data Sciences, Independent University, Bangladesh, Department of Computer Science and Engineering, Independent University, Bangladesh), M. A. Amin (Center for Computational and Data Sciences, Independent University, Bangladesh, Department of Computer Science and Engineering, Independent University, Bangladesh), A. Momen (Center for Computational and Data Sciences, Independent University, Bangladesh, Center for Astronomy, Space Science and Astrophysics, Independent University, Bangladesh, Department of Physical Sciences, Independent University, Bangladesh), M. Hasan (Department of Computer Science and Engineering, Independent University, Bangladesh), A. K. M. M. Rahman (Center for Computational and Data Sciences, Independent University, Bangladesh, Department of Computer Science and Engineering, Independent University, Bangladesh)
Abstract: Wide-angle tail (WAT) and narrow-angle tail (NAT) radio active galactic nuclei (RAGNs) are key tracers of dense environments in galaxy groups and clusters, yet no machine-learning classifier of bent RAGNs has been trained using both unlabeled data and purely visually inspected labels. We release the RGC Python package, which includes two newly preprocessed labeled datasets of 639 WATs and NATs derived from a publicly available catalog of visually inspected sources, along with a semi-supervised RGC model that leverages 20,000 unlabeled RAGNs. The two labeled datasets in RGC were preprocessed using PyBDSF which retains spurious sources, and Photutils which removes them. The RGC model integrates the self-supervised framework BYOL (Bootstrap YOur Latent) with the supervised E2CNN (E2-equivariant Convolutional Neural Network) to form a semi-supervised binary classifier. The RGC model, when trained and evaluated on a dataset devoid of spurious sources, reaches peak performance, attaining an accuracy of 88.88% along with F1-scores of 0.90 for WATs and 0.85 for NATs. The model's attention patterns amid class imbalance suggest that this work can serve as a stepping stone toward developing physics-informed foundation models capable of identifying a broad range of AGN physical properties.
Authors: Giora Simchoni, Saharon Rosset
Abstract: We present MMbeddings, a probabilistic embedding approach that reinterprets categorical embeddings through the lens of nonlinear mixed models, effectively bridging classical statistical theory with modern deep learning. By treating embeddings as latent random effects within a variational autoencoder framework, our method substantially decreases the number of parameters -- from the conventional embedding approach of cardinality $\times$ embedding dimension, which quickly becomes infeasible with large cardinalities, to a significantly smaller, cardinality-independent number determined primarily by the encoder architecture. This reduction dramatically mitigates overfitting and computational burden in high-cardinality settings. Extensive experiments on simulated and real datasets, encompassing collaborative filtering and tabular regression tasks using varied architectures, demonstrate that MMbeddings consistently outperforms traditional embeddings, underscoring its potential across diverse machine learning applications.
Authors: Ollie Olby, Andreea Bacalum, Rory Baggott, Namid Stillman
Abstract: Execution algorithms are vital to modern trading, they enable market participants to execute large orders while minimising market impact and transaction costs. As these algorithms grow more sophisticated, optimising them becomes increasingly challenging. In this work, we present a reinforcement learning (RL) framework for discovering optimal execution strategies, evaluated within a reactive agent-based market simulator. This simulator creates reactive order flow and allows us to decompose slippage into its constituent components: market impact and execution risk. We assess the RL agent's performance using the efficient frontier based on work by Almgren and Chriss, measuring its ability to balance risk and cost. Results show that the RL-derived strategies consistently outperform baselines and operate near the efficient frontier, demonstrating a strong ability to optimise for risk and impact. These findings highlight the potential of reinforcement learning as a powerful tool in the trader's toolkit.
Authors: Jialin Song (Jackie), Yingheng Tang (Jackie), Pu Ren (Jackie), Shintaro Takayoshi (Jackie), Saurabh Sawant (Jackie), Yujie Zhu (Jackie), Jia-Mian Hu (Jackie), Andy Nonaka (Jackie), Michael W. Mahoney (Jackie), Benjamin Erichson (Jackie), Zhi (Jackie), Yao
Abstract: Simulating hybrid magnonic quantum systems remains a challenge due to the large disparity between the timescales of the two systems. We present a massively parallel GPU-based simulation framework that enables fully coupled, large-scale modeling of on-chip magnon-photon circuits. Our approach resolves the dynamic interaction between ferromagnetic and electromagnetic fields with high spatiotemporal fidelity. To accelerate design workflows, we develop a physics-informed machine learning surrogate trained on the simulation data, reducing computational cost while maintaining accuracy. This combined approach reveals real-time energy exchange dynamics and reproduces key phenomena such as anti-crossing behavior and the suppression of ferromagnetic resonance under strong electromagnetic fields. By addressing the multiscale and multiphysics challenges in magnon-photon modeling, our framework enables scalable simulation and rapid prototyping of next-generation quantum and spintronic devices.
Authors: Guan-Yan Yang, Farn Wang
Abstract: The integration of Artificial Intelligence (AI) into safety-critical systems introduces a new reliability paradigm: silent failures, where AI produces confident but incorrect outputs that can be dangerous. This paper introduces the Formal Assurance and Monitoring Environment (FAME), a novel framework that confronts this challenge. FAME synergizes the mathematical rigor of offline formal synthesis with the vigilance of online runtime monitoring to create a verifiable safety net around opaque AI components. We demonstrate its efficacy in an autonomous vehicle perception system, where FAME successfully detected 93.5% of critical safety violations that were otherwise silent. By contextualizing our framework within the ISO 26262 and ISO/PAS 8800 standards, we provide reliability engineers with a practical, certifiable pathway for deploying trustworthy AI. FAME represents a crucial shift from accepting probabilistic performance to enforcing provable safety in next-generation systems.
Authors: Krishna Gurugubelli
Abstract: Automated dysarthria detection and severity assessment from speech have attracted significant research attention due to their potential clinical impact. Despite rapid progress in acoustic modeling and deep learning, models still fall short of human expert performance. This manuscript provides a comprehensive analysis of the reasons behind this gap, emphasizing a conceptual divergence we term the ``perceptual-statistical gap''. We detail human expert perceptual processes, survey machine learning representations and methods, review existing literature on feature sets and modeling strategies, and present a theoretical analysis of limits imposed by label noise and inter-rater variability. We further outline practical strategies to narrow the gap, perceptually motivated features, self-supervised pretraining, ASR-informed objectives, multimodal fusion, human-in-the-loop training, and explainability methods. Finally, we propose experimental protocols and evaluation metrics aligned with clinical goals to guide future research toward clinically reliable and interpretable dysarthria assessment tools.
Authors: Jahidul Arafat, Sanjaya Poudel
Abstract: Chromatin sensitive partial wave spectroscopic (csPWS) microscopy enables label free detection of nanoscale chromatin packing alterations that occur before visible cellular transformation. However, manual nuclear segmentation limits population scale analysis needed for biomarker discovery in early cancer detection. The lack of annotated csPWS imaging data prevents direct use of standard deep learning methods. We present CFU Net, a hierarchical segmentation architecture trained with a three stage curriculum on synthetic multimodal data. CFU Net achieves near perfect performance on held out synthetic test data that represent diverse spectroscopic imaging conditions without manual annotations (Dice 0.9879, IoU 0.9895). Our approach uses physics based rendering that incorporates empirically supported chromatin packing statistics, Mie scattering models, and modality specific noise, combined with a curriculum that progresses from adversarial RGB pretraining to spectroscopic fine tuning and histology validation. CFU Net integrates five architectural elements (ConvNeXt backbone, Feature Pyramid Network, UNet plus plus dense connections, dual attention, and deep supervision) that together improve Dice over a baseline UNet by 8.3 percent. We demonstrate deployment ready INT8 quantization with 74.9 percent compression and 0.15 second inference, giving a 240 times throughput gain over manual analysis. Applied to more than ten thousand automatically segmented nuclei from synthetic test data, the pipeline extracts chromatin biomarkers that distinguish normal from pre cancerous tissue with large effect sizes (Cohens d between 1.31 and 2.98), reaching 94 percent classification accuracy. This work provides a general framework for synthetic to real transfer learning in specialized microscopy and open resources for community validation on clinical specimens.
Authors: Imran Khan (Independent Researcher)
Abstract: Prompt engineering, particularly Chain-of-Thought (CoT) prompting, significantly enhances LLM reasoning capabilities. We introduce "Sculpting," a constrained, rule-based prompting method designed to improve upon standard CoT by reducing errors from semantic ambiguity and flawed common sense. We evaluate three prompting strategies (Zero Shot, standard CoT, and Sculpting) across three OpenAI model generations (gpt-4o-mini, gpt-4o, gpt-5) using the GSM8K mathematical reasoning benchmark (1,317 problems). Our findings reveal a "Prompting Inversion": Sculpting provides advantages on gpt-4o (97% vs. 93% for standard CoT), but becomes detrimental on gpt-5 (94.00% vs. 96.36% for CoT on full benchmark). We trace this to a "Guardrail-to-Handcuff" transition where constraints preventing common-sense errors in mid-tier models induce hyper-literalism in advanced models. Our detailed error analysis demonstrates that optimal prompting strategies must co-evolve with model capabilities, suggesting simpler prompts for more capable models.
Authors: Anum Paracha, Junaid Arshad, Mohamed Ben Farah, Khalid Ismail
Abstract: Data poisoning attacks are a potential threat to machine learning (ML) models, aiming to manipulate training datasets to disrupt their performance. Existing defenses are mostly designed to mitigate specific poisoning attacks or are aligned with particular ML algorithms. Furthermore, most defenses are developed to secure deep neural networks or binary classifiers. However, traditional multiclass classifiers need attention to be secure from data poisoning attacks, as these models are significant in developing multi-modal applications. Therefore, this paper proposes SecureLearn, a two-layer attack-agnostic defense to defend multiclass models from poisoning attacks. It comprises two components of data sanitization and a new feature-oriented adversarial training. To ascertain the effectiveness of SecureLearn, we proposed a 3D evaluation matrix with three orthogonal dimensions: data poisoning attack, data sanitization and adversarial training. Benchmarking SecureLearn in a 3D matrix, a detailed analysis is conducted at different poisoning levels (10%-20%), particularly analysing accuracy, recall, F1-score, detection and correction rates, and false discovery rate. The experimentation is conducted for four ML algorithms, namely Random Forest (RF), Decision Tree (DT), Gaussian Naive Bayes (GNB) and Multilayer Perceptron (MLP), trained with three public datasets, against three poisoning attacks and compared with two existing mitigations. Our results highlight that SecureLearn is effective against the provided attacks. SecureLearn has strengthened resilience and adversarial robustness of traditional multiclass models and neural networks, confirming its generalization beyond algorithm-specific defenses. It consistently maintained accuracy above 90%, recall and F1-score above 75%. For neural networks, SecureLearn achieved 97% recall and F1-score against all selected poisoning attacks.
Authors: Devon A. Kelly, Christiana Chamon
Abstract: Wide-bandgap (WBG) technologies offer unprecedented improvements in power system efficiency, size, and performance, but also introduce unique sensor corruption and cybersecurity risks in industrial control systems (ICS), particularly due to high-frequency noise and sophisticated cyber-physical threats. This proof-of-concept (PoC) study demonstrates the adaptation of a noise-driven physically unclonable function (PUF) and machine learning (ML)-assisted anomaly detection framework to the demanding environment of WBG-based ICS sensor pathways. By extracting entropy from unavoidable WBG switching noise (up to 100 kHz) as a PUF source, and simultaneously using this noise as a real-time threat indicator, the proposed system unites hardware-level authentication and anomaly detection. Our approach integrates hybrid machine learning (ML) models with adaptive Bayesian filtering, providing robust and low-latency detection capabilities resilient to both natural electromagnetic interference (EMI) and active adversarial manipulation. Through detailed simulations of WBG modules under benign and attack scenarios--including EMI injection, signal tampering, and node impersonation--we achieve 95% detection accuracy and sub-millisecond processing latency. These results demonstrate the feasibility of physics-driven, dual-use noise exploitation as a scalable ICS defense primitive. Our findings lay the groundwork for next-generation security strategies that leverage inherent device characteristics, bridging hardware and artificial intelligence (AI) for enhanced protection of critical ICS infrastructure.
Authors: Hans Jarett Ong, Yoichi Chikahara, Tomoharu Iwata
Abstract: Uncovering the underlying causal mechanisms of complex real-world systems remains a significant challenge, as these systems often entail high data collection costs and involve unknown interventions. We introduce MetaCaDI, the first framework to cast the joint discovery of a causal graph and unknown interventions as a meta-learning problem. MetaCaDI is a Bayesian framework that learns a shared causal graph structure across multiple experiments and is optimized to rapidly adapt to new, few-shot intervention target prediction tasks. A key innovation is our model's analytical adaptation, which uses a closed-form solution to bypass expensive and potentially unstable gradient-based bilevel optimization. Extensive experiments on synthetic and complex gene expression data demonstrate that MetaCaDI significantly outperforms state-of-the-art methods. It excels at both causal graph recovery and identifying intervention targets from as few as 10 data instances, proving its robustness in data-scarce scenarios.
Authors: Matthias J. Ehrhardt, Davide Murari, Ferdia Sherry
Abstract: The existence of instabilities, for example in the form of adversarial examples, has given rise to a highly active area of research concerning itself with understanding and enhancing the stability of neural networks. We focus on a popular branch within this area which draws on connections to continuous dynamical systems and optimal control, giving a bird's eye view of this area. We identify and describe the fundamental concepts that underlie much of the existing work in this area. Following this, we go into more detail on a specific approach to designing stable neural networks, developing the theoretical background and giving a description of how these networks can be implemented. We provide code that implements the approach that can be adapted and extended by the reader. The code further includes a notebook with a fleshed-out toy example on adversarial robustness of image classification that can be run without heavy requirements on the reader's computer. We finish by discussing this toy example so that the reader can interactively follow along on their computer. This work will be included as a chapter of a book on scientific machine learning, which is currently under revision and aimed at students.
Authors: Jing Wang, Jiajun Liang, Jie Liu, Henglin Liu, Gongye Liu, Jun Zheng, Wanyuan Pang, Ao Ma, Zhenyu Xie, Xintao Wang, Meng Wang, Pengfei Wan, Xiaodan Liang
Abstract: Recently, GRPO-based reinforcement learning has shown remarkable progress in optimizing flow-matching models, effectively improving their alignment with task-specific rewards. Within these frameworks, the policy update relies on importance-ratio clipping to constrain overconfident positive and negative gradients. However, in practice, we observe a systematic shift in the importance-ratio distribution-its mean falls below 1 and its variance differs substantially across timesteps. This left-shifted and inconsistent distribution prevents positive-advantage samples from entering the clipped region, causing the mechanism to fail in constraining overconfident positive updates. As a result, the policy model inevitably enters an implicit over-optimization stage-while the proxy reward continues to increase, essential metrics such as image quality and text-prompt alignment deteriorate sharply, ultimately making the learned policy impractical for real-world use. To address this issue, we introduce GRPO-Guard, a simple yet effective enhancement to existing GRPO frameworks. Our method incorporates ratio normalization, which restores a balanced and step-consistent importance ratio, ensuring that PPO clipping properly constrains harmful updates across denoising timesteps. In addition, a gradient reweighting strategy equalizes policy gradients over noise conditions, preventing excessive updates from particular timestep regions. Together, these designs act as a regulated clipping mechanism, stabilizing optimization and substantially mitigating implicit over-optimization without relying on heavy KL regularization. Extensive experiments on multiple diffusion backbones (e.g., SD3.5M, Flux.1-dev) and diverse proxy tasks demonstrate that GRPO-Guard significantly reduces over-optimization while maintaining or even improving generation quality.
Authors: Xiao Hu, Yuansheng Lian, Ke Zhang, Yunxuan Li, Yuelong Su, Meng Li
Abstract: This study proposes an interpretable prediction framework with literature-informed fine-tuned (LIFT) LLMs for truck driving risk prediction. The framework integrates an LLM-driven Inference Core that predicts and explains truck driving risk, a Literature Processing Pipeline that filters and summarizes domain-specific literature into a literature knowledge base, and a Result Evaluator that evaluates the prediction performance as well as the interpretability of the LIFT LLM. After fine-tuning on a real-world truck driving risk dataset, the LIFT LLM achieved accurate risk prediction, outperforming benchmark models by 26.7% in recall and 10.1% in F1-score. Furthermore, guided by the literature knowledge base automatically constructed from 299 domain papers, the LIFT LLM produced variable importance ranking consistent with that derived from the benchmark model, while demonstrating robustness in interpretation results to various data sampling conditions. The LIFT LLM also identified potential risky scenarios by detecting key combination of variables in truck driving risk, which were verified by PERMANOVA tests. Finally, we demonstrated the contribution of the literature knowledge base and the fine-tuning process in the interpretability of the LIFT LLM, and discussed the potential of the LIFT LLM in data-driven knowledge discovery.
Authors: Changti Wu, Shijie Lian, Zihao Liu, Lei Zhang, Laurence Tianruo Yang, Kai Chen
Abstract: Solid geometry problem solving demands spatial mathematical reasoning that integrates spatial intelligence and symbolic reasoning. However, most existing multimodal mathematical reasoning benchmarks focus primarily on 2D plane geometry, rely on static datasets prone to data contamination and memorization, and evaluate models solely by final answers, overlooking the reasoning process. To address these limitations, we introduce DynaSolidGeo, the first dynamic benchmark for evaluating genuine spatial reasoning in Vision-Language Models (VLMs). Constructed through a semi-automatic annotation pipeline, DynaSolidGeo contains 503 expert-curated seed questions that can, in principle, dynamically generate an unbounded number of diverse multimodal text-visual instances. Beyond answer accuracy, we incorporate process evaluation based on expert-annotated reasoning chains to measure logical validity and causal coherence. Experiments across representative open-source and closed-source VLMs reveal large performance gaps, severe degradation in dynamic settings, and poor performance on tasks requiring high-level spatial intelligence, such as mental rotation and visualization. The code and dataset are available at \href{https://zgca-ai4edu.github.io/DynaSolidGeo/}{DynaSolidGeo}.
Authors: Seyed Ahmad Hosseini Miangoleh, Amin Jalal Aghdasian, Farzaneh Abdollahi
Abstract: In this paper, we propose Bootstrapped Language-Image Pretraining-driven Fused State Representation in Proximal Policy Optimization (BLIP-FusePPO), a novel multimodal reinforcement learning (RL) framework for autonomous lane-keeping (LK), in which semantic embeddings generated by a vision-language model (VLM) are directly fused with geometric states, LiDAR observations, and Proportional-Integral-Derivative-based (PID) control feedback within the agent observation space. The proposed method lets the agent learn driving rules that are aware of their surroundings and easy to understand by combining high-level scene understanding from the VLM with low-level control and spatial signals. Our architecture brings together semantic, geometric, and control-aware representations to make policy learning more robust. A hybrid reward function that includes semantic alignment, LK accuracy, obstacle avoidance, and speed regulation helps learning to be more efficient and generalizable. Our method is different from the approaches that only use semantic models to shape rewards. Instead, it directly embeds semantic features into the state representation. This cuts down on expensive runtime inference and makes sure that semantic guidance is always available. The simulation results show that the proposed model is better at LK stability and adaptability than the best vision-based and multimodal RL baselines in a wide range of difficult driving situations. We make our code publicly available.
Authors: Xiyu Luo, Haodong LI, Xinxing Cheng, He Zhao, Yang Hu, Xuan Song, Tianyang Zhang
Abstract: Image-to-image translation models have achieved notable success in converting images across visual domains and are increasingly used for medical tasks such as predicting post-operative outcomes and modeling disease progression. However, most existing methods primarily aim to match the target distribution and often neglect spatial correspondences between the source and translated images. This limitation can lead to structural inconsistencies and hallucinations, undermining the reliability and interpretability of the predictions. These challenges are accentuated in clinical applications by the stringent requirement for anatomical accuracy. In this work, we present TraceTrans, a novel deformable image translation model designed for post-operative prediction that generates images aligned with the target distribution while explicitly revealing spatial correspondences with the pre-operative input. The framework employs an encoder for feature extraction and dual decoders for predicting spatial deformations and synthesizing the translated image. The predicted deformation field imposes spatial constraints on the generated output, ensuring anatomical consistency with the source. Extensive experiments on medical cosmetology and brain MRI datasets demonstrate that TraceTrans delivers accurate and interpretable post-operative predictions, highlighting its potential for reliable clinical deployment.
Authors: Nader Nemati
Abstract: Deep neural networks can convert ECG page images into analyzable waveforms, yet centralized training often conflicts with cross-institutional privacy and deployment constraints. A cross-silo federated digitization framework is presented that trains a full-model nnU-Net segmentation backbone without sharing images and aggregates updates across sites under realistic non-IID heterogeneity (layout, grid style, scanner profile, noise). The protocol integrates three standard server-side aggregators--FedAvg, FedProx, and FedAdam--and couples secure aggregation with central, user-level differential privacy to align utility with formal guarantees. Key features include: (i) end-to-end full-model training and synchronization across clients; (ii) secure aggregation so the server only observes a clipped, weighted sum once a participation threshold is met; (iii) central Gaussian DP with Renyi accounting applied post-aggregation for auditable user-level privacy; and (iv) a calibration-aware digitization pipeline comprising page normalization, trace segmentation, grid-leakage suppression, and vectorization to twelve-lead signals. Experiments on ECG pages rendered from PTB-XL show consistently faster convergence and higher late-round plateaus with adaptive server updates (FedAdam) relative to FedAvg and FedProx, while approaching centralized performance. The privacy mechanism maintains competitive accuracy while preventing exposure of raw images or per-client updates, yielding deployable, auditable guarantees suitable for multi-institution settings.
Authors: Satyandra Guthula, Jaber Daneshamooz, Charles Fleming, Ashish Kundu, Walter Willinger, Arpit Gupta
Abstract: Forecasting on widely used benchmark time series data (e.g., ETT, Electricity, Taxi, and Exchange Rate, etc.) has favored smooth, seasonal series, but network telemetry time series -- traffic measurements at service, IP, or subnet granularity -- are instead highly bursty and intermittent, with heavy-tailed bursts and highly variable inactive periods. These properties place the latter in the statistical regimes made famous and popularized more than 20 years ago by B.~Mandelbrot. Yet forecasting such time series with modern-day AI architectures remains underexplored. We introduce NetBurst, an event-centric framework that reformulates forecasting as predicting when bursts occur and how large they are, using quantile-based codebooks and dual autoregressors. Across large-scale sets of production network telemetry time series and compared to strong baselines, such as Chronos, NetBurst reduces Mean Average Scaled Error (MASE) by 13--605x on service-level time series while preserving burstiness and producing embeddings that cluster 5x more cleanly than Chronos. In effect, our work highlights the benefits that modern AI can reap from leveraging Mandelbrot's pioneering studies for forecasting in bursty, intermittent, and heavy-tailed regimes, where its operational value for high-stakes decision making is of paramount interest.
Authors: Kaihua Chang
Abstract: Student Growth Percentiles (SGPs), widely adopted across U.S. state assessment systems, employ independent quantile regression followed by post-hoc correction using an isotonic projection method (\texttt{isotonize=TRUE} in the \texttt{SGP} R package) to address quantile crossing. We demonstrate this approach contains a fundamental methodological inconsistency: interpolation between independently-estimated, potentially crossed quantiles requires monotonicity, yet the post-hoc correction alters estimates in ways that may violate the quantile property $P(Y \leq \hat{Q}_{\tau}(Y|X) \mid X) = \tau$. We term this the \emph{interpolation paradox}. While theoretically sound constrained joint quantile regression (CJQR) eliminates crossing by enforcing non-crossing constraints during optimization, we analyze its computational complexity (often scaling poorly, e.g., $\mathcal{O}((qn)^3)$ for standard LP solvers) rendering it intractable for large-scale educational data ($n > 100{,}000$). We examine the SGP package's switch to the Frisch-Newton interior point method (\texttt{rq.method.for.large.n="fn"}) for large $N$, noting that while efficient for \emph{independent} QR, it doesn't resolve the joint problem's complexity or the paradox. We propose neural network-based multi-quantile regression (NNQR) with shared hidden layers as a practical alternative. Leveraging the convexity of the composite pinball loss, SGD-based optimization used in NN training can reliably approach the global optimum, offering scalability ($O(n)$) and implicitly reducing crossing. Our empirical analysis shows independent QR yields crossing, while both CJQR and NNQR enforce monotonicity. NNQR emerges as a viable, scalable alternative for operational SGP systems, aligning theoretical validity with computational feasibility.
Authors: Sayantan Choudhury, Nicolas Loizou
Abstract: Introduced by Korpelevich in 1976, the extragradient method (EG) has become a cornerstone technique for solving min-max optimization, root-finding problems, and variational inequalities (VIs). Despite its longstanding presence and significant attention within the optimization community, most works focusing on understanding its convergence guarantees assume the strong L-Lipschitz condition. In this work, building on the proposed assumptions by Zhang et al. [2024b] for minimization and Vankov et al.[2024] for VIs, we focus on the more relaxed $\alpha$-symmetric $(L_0, L_1)$-Lipschitz condition. This condition generalizes the standard Lipschitz assumption by allowing the Lipschitz constant to scale with the operator norm, providing a more refined characterization of problem structures in modern machine learning. Under the $\alpha$-symmetric $(L_0, L_1)$-Lipschitz condition, we propose a novel step size strategy for EG to solve root-finding problems and establish sublinear convergence rates for monotone operators and linear convergence rates for strongly monotone operators. Additionally, we prove local convergence guarantees for weak Minty operators. We supplement our analysis with experiments validating our theory and demonstrating the effectiveness and robustness of the proposed step sizes for EG.
Authors: Mouyang Cheng, Qiwei Wan, Bowen Yu, Eunbi Rha, Michael J Landry, Mingda Li
Abstract: High-temperature superconductors are essential for next-generation energy and quantum technologies, yet their performance is often limited by the critical current density ($J_c$), which is strongly influenced by microstructural defects. Optimizing $J_c$ through defect engineering is challenging due to the complex interplay of defect type, density, and spatial correlation. Here we present an integrated workflow that combines reinforcement learning (RL) with time-dependent Ginzburg-Landau (TDGL) simulations to autonomously identify optimal defect configurations that maximize $J_c$. In our framework, TDGL simulations generate current-voltage characteristics to evaluate $J_c$, which serves as the reward signal that guides the RL agent to iteratively refine defect configurations. We find that the agent discovers optimal defect densities and correlations in two-dimensional thin-film geometries, enhancing vortex pinning and $J_c$ relative to the pristine thin-film, approaching 60\% of theoretical depairing limit with up to 15-fold enhancement compared to random initialization. This RL-driven approach provides a scalable strategy for defect engineering, with broad implications for advancing HTS applications in fusion magnets, particle accelerators, and other high-field technologies.
Authors: Vijay Veerabadran, Fanyi Xiao, Nitin Kamra, Pedro Matias, Joy Chen, Caley Drooff, Brett D Roads, Riley Williams, Ethan Henderson, Xuanyi Zhao, Kevin Carlberg, Joseph Tighe, Karl Ridgeway
Abstract: There has been a surge of interest in assistive wearable agents: agents embodied in wearable form factors (e.g., smart glasses) who take assistive actions toward a user's goal/query (e.g. "Where did I leave my keys?"). In this work, we consider the important complementary problem of inferring that goal from multi-modal contextual observations. Solving this "goal inference" problem holds the promise of eliminating the effort needed to interact with such an agent. This work focuses on creating WAGIBench, a strong benchmark to measure progress in solving this problem using vision-language models (VLMs). Given the limited prior work in this area, we collected a novel dataset comprising 29 hours of multimodal data from 348 participants across 3,477 recordings, featuring ground-truth goals alongside accompanying visual, audio, digital, and longitudinal contextual observations. We validate that human performance exceeds model performance, achieving 93% multiple-choice accuracy compared with 84% for the best-performing VLM. Generative benchmark results that evaluate several families of modern vision-language models show that larger models perform significantly better on the task, yet remain far from practical usefulness, as they produce relevant goals only 55% of the time. Through a modality ablation, we show that models benefit from extra information in relevant modalities with minimal performance degradation from irrelevant modalities.
Authors: Siddharth Vishwanath, Ery Arias-Castro
Abstract: We develop a formal statistical framework for classical multidimensional scaling (CMDS) applied to noisy dissimilarity data. We establish distributional convergence results for the embeddings produced by CMDS for various noise models, which enable the construction of \emph{bona~fide} uniform confidence sets for the latent configuration, up to rigid transformations. We further propose bootstrap procedures for constructing these confidence sets and provide theoretical guarantees for their validity. We find that the multiplier bootstrap adapts automatically to heteroscedastic noise such as multiplicative noise, while the empirical bootstrap seems to require homoscedasticity. Either form of bootstrap, when valid, is shown to substantially improve finite-sample accuracy. The empirical performance of the proposed methods is demonstrated through numerical experiments.
Authors: Nivasini Ananthakrishnan, Yuval Dagan, Kunhe Yang
Abstract: Motivated by the question of how a principal can maximize its utility in repeated interactions with a learning agent, we study repeated games between an principal and an agent employing a mean-based learning algorithm. Prior work has shown that computing or even approximating the global Stackelberg value in similar settings can require an exponential number of rounds in the size of the agent's action space, making it computationally intractable. In contrast, we shift focus to the computation of local Stackelberg equilibria and introduce an algorithm that, within the smoothed analysis framework, constitutes a Polynomial Time Approximation Scheme (PTAS) for finding an epsilon-approximate local Stackelberg equilibrium. Notably, the algorithm's runtime is polynomial in the size of the agent's action space yet exponential in (1/epsilon) - a dependency we prove to be unavoidable.
Authors: Min Chen, Bingzhi Zhang, Quntao Zhuang, Junyu Liu
Abstract: Quantum imaginary time evolution (QITE) algorithm is one of the most promising variational quantum algorithms (VQAs), bridging the current era of Noisy Intermediate-Scale Quantum devices and the future of fully fault-tolerant quantum computing. Although practical demonstrations of QITE and its potential advantages over the general VQA trained with vanilla gradient descent (GD) in certain tasks have been reported, a first-principle, theoretical understanding of QITE remains limited. Here, we aim to develop an analytic theory for the dynamics of QITE. First, we show that QITE can be interpreted as a form of a general VQA trained with Quantum Natural Gradient Descent (QNGD), where the inverse quantum Fisher information matrix serves as the learning-rate tensor. This equivalence is established not only at the level of gradient update rules, but also through the action principle: the variational principle can be directly connected to the geometric geodesic distance in the quantum Fisher information metric, up to an integration constant. Second, for wide quantum neural networks, we employ the quantum neural tangent kernel framework to construct an analytic model for QITE. We prove that QITE always converges faster than GD-based VQA, though this advantage is suppressed by the exponential growth of Hilbert space dimension. This helps explain certain experimental results in quantum computational chemistry. Our theory encompasses linear, quadratic, and more general loss functions. We validate the analytic results through numerical simulations. Our findings establish a theoretical foundation for QITE dynamics and provide analytic insights for the first-principle design of variational quantum algorithms.
Authors: Yuanhe Tian, Junjie Liu, Xican Yang, Haishan Ye, Yan Song
Abstract: Pruning provides a practical solution to reduce the resources required to run large language models (LLMs) to benefit from their effective capabilities as well as control their cost for training and inference. Research on LLM pruning often ranks the importance of LLM parameters using their magnitudes and calibration-data activations and removes (or masks) the less important ones, accordingly reducing LLMs' size. However, these approaches primarily focus on preserving the LLM's ability to generate fluent sentences, while neglecting performance on specific domains and tasks. In this paper, we propose a simple yet effective pruning approach for LLMs that preserves task-specific capabilities while shrinking their parameter space. We first analyze how conventional pruning minimizes loss perturbation under general-domain calibration and extend this formulation by incorporating task-specific feature distributions into the importance computation of existing pruning algorithms. Thus, our framework computes separate importance scores using both general and task-specific calibration data, partitions parameters into shared and exclusive groups based on activation-norm differences, and then fuses their scores to guide the pruning process. This design enables our method to integrate seamlessly with various foundation pruning techniques and preserve the LLM's specialized abilities under compression. Experiments on widely used benchmarks demonstrate that our approach is effective and consistently outperforms the baselines with identical pruning ratios and different settings.
Authors: Gareth Hardwick, Haizhao Yang
Abstract: Solving partial differential equations (PDEs) with highly oscillatory solutions on complex domains remains a challenging and important problem. High-frequency oscillations and intricate geometries often result in prohibitively expensive representations for traditional numerical methods and lead to difficult optimization landscapes for machine learning-based approaches. In this work, we introduce an enhanced Finite Expression Method (FEX) designed to address these challenges with improved accuracy, interpretability, and computational efficiency. The proposed framework incorporates three key innovations: a symbolic spectral composition module that enables FEX to learn and represent multiscale oscillatory behavior; a redesigned linear input layer that significantly expands the expressivity of the model; and an eigenvalue formulation that extends FEX to a new class of problems involving eigenvalue PDEs. Through extensive numerical experiments, we demonstrate that FEX accurately resolves oscillatory PDEs on domains containing multiple holes of varying shapes and sizes. Compared with existing neural network-based solvers, FEX achieves substantially higher accuracy while yielding interpretable, closed-form solutions that expose the underlying structure of the problem. These advantages, often absent in conventional finite element, finite difference, and black-box neural approaches, highlight FEX as a powerful and transparent framework for solving complex PDEs.
Authors: Sze Chai Leung, Di Zhou, H. Jane Bae
Abstract: Optimal sensor placement (OSP) is critical for efficient, accurate monitoring, control, and inference in complex real-world systems. We propose a machine-learning-based feature attribution framework to identify OSP for the prediction of quantities of interest. Feature attribution quantifies input contributions to a model's output; however, it struggles with highly correlated input data often encountered in real-world applications. To address this, we propose a Correlation-Aware Attribution Framework (CAAF), which introduces a clustering step before performing feature attribution to reduce redundancy and enhance generalizability. We first illustrate the core principles of the proposed framework through a series of validation cases, then demonstrate its effectiveness in real-world dynamical systems, such as structural health monitoring, airfoil lift prediction, and wall-normal velocity estimation for turbulent channel flow. The results show that the CAAF outperforms alternative approaches that typically struggle due to the presence of nonlinear dynamics, chaotic behavior, and multi-scale interactions, and enables the effective application of feature attribution for identifying OSP in real-world environments.
Authors: Yang Tian, Fan Liu, Jingyuan Zhang, Wei Bi, Yupeng Hu, Liqiang Nie
Abstract: Large Multimodal Models (LMMs) have achieved remarkable progress in generating photorealistic and prompt-aligned images, but they often produce outputs that contradict verifiable knowledge, especially when prompts involve fine-grained attributes or time-sensitive events. Conventional retrieval-augmented approaches attempt to address this issue by introducing external information, yet they are fundamentally incapable of grounding generation in accurate and evolving knowledge due to their reliance on static sources and shallow evidence integration. To bridge this gap, we introduce ORIG, an agentic open multimodal retrieval-augmented framework for Factual Image Generation (FIG), a new task that requires both visual realism and factual grounding. ORIG iteratively retrieves and filters multimodal evidence from the web and incrementally integrates the refined knowledge into enriched prompts to guide generation. To support systematic evaluation, we build FIG-Eval, a benchmark spanning ten categories across perceptual, compositional, and temporal dimensions. Experiments demonstrate that ORIG substantially improves factual consistency and overall image quality over strong baselines, highlighting the potential of open multimodal retrieval for factual image generation.
Authors: Yicong Jiang, Zheng Tracy Ke
Abstract: Vertex hunting (VH) is the task of estimating a simplex from noisy data points and has many applications in areas such as network and text analysis. We introduce a new variant, semi-supervised vertex hunting (SSVH), in which partial information is available in the form of barycentric coordinates for some data points, known only up to an unknown transformation. To address this problem, we develop a method that leverages properties of orthogonal projection matrices, drawing on novel insights from linear algebra. We establish theoretical error bounds for our method and demonstrate that it achieves a faster convergence rate than existing unsupervised VH algorithms. Finally, we apply SSVH to two practical settings, semi-supervised network mixed membership estimation and semi-supervised topic modeling, resulting in efficient and scalable algorithms.
Authors: Morgan Himes, Samiksha Krishnamurthy, Andrew Lizarraga, Srinath Saikrishnan, Vikram Seenivasan, Jonathan Soriano, Ying Nian Wu, Tuan Do
Abstract: Upcoming surveys will produce billions of galaxy images but comparatively few spectra, motivating models that learn cross-modal representations. We build a dataset of 134,533 galaxy images (HSC-PDR2) and spectra (DESI-DR1) and adapt a Multi-Modal Masked Autoencoder (MMAE) to embed both images and spectra in a shared representation. The MMAE is a transformer-based architecture, which we train by masking 75% of the data and reconstructing missing image and spectral tokens. We use this model to test three applications: spectral and image reconstruction from heavily masked data and redshift regression from images alone. It recovers key physical features, such as galaxy shapes, atomic emission line peaks, and broad continuum slopes, though it struggles with fine image details and line strengths. For redshift regression, the MMAE performs comparably or better than prior multi-modal models in terms of prediction scatter even when missing spectra in testing. These results highlight both the potential and limitations of masked autoencoders in astrophysics and motivate extensions to additional modalities, such as text, for foundation models.
Authors: Noshitha Padma Pratyusha Juttu, Sahithi Singireddy, Sravani Gona, Sujal Timilsina
Abstract: Large Language Models (LLMs) have transformed text understanding, yet their adaptation to specialized legal domains remains constrained by the cost of full fine-tuning. This study provides a systematic evaluation of fine tuning, parameter efficient adaptation (LoRA, QLoRA), and zero-shot prompting strategies for unfair clause detection in Terms of Service (ToS) documents, a key application in legal NLP. We finetune BERT and DistilBERT, apply 4-bit Low-Rank Adaptation (LoRA) to models such as TinyLlama, LLaMA 3B/7B, and SaulLM, and evaluate GPT-4o and O-versions in zero-shot settings. Experiments on the CLAUDETTE-ToS benchmark and the Multilingual Scraper Corpus show that full fine-tuning achieves the strongest precision recall balance, while LoRA-based models provide competitive recall with up to 3x lower memory cost. These findings highlight practical design trade-offs for efficient and domain-adapted LLMs, contributing open baselines for fine-tuning research in legal text processing.
Authors: Heekang Song, Wan Choi
Abstract: In this paper, we propose an optimally structured gradient coding scheme to mitigate the straggler problem in distributed learning. Conventional gradient coding methods often assume homogeneous straggler models or rely on excessive data replication, limiting performance in real-world heterogeneous systems. To address these limitations, we formulate an optimization problem minimizing residual error while ensuring unbiased gradient estimation by explicitly considering individual straggler probabilities. We derive closed-form solutions for optimal encoding and decoding coefficients via Lagrangian duality and convex optimization, and propose data allocation strategies that reduce both redundancy and computation load. We also analyze convergence behavior for $\lambda$-strongly convex and $\mu$-smooth loss functions. Numerical results show that our approach significantly reduces the impact of stragglers and accelerates convergence compared to existing methods.
Authors: Pedro Chumpitaz-Flores, My Duong, Ying Mao, Kaixun Hua
Abstract: Clustering on NISQ hardware is constrained by data loading and limited qubits. We present \textbf{qc-kmeans}, a hybrid compressive $k$-means that summarizes a dataset with a constant-size Fourier-feature sketch and selects centroids by solving small per-group QUBOs with shallow QAOA circuits. The QFF sketch estimator is unbiased with mean-squared error $O(\varepsilon^2)$ for $B,S=\Theta(\varepsilon^{-2})$, and the peak-qubit requirement $q_{\text{peak}}=\max\{D,\lceil \log_2 B\rceil + 1\}$ does not scale with the number of samples. A refinement step with elitist retention ensures non-increasing surrogate cost. In Qiskit Aer simulations (depth $p{=}1$), the method ran with $\le 9$ qubits on low-dimensional synthetic benchmarks and achieved competitive sum-of-squared errors relative to quantum baselines; runtimes are not directly comparable. On nine real datasets (up to $4.3\times 10^5$ points), the pipeline maintained constant peak-qubit usage in simulation. Under IBM noise models, accuracy was similar to the idealized setting. Overall, qc-kmeans offers a NISQ-oriented formulation with shallow, bounded-width circuits and competitive clustering quality in simulation.
Authors: Dongyi Liu, Jiangtong Li, Dawei Cheng, Changjun Jiang
Abstract: Graph Neural Networks(GNNs) are vulnerable to backdoor attacks, where adversaries implant malicious triggers to manipulate model predictions. Existing trigger generators are often simplistic in structure and overly reliant on specific features, confining them to a single graph learning paradigm, such as graph supervised learning, graph contrastive learning, or graph prompt learning. This specialized design, which aligns the trigger with one learning objective, results in poor transferability when applied to other learning paradigms. For instance, triggers generated for the graph supervised learning paradigm perform poorly when tested within graph contrastive learning or graph prompt learning environments. Furthermore, these simple generators often fail to utilize complex structural information or node diversity within the graph data. These constraints limit the attack success rates of such methods in general testing scenarios. Therefore, to address these limitations, we propose Cross-Paradigm Graph Backdoor Attacks with Promptable Subgraph Triggers(CP-GBA), a new transferable graph backdoor attack that employs graph prompt learning(GPL) to train a set of universal subgraph triggers. First, we distill a compact yet expressive trigger set from target graphs, which is structured as a queryable repository, by jointly enforcing class-awareness, feature richness, and structural fidelity. Second, we conduct the first exploration of the theoretical transferability of GPL to train these triggers under prompt-based objectives, enabling effective generalization to diverse and unseen test-time paradigms. Extensive experiments across multiple real-world datasets and defense scenarios show that CP-GBA achieves state-of-the-art attack success rates.
Authors: Ibuki Maeda, Rentian Yao, Atsushi Nitanda
Abstract: The Schr\"odinger bridge problem seeks the optimal stochastic process that connects two given probability distributions with minimal energy modification. While the Sinkhorn algorithm is widely used to solve the static optimal transport problem, a recent work (Pooladian and Niles-Weed, 2024) proposed the Sinkhorn bridge, which estimates Schr\"odinger bridges by plugging optimal transport into the time-dependent drifts of SDEs, with statistical guarantees in the one-sample estimation setting where the true source distribution is fully accessible. In this work, to further justify this method, we study the statistical performance of intermediate Sinkhorn iterations in the two-sample estimation setting, where only finite samples from both source and target distributions are available. Specifically, we establish a statistical bound on the squared total variation error of Sinkhorn bridge iterations: $O(1/m+1/n + r^{4k})~(r \in (0,1))$, where $m$ and $n$ are the sample sizes from the source and target distributions, respectively, and $k$ is the number of Sinkhorn iterations. This result provides a theoretical guarantee for the finite-sample performance of the Schr\"odinger bridge estimator and offers practical guidance for selecting sample sizes and the number of Sinkhorn iterations. Notably, our theoretical results apply to several representative methods such as [SF]$^2$M, DSBM-IMF, BM2, and LightSB(-M) under specific settings, through the previously unnoticed connection between these estimators.
Authors: Archer Moore, Heejung Shim, Jingge Zhu, Mingming Gong
Abstract: Semi-supervised learning (SSL) aims to train a machine learning model using both labelled and unlabelled data. While the unlabelled data have been used in various ways to improve the prediction accuracy, the reason why unlabelled data could help is not fully understood. One interesting and promising direction is to understand SSL from a causal perspective. In light of the independent causal mechanisms principle, the unlabelled data can be helpful when the label causes the features but not vice versa. However, the causal relations between the features and labels can be complex in real world applications. In this paper, we propose a SSL framework that works with general causal models in which the variables have flexible causal relations. More specifically, we explore the causal graph structures and design corresponding causal generative models which can be learned with the help of unlabelled data. The learned causal generative model can generate synthetic labelled data for training a more accurate predictive model. We verify the effectiveness of our proposed method by empirical studies on both simulated and real data.
Authors: Bingqing Song, Jiaxiang Li, Rong Wang, Songtao Lu, Mingyi Hong
Abstract: Pre-trained large language models have demonstrated a strong ability to learn from context, known as in-context learning (ICL). Despite a surge of recent applications that leverage such capabilities, it is by no means clear, at least theoretically, how the ICL capabilities arise, and in particular, what is the precise role played by key factors such as pre-training procedure as well as context construction. In this work, we propose a new framework to analyze the ICL performance, for a class of realistic settings, which includes network architectures, data encoding, data generation, and prompt construction process. As a first step, we construct a simple example with a one-layer transformer, and show an interesting result, namely when the pre-train data distribution is different from the query task distribution, a properly constructed context can shift the output distribution towards the query task distribution, in a quantifiable manner, leading to accurate prediction on the query topic. We then extend the findings in the previous step to a more general case, and derive the precise relationship between ICL performance, context length and the KL divergence between pre-train and query task distribution. Finally, we provide experiments to validate our theoretical results.
Authors: Julia Bazinska, Max Mathys, Francesco Casucci, Mateo Rojas-Carulla, Xander Davies, Alexandra Souly, Niklas Pfister
Abstract: AI agents powered by large language models (LLMs) are being deployed at scale, yet we lack a systematic understanding of how the choice of backbone LLM affects agent security. The non-deterministic sequential nature of AI agents complicates security modeling, while the integration of traditional software with AI components entangles novel LLM vulnerabilities with conventional security risks. Existing frameworks only partially address these challenges as they either capture specific vulnerabilities only or require modeling of complete agents. To address these limitations, we introduce threat snapshots: a framework that isolates specific states in an agent's execution flow where LLM vulnerabilities manifest, enabling the systematic identification and categorization of security risks that propagate from the LLM to the agent level. We apply this framework to construct the $\operatorname{b}^3$ benchmark, a security benchmark based on 194331 unique crowdsourced adversarial attacks. We then evaluate 31 popular LLMs with it, revealing, among other insights, that enhanced reasoning capabilities improve security, while model size does not correlate with security. We release our benchmark, dataset, and evaluation code to facilitate widespread adoption by LLM providers and practitioners, offering guidance for agent developers and incentivizing model developers to prioritize backbone security improvements.
Authors: Jose Luis Ponton, Sheldon Andrews, Carlos Andujar, Nuria Pelechano
Abstract: Interactive applications demand believable characters that respond naturally to dynamic environments. Traditional character animation techniques often struggle to handle arbitrary situations, leading to a growing trend of dynamically selecting motion-captured animations based on predefined features. While Motion Matching has proven effective for locomotion by aligning to target trajectories, animating environment interactions and crowd behaviors remains challenging due to the need to consider surrounding elements. Existing approaches often involve manual setup or lack the naturalism of motion capture. Furthermore, in crowd animation, body animation is frequently treated as a separate process from trajectory planning, leading to inconsistencies between body pose and root motion. To address these limitations, we present Environment-aware Motion Matching, a novel real-time system for full-body character animation that dynamically adapts to obstacles and other agents, emphasizing the bidirectional relationship between pose and trajectory. In a preprocessing step, we extract shape, pose, and trajectory features from a motion capture database. At runtime, we perform an efficient search that matches user input and current pose while penalizing collisions with a dynamic environment. Our method allows characters to naturally adjust their pose and trajectory to navigate crowded scenes.
Authors: Shunta Akiyama
Abstract: In this paper, we consider a block coordinate descent (BCD) algorithm for training deep neural networks and provide a new global convergence guarantee under strictly monotonically increasing activation functions. While existing works demonstrate convergence to stationary points for BCD in neural networks, our contribution is the first to prove convergence to global minima, ensuring arbitrarily small loss. We show that the loss with respect to the output layer decreases exponentially while the loss with respect to the hidden layers remains well-controlled. Additionally, we derive generalization bounds using the Rademacher complexity framework, demonstrating that BCD not only achieves strong optimization guarantees but also provides favorable generalization performance. Moreover, we propose a modified BCD algorithm with skip connections and non-negative projection, extending our convergence guarantees to ReLU activation, which are not strictly monotonic. Empirical experiments confirm our theoretical findings, showing that the BCD algorithm achieves a small loss for strictly monotonic and ReLU activations.
Authors: Ruslan Berdichevsky, Shai Nahum-Gefen, Elad Ben Zaken
Abstract: Despite their impressive generalization capabilities, instruction-tuned Large Language Models often underperform on text classification benchmarks. We introduce SALSA, a coherent pipeline that combines structured prompting, class-to-token mapping, and parameter-efficient fine-tuning, thereby avoiding cold-start training. Each class label is mapped to a distinct output token, and prompts are constructed to elicit a single-token response. During inference, the model's output is projected only onto the logits of the relevant class tokens, enabling efficient and accurate classification in a single forward pass. SALSA achieves state-of-the-art results across diverse benchmarks, demonstrating its robustness and scalability for LLM-based classification applications.
Authors: Changwon Lee, Tak Hur, Daniel K. Park
Abstract: Real-time, scalable, and accurate decoding is a critical component for realizing a fault-tolerant quantum computer. While Transformer-based neural decoders such as \textit{AlphaQubit} have demonstrated high accuracy, the computational complexity of their core attention mechanism, which scales as $\mathcal{O}(d^4)$ with code distance $d$, results in decoding speeds insufficient for practical real-time applications. In this work, we introduce and evaluate a \textit{Mamba}-based decoder, a state-space model with $\mathcal{O}(d^2)$ complexity. In memory experiments using Sycamore hardware data, our Mamba decoder matches the performance of its Transformer-based counterpart, providing that its superior efficiency does not come at the cost of performance. Crucially, in simulated real-time scenarios that account for decoder-induced noise, the Mamba decoder significantly outperforms the Transformer, exhibiting a higher error threshold of $0.0104$ compared to $0.0097$. These results demonstrate that Mamba decoders offer a compelling balance between speed and accuracy, making them a promising architecture for scalable, real-time quantum error correction.
Authors: Kanad Pardeshi, Bryan Wilder, Aarti Singh
Abstract: Online learning algorithms continually update their models as data arrive, making it essential to accurately estimate the expected loss at the current time step. The prequential method is an effective estimation approach which can be practically deployed in various ways. However, theoretical guarantees have previously been established under strong conditions on the algorithm, and practical algorithms have hyperparameters which require careful tuning. We introduce OEUVRE, an estimator that evaluates each incoming sample on the function learned at the current and previous time steps, recursively updating the loss estimate in constant time and memory. We use algorithmic stability, a property satisfied by many popular online learners, for optimal updates and prove consistency, convergence rates, and concentration bounds for our estimator. We design a method to adaptively tune OEUVRE's hyperparameters and test it across diverse online and stochastic tasks. We observe that OEUVRE matches or outperforms other estimators even when their hyperparameters are tuned with oracle access to ground truth.
Authors: Swagatam Das
Abstract: This work develops a rigorous framework for diffusion-based generative modeling in the setting of free probability. We extend classical denoising diffusion probabilistic models to free diffusion processes -- stochastic dynamics acting on noncommutative random variables whose spectral measures evolve by free additive convolution. The forward dynamics satisfy a free Fokker--Planck equation that increases Voiculescu's free entropy and dissipates free Fisher information, providing a noncommutative analogue of the classical de Bruijn identity. Using tools from free stochastic analysis, including a free Malliavin calculus and a Clark--Ocone representation, we derive the reverse-time stochastic differential equation driven by the conjugate variable, the free analogue of the score function. We further develop a variational formulation of these flows in the free Wasserstein space, showing that the resulting gradient-flow structure converges to the semicircular equilibrium law. Together, these results connect modern diffusion models with the information geometry of free entropy and establish a mathematical foundation for generative modeling with operator-valued or high-dimensional structured data.
Authors: Michael Ungersb\"ock, Florian Gr\"otschla, Luca A. Lanzend\"orfer, June Young Yi, Changho Choi, Roger Wattenhofer
Abstract: Generative models have made significant progress in synthesizing high-fidelity audio from short textual descriptions. However, editing existing audio using natural language has remained largely underexplored. Current approaches either require the complete description of the edited audio or are constrained to predefined edit instructions that lack flexibility. In this work, we introduce SAO-Instruct, a model based on Stable Audio Open capable of editing audio clips using any free-form natural language instruction. To train our model, we create a dataset of audio editing triplets (input audio, edit instruction, output audio) using Prompt-to-Prompt, DDPM inversion, and a manual editing pipeline. Although partially trained on synthetic data, our model generalizes well to real in-the-wild audio clips and unseen edit instructions. We demonstrate that SAO-Instruct achieves competitive performance on objective metrics and outperforms other audio editing approaches in a subjective listening study. To encourage future research, we release our code and model weights.
Authors: Thu Phuong Nguyen, Duc M. Nguyen, Hyotaek Jeon, Hyunwook Lee, Hyunmin Song, Sungahn Ko, Taehwan Kim
Abstract: Automatically assessing handwritten mathematical solutions is an important problem in educational technology with practical applications, but it remains a significant challenge due to the diverse formats, unstructured layouts, and symbolic complexity of student work. To address this challenge, we introduce VEHME-a Vision-Language Model for Evaluating Handwritten Mathematics Expressions-designed to assess open-form handwritten math responses with high accuracy and interpretable reasoning traces. VEHME integrates a two-phase training pipeline: (i) supervised fine-tuning using structured reasoning data, and (ii) reinforcement learning that aligns model outputs with multi-dimensional grading objectives, including correctness, reasoning depth, and error localization. To enhance spatial understanding, we propose an Expression-Aware Visual Prompting Module, trained on our synthesized multi-line math expressions dataset to robustly guide attention in visually heterogeneous inputs. Evaluated on AIHub and FERMAT datasets, VEHME achieves state-of-the-art performance among open-source models and approaches the accuracy of proprietary systems, demonstrating its potential as a scalable and accessible tool for automated math assessment. Our training and experiment code is publicly available at our GitHub repository.
Authors: Zahraa Al Sahili, Maryam Fetanat, Maimuna Nowaz, Ioannis Patras, Matthew Purver
Abstract: Text-to-image (T2I) systems lack simple, reproducible ways to evaluate how well images match prompts and how models treat social attributes. Common proxies -- face classifiers and contrastive similarity -- reward surface cues, lack calibrated abstention, and miss attributes only weakly visible (for example, religion, culture, disability). We present FairJudge, a lightweight protocol that treats instruction-following multimodal LLMs as fair judges. It scores alignment with an explanation-oriented rubric mapped to [-1, 1]; constrains judgments to a closed label set; requires evidence grounded in the visible content; and mandates abstention when cues are insufficient. Unlike CLIP-only pipelines, FairJudge yields accountable, evidence-aware decisions; unlike mitigation that alters generators, it targets evaluation fairness. We evaluate gender, race, and age on FairFace, PaTA, and FairCoT; extend to religion, culture, and disability; and assess profession correctness and alignment on IdenProf, FairCoT-Professions, and our new DIVERSIFY-Professions. We also release DIVERSIFY, a 469-image corpus of diverse, non-iconic scenes. Across datasets, judge models outperform contrastive and face-centric baselines on demographic prediction and improve mean alignment while maintaining high profession accuracy, enabling more reliable, reproducible fairness audits.
Authors: Haowei Hua (Princeton University), Hong Jiao (University of Maryland, College Park), Xinyi Wang (University of Maryland, College Park & Beijing Normal University)
Abstract: BERT and its variants are extensively explored for automated scoring. However, a limit of 512 tokens for these encoder-based models showed the deficiency in automated scoring of long essays. Thus, this research explores generative language models for automated scoring of long essays via summarization and prompting. The results revealed great improvement of scoring accuracy with QWK increased from 0.822 to 0.8878 for the Learning Agency Lab Automated Essay Scoring 2.0 dataset.
Authors: Long H Dang, David Rawlinson
Abstract: The Hierarchical Reasoning Model (HRM) has impressive reasoning abilities given its small size, but has only been applied to supervised, static, fully-observable problems. One of HRM's strengths is its ability to adapt its computational effort to the difficulty of the problem. However, in its current form it cannot integrate and reuse computation from previous time-steps if the problem is dynamic, uncertain or partially observable, or be applied where the correct action is undefined, characteristics of many real-world problems. This paper presents HRM-Agent, a variant of HRM trained using only reinforcement learning. We show that HRM can learn to navigate to goals in dynamic and uncertain maze environments. Recent work suggests that HRM's reasoning abilities stem from its recurrent inference process. We explore the dynamics of the recurrent inference process and find evidence that it is successfully reusing computation from earlier environment time-steps.
Authors: Adrian Orenstein, Jessica Chen, Gwyneth Anne Delos Santos, Bayley Sapara, Michael Bowling
Abstract: While reinforcement learning agents can achieve superhuman performance in many complex tasks, they typically do not become more computationally efficient as they improve. In contrast, humans gradually require less cognitive effort as they become more proficient at a task. If agents could reason about their compute as they learn, could they similarly reduce their computation footprint? If they could, we could have more energy efficient agents or free up compute cycles for other processes like planning. In this paper, we experiment with showing agents the cost of their computation and giving them the ability to control when they use compute. We conduct our experiments on the Arcade Learning Environment, and our results demonstrate that with the same training compute budget, agents that reason about their compute perform better on 75% of games. Furthermore, these agents use three times less compute on average. We analyze individual games and show where agents gain these efficiencies.
Authors: Adam Stein, Neelay Velingker, Mayur Naik, Eric Wong
Abstract: Large language models (LLMs) excel at zero-shot inference but continue to struggle with complex, multi-step reasoning. Recent methods that augment LLMs with intermediate reasoning steps such as Chain of Thought (CoT) and Program of Thought (PoT) improve performance but often produce undesirable solutions, especially in algorithmic domains. We introduce Per-Instance Program Synthesis (PIPS), a method that generates and refines programs at the instance-level using structural feedback without relying on task-specific guidance or explicit test cases. To further improve performance, PIPS incorporates a confidence metric that dynamically chooses between direct inference and program synthesis on a per-instance basis. Experiments across three frontier LLMs and 30 benchmarks including all tasks of Big Bench Extra Hard (BBEH), visual question answering tasks, relational reasoning tasks, and mathematical reasoning tasks show that PIPS improves the absolute harmonic mean accuracy by up to 8.6% and 9.4% compared to PoT and CoT respectively, and reduces undesirable program generations by 65.1% on the algorithmic tasks compared to PoT with Gemini-2.0-Flash.
Authors: Tiasa Singha Roy, Ayush Rajesh Jhaveri, Ilias Triantafyllopoulos
Abstract: Despite their impressive capabilities, Large Language Models (LLMs) exhibit unwanted uncertainty, a phenomenon where a model changes a previously correct answer into an incorrect one when re-prompted. This behavior undermines trust and poses serious risks in high-stakes domains. In this work, we investigate the mechanisms that drive this phenomenon. We adapt the Needle-in-a-Haystack retrieval framework and integrate a Flip-style re-evaluation prompt to simulate realistic answer-flipping scenarios. We find that retrieval heads are not primarily responsible for avoiding uncertainty. Instead, we identify a small set of non-retrieval attention heads that disproportionately attend to misleading tokens in uncertain contexts. Masking these heads yields significant improvements, reducing flip behavior by up to 15% without introducing incoherence or overcorrection. However, when tested for downstream tasks, we observe trade-offs with flip behavior. Our findings contribute to the growing field of mechanistic interpretability and present a simple yet effective technique for mitigating uncertainty-driven failure modes in LLMs.
Authors: Thanyanee Srichaisak, Arissa Ieochai, Aueaphum Aueawattthanaphisut
Abstract: Background: Upper-limb weakness and tremor (4--12 Hz) limit activities of daily living (ADL) and reduce adherence to home rehabilitation. Objective: To assess technical feasibility and clinician-relevant signals of a sensor-fused wearable targeting the triceps brachii and extensor pollicis brevis. Methods: A lightweight node integrates surface EMG (1 kHz), IMU (100--200 Hz), and flex/force sensors with on-device INT8 inference (Tiny 1D-CNN/Transformer) and a safety-bounded assist policy (angle/torque/jerk limits; stall/time-out). Healthy adults (n = 12) performed three ADL-like tasks. Primary outcomes: Tremor Index (TI), range of motion (ROM), repetitions (Reps min$^{-1}$). Secondary: EMG median-frequency slope (fatigue trend), closed-loop latency, session completion, and device-related adverse events. Analyses used subject-level paired medians with BCa 95\% CIs; exact Wilcoxon $p$-values are reported in the Results. Results: Assistance was associated with lower tremor prominence and improved task throughput: TI decreased by $-0.092$ (95\% CI [$-0.102$, $-0.079$]), ROM increased by $+12.65\%$ (95\% CI [$+8.43$, $+13.89$]), and Reps rose by $+2.99$ min$^{-1}$ (95\% CI [$+2.61$, $+3.35$]). Median on-device latency was 8.7 ms at a 100 Hz loop rate; all sessions were completed with no device-related adverse events. Conclusions: Multimodal sensing with low-latency, safety-bounded assistance produced improved movement quality (TI $\downarrow$) and throughput (ROM, Reps $\uparrow$) in a pilot technical-feasibility setting, supporting progression to IRB-approved patient studies. Trial registration: Not applicable (pilot non-clinical).
Authors: Matthew So, Judah Goldfeder, Mark Lis, Hod Lipson
Abstract: There has been a historic assumption that the biometrics of an individual are statistically uncorrelated. We test this assumption by training Bi-Encoder networks on three verification tasks, including fingerprint-to-fingerprint matching, iris-to-iris matching, and cross-modal fingerprint-to-iris matching using 274 subjects with $\sim$100k fingerprints and 7k iris images. We trained ResNet-50 and Vision Transformer backbones in Bi-Encoder architectures such that the contrastive loss between images sampled from the same individual is minimized. The iris ResNet architecture reaches 91 ROC AUC score for iris-to-iris matching, providing clear evidence that the left and right irises of an individual are correlated. Fingerprint models reproduce the positive intra-subject suggested by prior work in this space. This is the first work attempting to use Vision Transformers for this matching. Cross-modal matching rises only slightly above chance, which suggests that more data and a more sophisticated pipeline is needed to obtain compelling results. These findings continue challenge independence assumptions of biometrics and we plan to extend this work to other biometrics in the future. Code available: https://github.com/MatthewSo/bio_fingerprints_iris.
Authors: Omar Allam, Brook Wander, Aayush R. Singh
Abstract: Large-scale datasets have enabled highly accurate machine learning interatomic potentials (MLIPs) for general-purpose heterogeneous catalysis modeling. There are, however, some limitations in what can be treated with these potentials because of gaps in the underlying training data. To extend these capabilities, we introduce AQCat25, a complementary dataset of 13.5 million density functional theory (DFT) single point calculations designed to improve the treatment of systems where spin polarization and/or higher fidelity are critical. We also investigate methodologies for integrating new datasets, such as AQCat25, with the broader Open Catalyst 2020 (OC20) dataset to create spin-aware models without sacrificing generalizability. We find that directly tuning a general model on AQCat25 leads to catastrophic forgetting of the original dataset's knowledge. Conversely, joint training strategies prove effective for improving accuracy on the new data without sacrificing general performance. This joint approach introduces a challenge, as the model must learn from a dataset containing both mixed-fidelity calculations and mixed-physics (spin-polarized vs. unpolarized). We show that explicitly conditioning the model on this system-specific metadata, for example by using Feature-wise Linear Modulation (FiLM), successfully addresses this challenge and further enhances model accuracy. Ultimately, our work establishes an effective protocol for bridging DFT fidelity domains to advance the predictive power of foundational models in catalysis.
Authors: Rishit Dagli, Donglai Xiang, Vismay Modi, Charles Loop, Clement Fuji Tsang, Anka He Chen, Anita Hu, Gavriel State, David I. W. Levin, Maria Shugrina
Abstract: Physical simulation relies on spatially-varying mechanical properties, often laboriously hand-crafted. VoMP is a feed-forward method trained to predict Young's modulus ($E$), Poisson's ratio ($\nu$), and density ($\rho$) throughout the volume of 3D objects, in any representation that can be rendered and voxelized. VoMP aggregates per-voxel multi-view features and passes them to our trained Geometry Transformer to predict per-voxel material latent codes. These latents reside on a manifold of physically plausible materials, which we learn from a real-world dataset, guaranteeing the validity of decoded per-voxel materials. To obtain object-level training data, we propose an annotation pipeline combining knowledge from segmented 3D datasets, material databases, and a vision-language model, along with a new benchmark. Experiments show that VoMP estimates accurate volumetric properties, far outperforming prior art in accuracy and speed.
Authors: Rui Hespanha, Elliot McGuire, Jo\~ao Hespanha
Abstract: Conventional fluid simulations can be time consuming and energy intensive. We researched the viability of a neural network for simulating incompressible fluids in a randomized obstacle-heavy environment, as an alternative to the numerical simulation of the Navier-Stokes equation. We hypothesized that the neural network predictions would have a relatively low error for simulations over a small number of time steps, but errors would eventually accumulate to the point that the output would become very noisy. Over a rich set of obstacle configurations, we achieved a root mean square error of 0.32% on our training dataset and 0.36% on a testing dataset. These errors only grew to 1.45% and 2.34% at t = 10 and, 2.11% and 4.16% at timestep t = 20. We also found that our selected neural network was approximately 8,800 times faster at predicting the flow than a conventional simulation. Our findings indicate neural networks can be extremely useful at simulating fluids in obstacle-heavy environments. Useful applications include modeling forest fire smoke, pipe fluid flow, and underwater/flood currents.
Authors: Wenxi Cai, Yuheng Wang, Naichen Shi
Abstract: We introduce Coupled Flow Matching (CPFM), a framework that integrates controllable dimensionality reduction and high-fidelity reconstruction. CPFM learns coupled continuous flows for both the high-dimensional data x and the low-dimensional embedding y, which enables sampling p(y|x) via a latent-space flow and p(x|y) via a data-space flow. Unlike classical dimension-reduction methods, where information discarded during compression is often difficult to recover, CPFM preserves the knowledge of residual information within the weights of a flow network. This design provides bespoke controllability: users may decide which semantic factors to retain explicitly in the latent space, while the complementary information remains recoverable through the flow network. Coupled flow matching builds on two components: (i) an extended Gromov-Wasserstein optimal transport objective that establishes a probabilistic correspondence between data and embeddings, and (ii) a dual-conditional flow-matching network that extrapolates the correspondence to the underlying space. Experiments on multiple benchmarks show that CPFM yields semantically rich embeddings and reconstructs data with higher fidelity than existing baselines.
Authors: Ran Xu, Jingjing Chen, Jiayu Ye, Yu Wu, Jun Yan, Carl Yang, Hongkun Yu
Abstract: Large Language Models (LLMs) are widely used as judges to evaluate response quality, providing a scalable alternative to human evaluation. However, most LLM judges operate solely on intrinsic text-based reasoning, limiting their ability to verify complex constraints or perform accurate computation. Motivated by the success of tool-integrated reasoning (TIR) in numerous tasks, we propose TIR-Judge, an end-to-end RL framework for training LLM judges that integrates a code executor for precise evaluation. TIR-Judge is built on three principles: (i) diverse training across verifiable and non-verifiable domains, (ii) flexible judgment formats (pointwise, pairwise, listwise), and (iii) iterative RL that bootstraps directly from the initial model without distillation. On seven public benchmarks, TIR-Judge surpasses strong reasoning-based judges by up to 6.4% (pointwise) and 7.7% (pairwise), and achieves listwise performance comparable to Claude-Opus-4 despite having only 8B parameters. Remarkably, TIR-Judge-Zero - trained entirely without distilled judge trajectories, matches the performance of distilled variants, demonstrating that tool-augmented judges can self-evolve through iterative reinforcement learning.
Authors: Jan Niklas Groeneveld, Xi Qin, Alexander Schaefer, Yaad Oren
Abstract: Generating high-quality code remains a challenge for Large Language Models (LLMs). For the evolution of reasoning models on this task, reward models are a necessary intermediate step. These models judge outcomes or intermediate steps. Decoder-only transformer models can be turned into reward models by introducing a regression layer and supervised fine-tuning. While it is known that reflection capabilities generally increase with the size of a model, we want to investigate whether state-of-the-art small language models like the Phi-4 family can be turned into usable reward models blending the consideration of process rewards and outcome rewards. Targeting this goal, we construct a dataset of code samples with correctness labels derived from the APPS coding challenge benchmark. We then train a value-head model to estimate the success probability of intermediate outputs. Our evaluation shows that small LLMs are capable of serving as effective reward models or code evaluation critics, successfully identifying correct solutions among multiple candidates. Using this critic, we achieve over a 20% improvement in the search capability of the most accurate code out of multiple generations.
Authors: Shiwei Li, Xiandi Luo, Haozhao Wang, Xing Tang, Ziqiang Cui, Dugang Liu, Yuhua Li, Xiuqiang He, Ruixuan Li
Abstract: Low-rank adaptation (LoRA) is a parameter-efficient fine-tuning (PEFT) method widely used in large language models (LLMs). LoRA essentially describes the projection of an input space into a low-dimensional output space, with the dimensionality determined by the LoRA rank. In standard LoRA, all input tokens share the same weights and undergo an identical input-output projection. This limits LoRA's ability to capture token-specific information due to the inherent semantic differences among tokens. To address this limitation, we propose Token-wise Projected Low-Rank Adaptation (TopLoRA), which dynamically adjusts LoRA weights according to the input token, thereby learning token-wise input-output projections in an end-to-end manner. Formally, the weights of TopLoRA can be expressed as $B\Sigma_X A$, where $A$ and $B$ are low-rank matrices (as in standard LoRA), and $\Sigma_X$ is a diagonal matrix generated from each input token $X$. Notably, TopLoRA does not increase the rank of LoRA weights but achieves more granular adaptation by learning token-wise LoRA weights (i.e., token-wise input-output projections). Extensive experiments across multiple models and datasets demonstrate that TopLoRA consistently outperforms LoRA and its variants. The code is available at https://github.com/Leopold1423/toplora-neurips25.
Authors: Rupasree Dey, Abdul Matin, Everett Lewark, Tanjim Bin Faruk, Andrei Bachinin, Sam Leuthold, M. Francesca Cotrufo, Shrideep Pallickara, Sangmi Lee Pallickara
Abstract: Soil salinization poses a significant threat to both ecosystems and agriculture because it limits plants' ability to absorb water and, in doing so, reduces crop productivity. This phenomenon alters the soil's spectral properties, creating a measurable relationship between salinity and light reflectance that enables remote monitoring. While laboratory spectroscopy provides precise measurements, its reliance on in-situ sampling limits scalability to regional or global levels. Conversely, hyperspectral satellite imagery enables wide-area observation but lacks the fine-grained interpretability of laboratory instruments. To bridge this gap, we introduce DeepSalt, a deep-learning-based spectral transfer framework that leverages knowledge distillation and a novel Spectral Adaptation Unit to transfer high-resolution spectral insights from laboratory-based spectroscopy to satellite-based hyperspectral sensing. Our approach eliminates the need for extensive ground sampling while enabling accurate, large-scale salinity estimation, as demonstrated through comprehensive empirical benchmarks. DeepSalt achieves significant performance gains over methods without explicit domain adaptation, underscoring the impact of the proposed Spectral Adaptation Unit and the knowledge distillation strategy. The model also effectively generalized to unseen geographic regions, explaining a substantial portion of the salinity variance.
Authors: Sarabeth S. Mullins, Georg G\"otz, Eric Bezzam, Steven Zheng, Daniel Gert Nielsen
Abstract: Accurate far-field speech datasets are critical for tasks such as automatic speech recognition (ASR), dereverberation, speech enhancement, and source separation. However, current datasets are limited by the trade-off between acoustic realism and scalability. Measured corpora provide faithful physics but are expensive, low-coverage, and rarely include paired clean and reverberant data. In contrast, most simulation-based datasets rely on simplified geometrical acoustics, thus failing to reproduce key physical phenomena like diffraction, scattering, and interference that govern sound propagation in complex environments. We introduce Treble10, a large-scale, physically accurate room-acoustic dataset. Treble10 contains over 3000 broadband room impulse responses (RIRs) simulated in 10 fully furnished real-world rooms, using a hybrid simulation paradigm implemented in the Treble SDK that combines a wave-based and geometrical acoustics solver. The dataset provides six complementary subsets, spanning mono, 8th-order Ambisonics, and 6-channel device RIRs, as well as pre-convolved reverberant speech scenes paired with LibriSpeech utterances. All signals are simulated at 32 kHz, accurately modelling low-frequency wave effects and high-frequency reflections. Treble10 bridges the realism gap between measurement and simulation, enabling reproducible, physically grounded evaluation and large-scale data augmentation for far-field speech tasks. The dataset is openly available via the Hugging Face Hub, and is intended as both a benchmark and a template for next-generation simulation-driven audio research.
Authors: Diego Marcondes
Abstract: Physics-informed statistical learning (PISL) integrates empirical data with physical knowledge to enhance the statistical performance of estimators. While PISL methods are widely used in practice, a comprehensive theoretical understanding of how informed regularization affects statistical properties is still missing. Specifically, two fundamental questions have yet to be fully addressed: (1) what is the trade-off between considering soft penalties versus hard constraints, and (2) what is the statistical gain of incorporating physical knowledge compared to purely data-driven empirical error minimisation. In this paper, we address these questions for PISL in convex classes of functions under physical knowledge expressed as linear equations by developing appropriate complexity dependent error rates based on the small-ball method. We show that, under suitable assumptions, (1) the error rates of physics-informed estimators are comparable to those of hard constrained empirical error minimisers, differing only by constant terms, and that (2) informed penalization can effectively reduce model complexity, akin to dimensionality reduction, thereby improving learning performance. This work establishes a theoretical framework for evaluating the statistical properties of physics-informed estimators in convex classes of functions, contributing to closing the gap between statistical theory and practical PISL, with potential applications to cases not yet explored in the literature.
Authors: Sixian Liu, Chen Xu, Qiang Wang, Donghai Shi, Yiwen Li
Abstract: Multimodal camera-LiDAR fusion technology has found extensive application in 3D object detection, demonstrating encouraging performance. However, existing methods exhibit significant performance degradation in challenging scenarios characterized by sensor degradation or environmental disturbances. We propose a novel Adaptive Gated Fusion (AG-Fusion) approach that selectively integrates cross-modal knowledge by identifying reliable patterns for robust detection in complex scenes. Specifically, we first project features from each modality into a unified BEV space and enhance them using a window-based attention mechanism. Subsequently, an adaptive gated fusion module based on cross-modal attention is designed to integrate these features into reliable BEV representations robust to challenging environments. Furthermore, we construct a new dataset named Excavator3D (E3D) focusing on challenging excavator operation scenarios to benchmark performance in complex conditions. Our method not only achieves competitive performance on the standard KITTI dataset with 93.92% accuracy, but also significantly outperforms the baseline by 24.88% on the challenging E3D dataset, demonstrating superior robustness to unreliable modal information in complex industrial scenes.
Authors: Zakaria Boutakka, Nouhaila Innan, Muhammed Shafique, Mohamed Bennai, Z. Sakhi
Abstract: Quantum computing presents a promising path toward precise quantum chemical simulations, particularly for systems that challenge classical methods. This work investigates the performance of the Variational Quantum Eigensolver (VQE) in estimating the ground-state energy of the silicon atom, a relatively heavy element that poses significant computational complexity. Within a hybrid quantum-classical optimization framework, we implement VQE using a range of ansatz, including Double Excitation Gates, ParticleConservingU2, UCCSD, and k-UpCCGSD, combined with various optimizers such as gradient descent, SPSA, and ADAM. The main contribution of this work lies in a systematic methodological exploration of how these configuration choices interact to influence VQE performance, establishing a structured benchmark for selecting optimal settings in quantum chemical simulations. Key findings show that parameter initialization plays a decisive role in the algorithm's stability, and that the combination of a chemically inspired ansatz with adaptive optimization yields superior convergence and precision compared to conventional approaches.
Authors: Arnav Sukhija, Lenart Treven, Jin Cheng, Florian D\"orfler, Stelian Coros, Andreas Krause
Abstract: Fixed-frequency control in robotics imposes a trade-off between the efficiency of low-frequency control and the robustness of high-frequency control, a limitation not seen in adaptable biological systems. We address this with a reinforcement learning approach in which policies jointly select control actions and their application durations, enabling robots to autonomously modulate their control frequency in response to situational demands. We validate our method with zero-shot sim-to-real experiments on two distinct hardware platforms: a high-speed RC car and a quadrupedal robot. Our method matches or outperforms fixed-frequency baselines in terms of rewards while significantly reducing the control frequency and exhibiting adaptive frequency control under real-world conditions.
Authors: Andrij Vasylenko, Federico Ottomano, Christopher M. Collins, Rahul Savani, Matthew S. Dyer, Matthew J. Rosseinsky
Abstract: Discovering materials with previously unreported crystal frameworks is key to achieving transformative functionality. Generative artificial intelligence offers a scalable means to propose candidate crystal structures, however existing approaches mainly reproduce decorated variants of established motifs rather than uncover new configurations. Here we develop a physics-informed diffusion method, supported by chemically grounded validation protocol, which embeds descriptors of compactness and local environment diversity to balance physical plausibility with structural novelty. Conditioning on these metrics improves generative performance across architectures, increasing the fraction of structures outside 100 most common prototypes up to 67%. When crystal structure prediction (CSP) is seeded with generative structures, most candidates (97%) are reconstructed by CSP, yielding 145 (66%) low-energy frameworks not matching any known prototypes. These results show that while generative models are not substitutes for CSP, their chemically informed, diversity-guided outputs can enhance CSP efficiency, establishing a practical generative-CSP synergy for discovery-oriented exploration of chemical space.
Authors: Ali Fata, Hossein Rahmani, Parinaz Soltanzadeh, Amirhossein Derakhshan, Behrouz Minaei Bidgoli
Abstract: Relation extraction between drugs plays a crucial role in identifying drug drug interactions and predicting side effects. The advancement of machine learning methods in relation extraction, along with the development of large medical text databases, has enabled the low cost extraction of such relations compared to other approaches that typically require expert knowledge. However, to the best of our knowledge, there are limited datasets specifically designed for drug drug relation extraction currently available. Therefore, employing transfer learning becomes necessary to apply machine learning methods in this domain. In this study, we propose DREAM, a method that first employs a trained relation extraction model to discover relations between entities and then applies this model to a corpus of medical texts to construct an ontology of drug relationships. The extracted relations are subsequently validated using a large language model. Quantitative results indicate that the LLM agreed with 71 of the relations extracted from a subset of PubMed abstracts. Furthermore, our qualitative analysis indicates that this approach can uncover ambiguities in the medical domain, highlighting the challenges inherent in relation extraction in this field.
Authors: Junpei Komiyama, Kyoungseok Jang, Junya Honda
Abstract: We consider the best arm identification problem, where the goal is to identify the arm with the highest mean reward from a set of $K$ arms under a limited sampling budget. This problem models many practical scenarios such as A/B testing. We consider a class of algorithms for this problem, which is provably minimax optimal up to a constant factor. This idea is a generalization of existing works in fixed-budget best arm identification, which are limited to a particular choice of risk measures. Based on the framework, we propose Almost Tracking, a closed-form algorithm that has a provable guarantee on the popular risk measure $H_1$. Unlike existing algorithms, Almost Tracking does not require the total budget in advance nor does it need to discard a significant part of samples, which gives a practical advantage. Through experiments on synthetic and real-world datasets, we show that our algorithm outperforms existing anytime algorithms as well as fixed-budget algorithms.
Authors: Alessandro Sestini, Joakim Bergdahl, Jean-Philippe Barrette-LaPierre, Florian Fuchs, Brady Chen, Micheal Jones, Linus Gissl\'en
Abstract: While several high profile video games have served as testbeds for Deep Reinforcement Learning (DRL), this technique has rarely been employed by the game industry for crafting authentic AI behaviors. Previous research focuses on training super-human agents with large models, which is impractical for game studios with limited resources aiming for human-like agents. This paper proposes a sample-efficient DRL method tailored for training and fine-tuning agents in industrial settings such as the video game industry. Our method improves sample efficiency of value-based DRL by leveraging pre-collected data and increasing network plasticity. We evaluate our method training a goalkeeper agent in EA SPORTS FC 25, one of the best-selling football simulations today. Our agent outperforms the game's built-in AI by 10% in ball saving rate. Ablation studies show that our method trains agents 50% faster compared to standard DRL methods. Finally, qualitative evaluation from domain experts indicates that our approach creates more human-like gameplay compared to hand-crafted agents. As a testimony of the impact of the approach, the method is intended to replace the hand-crafted counterpart in next iterations of the series.
Authors: Stefan M. Fischer, Johannes Kiechle, Laura Daza, Lina Felsner, Richard Osuala, Daniel M. Lang, Karim Lekadir, Jan C. Peeken, Julia A. Schnabel
Abstract: In this work, we introduce Progressive Growing of Patch Size, an automatic curriculum learning approach for 3D medical image segmentation. Our approach progressively increases the patch size during model training, resulting in an improved class balance for smaller patch sizes and accelerated convergence of the training process. We evaluate our curriculum approach in two settings: a resource-efficient mode and a performance mode, both regarding Dice score performance and computational costs across 15 diverse and popular 3D medical image segmentation tasks. The resource-efficient mode matches the Dice score performance of the conventional constant patch size sampling baseline with a notable reduction in training time to only 44%. The performance mode improves upon constant patch size segmentation results, achieving a statistically significant relative mean performance gain of 1.28% in Dice Score. Remarkably, across all 15 tasks, our proposed performance mode manages to surpass the constant patch size baseline in Dice Score performance, while simultaneously reducing training time to only 89%. The benefits are particularly pronounced for highly imbalanced tasks such as lesion segmentation tasks. Rigorous experiments demonstrate that our performance mode not only improves mean segmentation performance but also reduces performance variance, yielding more trustworthy model comparison. Furthermore, our findings reveal that the proposed curriculum sampling is not tied to a specific architecture but represents a broadly applicable strategy that consistently boosts performance across diverse segmentation models, including UNet, UNETR, and SwinUNETR. In summary, we show that this simple yet elegant transformation on input data substantially improves both Dice Score performance and training runtime, while being compatible across diverse segmentation backbones.
Authors: Tianyi Ma, Tengyao Wang, Richard J. Samworth
Abstract: We study in-context learning problems where a Transformer is pretrained on tasks drawn from a mixture distribution $\pi=\sum_{\alpha\in\mathcal{A}} \lambda_{\alpha} \pi_{\alpha}$, called the pretraining prior, in which each mixture component $\pi_{\alpha}$ is a distribution on tasks of a specific difficulty level indexed by $\alpha$. Our goal is to understand the performance of the pretrained Transformer when evaluated on a different test distribution $\mu$, consisting of tasks of fixed difficulty $\beta\in\mathcal{A}$, and with potential distribution shift relative to $\pi_\beta$, subject to the chi-squared divergence $\chi^2(\mu,\pi_{\beta})$ being at most $\kappa$. In particular, we consider nonparametric regression problems with random smoothness, and multi-index models with random smoothness as well as random effective dimension. We prove that a large Transformer pretrained on sufficient data achieves the optimal rate of convergence corresponding to the difficulty level $\beta$, uniformly over test distributions $\mu$ in the chi-squared divergence ball. Thus, the pretrained Transformer is able to achieve faster rates of convergence on easier tasks and is robust to distribution shift at test time. Finally, we prove that even if an estimator had access to the test distribution $\mu$, the convergence rate of its expected risk over $\mu$ could not be faster than that of our pretrained Transformers, thereby providing a more appropriate optimality guarantee than minimax lower bounds.
Authors: Riko Yokozawa, Kentaro Fujii, Yuta Nomura, Shingo Murata
Abstract: Autonomous robotic navigation in real-world environments requires exploration to acquire environmental information as well as goal-directed navigation in order to reach specified targets. Active inference (AIF) based on the free-energy principle provides a unified framework for these behaviors by minimizing the expected free energy (EFE), thereby combining epistemic and extrinsic values. To realize this practically, we propose a deep AIF framework that integrates a diffusion policy as the policy model and a multiple timescale recurrent state-space model (MTRSSM) as the world model. The diffusion policy generates diverse candidate actions while the MTRSSM predicts their long-horizon consequences through latent imagination, enabling action selection that minimizes EFE. Real-world navigation experiments demonstrated that our framework achieved higher success rates and fewer collisions compared with the baselines, particularly in exploration-demanding scenarios. These results highlight how AIF based on EFE minimization can unify exploration and goal-directed navigation in real-world robotic settings.
Authors: Mouhand Alkadri, Dania Desouki, Khloud Al Jallad
Abstract: The performance of Artificial Intelligence (AI) systems fundamentally depends on high-quality training data. However, low-resource languages like Arabic suffer from severe data scarcity. Moreover, the absence of child-specific speech corpora is an essential gap that poses significant challenges. To address this gap, we present our created dataset, Arabic Little STT, a dataset of Levantine Arabic child speech recorded in classrooms, containing 355 utterances from 288 children (ages 6 - 13). We further conduct a systematic assessment of Whisper, a state-of-the-art automatic speech recognition (ASR) model, on this dataset and compare its performance with adult Arabic benchmarks. Our evaluation across eight Whisper variants reveals that even the best-performing model (Large_v3) struggles significantly, achieving a 0.66 word error rate (WER) on child speech, starkly contrasting with its sub 0.20 WER on adult datasets. These results align with other research on English speech. Results highlight the critical need for dedicated child speech benchmarks and inclusive training data in ASR development. Emphasizing that such data must be governed by strict ethical and privacy frameworks to protect sensitive child information. We hope that this study provides an initial step for future work on equitable speech technologies for Arabic-speaking children. We hope that our publicly available dataset enrich the children's demographic representation in ASR datasets.
Authors: Cristian Simionescu
Abstract: This thesis works to address a pivotal challenge in medical image analysis: the reliance on extensive labeled datasets, which are often limited due to the need for expert annotation and constrained by privacy and legal issues. By focusing on the development of self-supervised learning techniques and domain adaptation methods, this research aims to circumvent these limitations, presenting a novel approach to enhance the utility and efficacy of deep learning in medical imaging. Central to this thesis is the development of the Medformer, an innovative neural network architecture designed for multitask learning and deep domain adaptation. This model is adept at pre-training on diverse medical image datasets, handling varying sizes and modalities, and is equipped with a dynamic input-output adaptation mechanism. This enables efficient processing and integration of a wide range of medical image types, from 2D X-rays to complex 3D MRIs, thus mitigating the dependency on large labeled datasets. Further, the thesis explores the current state of self-supervised learning in medical imaging. It introduces novel pretext tasks that are capable of extracting meaningful information from unlabeled data, significantly advancing the model's interpretative abilities. This approach is validated through rigorous experimentation, including the use of the MedMNIST dataset, demonstrating the model's proficiency in learning generalized features applicable to various downstream tasks. In summary, this thesis contributes to the advancement of medical image analysis by offering a scalable, adaptable framework that reduces reliance on labeled data. It paves the way for more accurate, efficient diagnostic tools in healthcare, signifying a major step forward in the application of deep learning in medical imaging.
Authors: Keiya Hirashima, Michiko S. Fujii, Takayuki R. Saitoh, Naoto Harada, Kentaro Nomura, Kohji Yoshikawa, Yutaka Hirai, Tetsuro Asano, Kana Moriwaki, Masaki Iwasawa, Takashi Okamoto, Junichiro Makino
Abstract: A major goal of computational astrophysics is to simulate the Milky Way Galaxy with sufficient resolution down to individual stars. However, the scaling fails due to some small-scale, short-timescale phenomena, such as supernova explosions. We have developed a novel integration scheme of $N$-body/hydrodynamics simulations working with machine learning. This approach bypasses the short timesteps caused by supernova explosions using a surrogate model, thereby improving scalability. With this method, we reached 300 billion particles using 148,900 nodes, equivalent to 7,147,200 CPU cores, breaking through the billion-particle barrier currently faced by state-of-the-art simulations. This resolution allows us to perform the first star-by-star galaxy simulation, which resolves individual stars in the Milky Way Galaxy. The performance scales over $10^4$ CPU cores, an upper limit in the current state-of-the-art simulations using both A64FX and X86-64 processors and NVIDIA CUDA GPUs.
Authors: Shovon Sengupta, Sunny Kumar Singh, Tanujit Chakraborty
Abstract: Accurate macroeconomic forecasting has become harder amid geopolitical disruptions, policy reversals, and volatile financial markets. Conventional vector autoregressions (VARs) overfit in high dimensional settings, while threshold VARs struggle with time varying interdependencies and complex parameter structures. We address these limitations by extending the Sims Zha Bayesian VAR with exogenous variables (SZBVARx) to incorporate domain-informed shrinkage and four newspaper based uncertainty shocks such as economic policy uncertainty, geopolitical risk, US equity market volatility, and US monetary policy uncertainty. The framework improves structural interpretability, mitigates dimensionality, and imposes empirically guided regularization. Using G7 data, we study spillovers from uncertainty shocks to five core variables (unemployment, real broad effective exchange rates, short term rates, oil prices, and CPI inflation), combining wavelet coherence (time frequency dynamics) with nonlinear local projections (state dependent impulse responses). Out-of-sample results at 12 and 24 month horizons show that SZBVARx outperforms 14 benchmarks, including classical VARs and leading machine learning models, as confirmed by Murphy difference diagrams, multivariate Diebold Mariano tests, and Giacomini White predictability tests. Credible Bayesian prediction intervals deliver robust uncertainty quantification for scenario analysis and risk management. The proposed SZBVARx offers G7 policymakers a transparent, well calibrated tool for modern macroeconomic forecasting under pervasive uncertainty.
Authors: Pratik N. Kalamkar, A. G. Phakatkar
Abstract: Opinions are central to almost all human activities and are key influencers of our behaviors. In current times due to growth of social networking website and increase in number of e-commerce site huge amount of opinions are now available on web. Given a set of evaluative statements that contain opinions (or sentiments) about an Entity, opinion mining aims to extract attributes and components of the object that have been commented on in each statement and to determine whether the comments are positive, negative or neutral. While lot of research recently has been done in field of opinion mining and some of it dealing with ranking of entities based on review or opinion set, classifying opinions into finer granularity level and then ranking entities has never been done before. In this paper method for opinion mining from statements at a deeper level of granularity is proposed. This is done by using fuzzy logic reasoning, after which entities are ranked as per this information.
Authors: Edoardo Manino, Bruno Farias, Rafael S\'a Menezes, Fedor Shmarov, Lucas C. Cordeiro
Abstract: The behaviour of neural network components must be proven correct before deployment in safety-critical systems. Unfortunately, existing neural network verification techniques cannot certify the absence of faults at the software level. In this paper, we show how to specify and verify that neural networks are safe, by explicitly reasoning about their floating-point implementation. In doing so, we construct NeuroCodeBench 2.0, a benchmark comprising 912 neural network verification examples that cover activation functions, common layers, and full neural networks of up to 170K parameters. Our verification suite is written in plain C and is compatible with the format of the International Competition on Software Verification (SV-COMP). Thanks to it, we can conduct the first rigorous evaluation of eight state-of-the-art software verifiers on neural network code. The results show that existing automated verification tools can correctly solve an average of 11% of our benchmark, while producing around 3% incorrect verdicts. At the same time, a historical analysis reveals that the release of our benchmark has already had a significantly positive impact on the latter.
Authors: Abolfazl Younesi, Zahra Najafabadi Samani, Thomas Fahringer
Abstract: Data pipelines are essential in stream processing as they enable the efficient collection, processing, and delivery of real-time data, supporting rapid data analysis. In this paper, we present AutoStreamPipe, a novel framework that employs Large Language Models (LLMs) to automate the design, generation, and deployment of stream processing pipelines. AutoStreamPipe bridges the semantic gap between high-level user intent and platform-specific implementations across distributed stream processing systems for structured multi-agent reasoning by integrating a Hypergraph of Thoughts (HGoT) as an extended version of GoT. AutoStreamPipe combines resilient execution strategies, advanced query analysis, and HGoT to deliver pipelines with good accuracy. Experimental evaluations on diverse pipelines demonstrate that AutoStreamPipe significantly reduces development time (x6.3) and error rates (x5.19), as measured by a novel Error-Free Score (EFS), compared to LLM code-generation methods.
Authors: Shayan Kiyani, Hamed Hassani, George Pappas, Aaron Roth
Abstract: Calibration has emerged as a foundational goal in ``trustworthy machine learning'', in part because of its strong decision theoretic semantics. Independent of the underlying distribution, and independent of the decision maker's utility function, calibration promises that amongst all policies mapping predictions to actions, the uniformly best policy is the one that ``trusts the predictions'' and acts as if they were correct. But this is true only of \emph{fully calibrated} forecasts, which are tractable to guarantee only for very low dimensional prediction problems. For higher dimensional prediction problems (e.g. when outcomes are multiclass), weaker forms of calibration have been studied that lack these decision theoretic properties. In this paper we study how a conservative decision maker should map predictions endowed with these weaker (``partial'') calibration guarantees to actions, in a way that is robust in a minimax sense: i.e. to maximize their expected utility in the worst case over distributions consistent with the calibration guarantees. We characterize their minimax optimal decision rule via a duality argument, and show that surprisingly, ``trusting the predictions and acting accordingly'' is recovered in this minimax sense by \emph{decision calibration} (and any strictly stronger notion of calibration), a substantially weaker and more tractable condition than full calibration. For calibration guarantees that fall short of decision calibration, the minimax optimal decision rule is still efficiently computable, and we provide an empirical evaluation of a natural one that applies to any regression model solved to optimize squared error.
Authors: Milad Sefidgaran, Kimia Nadjahi, Abdellatif Zaidi
Abstract: In this paper, we leverage stochastic projection and lossy compression to establish new conditional mutual information (CMI) bounds on the generalization error of statistical learning algorithms. It is shown that these bounds are generally tighter than the existing ones. In particular, we prove that for certain problem instances for which existing MI and CMI bounds were recently shown in Attias et al. [2024] and Livni [2023] to become vacuous or fail to describe the right generalization behavior, our bounds yield suitable generalization guarantees of the order of $\mathcal{O}(1/\sqrt{n})$, where $n$ is the size of the training dataset. Furthermore, we use our bounds to investigate the problem of data "memorization" raised in those works, and which asserts that there are learning problem instances for which any learning algorithm that has good prediction there exist distributions under which the algorithm must "memorize" a big fraction of the training dataset. We show that for every learning algorithm, there exists an auxiliary algorithm that does not memorize and which yields comparable generalization error for any data distribution. In part, this shows that memorization is not necessary for good generalization.
Authors: Hemish Ahuja, Samradh Bhardwaj, Kirti Dhir, Roman Bagdasarian, Ziwoong Jang
Abstract: Quantum phase transitions in Rydberg atom arrays present significant opportunities for studying many-body physics, yet distinguishing between different ordered phases without explicit order parameters remains challenging. We present a resource-efficient quantum machine learning approach combining classical shadow tomography with variational quantum circuits (VQCs) for binary phase classification of Z2 and Z3 ordered phases. Our pipeline processes 500 randomized measurements per 51-atom chain state, reconstructs shadow operators, performs PCA dimensionality reduction (514 features), and encodes features using angle embedding onto a 2-qubit parameterized circuit. The circuit employs RY-RZ angle encoding, strong entanglement via all-to-all CZ gates, and a minimal 2-parameter ansatz achieving depth 7. Training via simultaneous perturbation stochastic approximation (SPSA) with hinge loss converged in 120 iterations. The model achieved 100% test accuracy with perfect precision, recall, and F1 scores, demonstrating that minimal quantum resources suffice for high-accuracy phase classification. This work establishes pathways for quantum-enhanced condensed matter physics on near-term quantum devices.
Authors: Fatemeh Zahra Safaeipour, Jacob Chakareski, Morteza Hashemi
Abstract: Mobile edge devices (e.g., AR/VR headsets) typically need to complete timely inference tasks while operating with limited on-board computing and energy resources. In this paper, we investigate the problem of collaborative inference in wireless edge networks, where energy-constrained edge devices aim to complete inference tasks within given deadlines. These tasks are carried out using neural networks, and the edge device seeks to optimize inference performance under energy and delay constraints. The inference process can be split between the edge device and an edge server, thereby achieving collaborative inference over wireless networks. We formulate an inference utility optimization problem subject to energy and delay constraints, and propose a novel solution called Bayes-Split-Edge, which leverages Bayesian optimization for collaborative split inference over wireless edge networks. Our solution jointly optimizes the transmission power and the neural network split point. The Bayes-Split-Edge framework incorporates a novel hybrid acquisition function that balances inference task utility, sample efficiency, and constraint violation penalties. We evaluate our approach using the VGG19 model on the ImageNet-Mini dataset, and Resnet101 on Tiny-ImageNet, and real-world mMobile wireless channel datasets. Numerical results demonstrate that Bayes-Split-Edge achieves up to 2.4x reduction in evaluation cost compared to standard Bayesian optimization and achieves near-linear convergence. It also outperforms several baselines, including CMA-ES, DIRECT, exhaustive search, and Proximal Policy Optimization (PPO), while matching exhaustive search performance under tight constraints. These results confirm that the proposed framework provides a sample-efficient solution requiring maximum 20 function evaluations and constraint-aware optimization for wireless split inference in edge computing systems.
Authors: KC Santosh, Rodrigue Rizk, Longwei Wang
Abstract: The rapid advancement of Artificial Intelligence (AI) has led to unprecedented computational demands, raising significant environmental and ethical concerns. This paper critiques the prevailing reliance on large-scale, static datasets and monolithic training paradigms, advocating for a shift toward human-inspired, sustainable AI solutions. We introduce a novel framework, Human AI (HAI), which emphasizes incremental learning, carbon-aware optimization, and human-in-the-loop collaboration to enhance adaptability, efficiency, and accountability. By drawing parallels with biological cognition and leveraging dynamic architectures, HAI seeks to balance performance with ecological responsibility. We detail the theoretical foundations, system design, and operational principles that enable AI to learn continuously and contextually while minimizing carbon footprints and human annotation costs. Our approach addresses pressing challenges in active learning, continual adaptation, and energy-efficient model deployment, offering a pathway toward responsible, human-centered artificial intelligence.
Authors: Bernardo Torres, Manuel Moussallam, Gabriel Meseguer-Brocal
Abstract: Audio autoencoders learn useful, compressed audio representations, but their non-linear latent spaces prevent intuitive algebraic manipulation such as mixing or scaling. We introduce a simple training methodology to induce linearity in a high-compression Consistency Autoencoder (CAE) by using data augmentation, thereby inducing homogeneity (equivariance to scalar gain) and additivity (the decoder preserves addition) without altering the model's architecture or loss function. When trained with our method, the CAE exhibits linear behavior in both the encoder and decoder while preserving reconstruction fidelity. We test the practical utility of our learned space on music source composition and separation via simple latent arithmetic. This work presents a straightforward technique for constructing structured latent spaces, enabling more intuitive and efficient audio processing.
Authors: Anirban Das, Irtaza Khalid, Rafael Pe\~naloza, Steven Schockaert
Abstract: Designing models that can learn to reason in a systematic way is an important and long-standing challenge. In recent years, a wide range of solutions have been proposed for the specific case of systematic relational reasoning, including Neuro-Symbolic approaches, variants of the Transformer architecture, and specialised Graph Neural Networks. However, existing benchmarks for systematic relational reasoning focus on an overly simplified setting, based on the assumption that reasoning can be reduced to composing relational paths. In fact, this assumption is hard-baked into the architecture of several recent models, leading to approaches that can perform well on existing benchmarks but are difficult to generalise to other settings. To support further progress in the field of systematic relational reasoning with neural networks, we introduce NoRA, a new benchmark which adds several levels of difficulty and requires models to go beyond path-based reasoning.
Authors: Masahiro Kato
Abstract: We develop a direct debiased machine learning framework comprising Neyman targeted estimation and generalized Riesz regression. Our framework unifies Riesz regression for automatic debiased machine learning, covariate balancing, targeted maximum likelihood estimation (TMLE), and density-ratio estimation. In many problems involving causal effects or structural models, the parameters of interest depend on regression functions. Plugging regression functions estimated by machine learning methods into the identifying equations can yield poor performance because of first-stage bias. To reduce such bias, debiased machine learning employs Neyman orthogonal estimating equations. Debiased machine learning typically requires estimation of the Riesz representer and the regression function. For this problem, we develop a direct debiased machine learning framework with an end-to-end algorithm. We formulate estimation of the nuisance parameters, the regression function and the Riesz representer, as minimizing the discrepancy between Neyman orthogonal scores computed with known and unknown nuisance parameters, which we refer to as Neyman targeted estimation. Neyman targeted estimation includes Riesz representer estimation, and we measure discrepancies using the Bregman divergence. The Bregman divergence encompasses various loss functions as special cases, where the squared loss yields Riesz regression and the Kullback-Leibler divergence yields entropy balancing. We refer to this Riesz representer estimation as generalized Riesz regression. Neyman targeted estimation also yields TMLE as a special case for regression function estimation. Furthermore, for specific pairs of models and Riesz representer estimation methods, we can automatically obtain the covariate balancing property without explicitly solving the covariate balancing objective.
Authors: William R\'eveillard, Vasileios Saketos, Alexandre Proutiere, Richard Combes
Abstract: We introduce and study an online problem arising in question answering systems. In this problem, an agent must sequentially classify user-submitted queries represented by $d$-dimensional embeddings drawn i.i.d. from an unknown distribution. The agent may consult a costly human expert for the correct label, or guess on her own without receiving feedback. The goal is to minimize regret against an oracle with free expert access. When the time horizon $T$ is at least exponential in the embedding dimension $d$, one can learn the geometry of the class regions: in this regime, we propose the Conservative Hull-based Classifier (CHC), which maintains convex hulls of expert-labeled queries and calls the expert as soon as a query lands outside all known hulls. CHC attains $\mathcal{O}(\log^d T)$ regret in $T$ and is minimax optimal for $d=1$. Otherwise, the geometry cannot be reliably learned without additional distributional assumptions. We show that when the queries are drawn from a subgaussian mixture, for $T \le e^d$, a Center-based Classifier (CC) achieves regret proportional to $N\log{N}$ where $N$ is the number of labels. To bridge these regimes, we introduce the Generalized Hull-based Classifier (GHC), a practical extension of CHC that allows for more aggressive guessing via a tunable threshold parameter. Our approach is validated with experiments, notably on real-world question-answering datasets using embeddings derived from state-of-the-art large language models.
Authors: Zhaoyang Yu, Jiayi Zhang, Huixue Su, Yufan Zhao, Yifan Wu, Mingyi Deng, Jinyu Xiang, Yizhang Lin, Lingxiao Tang, Yingchao Li, Yuyu Luo, Bang Liu, Chenglin Wu
Abstract: Real-world tasks require decisions at varying granularities, and humans excel at this by leveraging a unified cognitive representation where planning is fundamentally understood as a high-level form of action. However, current Large Language Model (LLM)-based agents lack this crucial capability to operate fluidly across decision granularities. This limitation stems from existing paradigms that enforce a rigid separation between high-level planning and low-level action, which impairs dynamic adaptability and limits generalization. We propose ReCode (Recursive Code Generation), a novel paradigm that addresses this limitation by unifying planning and action within a single code representation. In this representation, ReCode treats high-level plans as abstract placeholder functions, which the agent then recursively decomposes into finer-grained sub-functions until reaching primitive actions. This recursive approach dissolves the rigid boundary between plan and action, enabling the agent to dynamically control its decision granularity. Furthermore, the recursive structure inherently generates rich, multi-granularity training data, enabling models to learn hierarchical decision-making processes. Extensive experiments show ReCode significantly surpasses advanced baselines in inference performance and demonstrates exceptional data efficiency in training, validating our core insight that unifying planning and action through recursive code generation is a powerful and effective approach to achieving universal granularity control. The code is available at https://github.com/FoundationAgents/ReCode.
Authors: Yash Jangir, Yidi Zhang, Kashu Yamazaki, Chenyu Zhang, Kuan-Hsun Tu, Tsung-Wei Ke, Lei Ke, Yonatan Bisk, Katerina Fragkiadaki
Abstract: The pursuit of robot generalists - instructable agents capable of performing diverse tasks across diverse environments - demands rigorous and scalable evaluation. Yet real-world testing of robot policies remains fundamentally constrained: it is labor-intensive, slow, unsafe at scale, and difficult to reproduce. Existing simulation benchmarks are similarly limited, as they train and test policies within the same synthetic domains and cannot assess models trained from real-world demonstrations or alternative simulation environments. As policies expand in scope and complexity, these barriers only intensify, since defining "success" in robotics often hinges on nuanced human judgments of execution quality. In this paper, we introduce a new benchmarking framework that overcomes these challenges by shifting VLA evaluation into large-scale simulated environments augmented with online human feedback. Leveraging advances in vision-language models, 2D-to-3D generative modeling, and differentiable rendering, our approach automatically converts video demonstrations from widely used robot datasets into simulated counterparts. Within these digital twins, we assess VLA policies using both automated VLM-guided scoring and scalable human preference judgments collected from crowdworkers, transforming human involvement from tedious scene setup, resetting, and safety supervision into lightweight preference comparisons. To measure robustness, we systematically perturb simulated environments along multiple axes, such as textures and object placements, stress-testing policy generalization under controlled variation. The result is a continuously evolving, reproducible, and scalable benchmark for real-world trained robot manipulation policies, addressing a critical missing capability in today's robotics landscape.
Authors: Junyoung Seo, Rodrigo Mira, Alexandros Haliassos, Stella Bounareli, Honglie Chen, Linh Tran, Seungryong Kim, Zoe Landgraf, Jie Shen
Abstract: Audio-driven human animation models often suffer from identity drift during temporal autoregressive generation, where characters gradually lose their identity over time. One solution is to generate keyframes as intermediate temporal anchors that prevent degradation, but this requires an additional keyframe generation stage and can restrict natural motion dynamics. To address this, we propose Lookahead Anchoring, which leverages keyframes from future timesteps ahead of the current generation window, rather than within it. This transforms keyframes from fixed boundaries into directional beacons: the model continuously pursues these future anchors while responding to immediate audio cues, maintaining consistent identity through persistent guidance. This also enables self-keyframing, where the reference image serves as the lookahead target, eliminating the need for keyframe generation entirely. We find that the temporal lookahead distance naturally controls the balance between expressivity and consistency: larger distances allow for greater motion freedom, while smaller ones strengthen identity adherence. When applied to three recent human animation models, Lookahead Anchoring achieves superior lip synchronization, identity preservation, and visual quality, demonstrating improved temporal conditioning across several different architectures. Video results are available at the following link: https://lookahead-anchoring.github.io.
Authors: Shuhong Zheng, Ashkan Mirzaei, Igor Gilitschenski
Abstract: Current 3D/4D generation methods are usually optimized for photorealism, efficiency, and aesthetics. However, they often fail to preserve the semantic identity of the subject across different viewpoints. Adapting generation methods with one or few images of a specific subject (also known as Personalization or Subject-driven generation) allows generating visual content that align with the identity of the subject. However, personalized 3D/4D generation is still largely underexplored. In this work, we introduce TIRE (Track, Inpaint, REsplat), a novel method for subject-driven 3D/4D generation. It takes an initial 3D asset produced by an existing 3D generative model as input and uses video tracking to identify the regions that need to be modified. Then, we adopt a subject-driven 2D inpainting model for progressively infilling the identified regions. Finally, we resplat the modified 2D multi-view observations back to 3D while still maintaining consistency. Extensive experiments demonstrate that our approach significantly improves identity preservation in 3D/4D generation compared to state-of-the-art methods. Our project website is available at https://zsh2000.github.io/track-inpaint-resplat.github.io/.
URLs: https://zsh2000.github.io/track-inpaint-resplat.github.io/.
Authors: Mara Daniels, Paul Hand, Reinhard Heckel
Abstract: Generative models, such as GANs, learn an explicit low-dimensional representation of a particular class of images, and so they may be used as natural image priors for solving inverse problems such as image restoration and compressive sensing. GAN priors have demonstrated impressive performance on these tasks, but they can exhibit substantial representation error for both in-distribution and out-of-distribution images, because of the mismatch between the learned, approximate image distribution and the data generating distribution. In this paper, we demonstrate a method for reducing the representation error of GAN priors by modeling images as the linear combination of a GAN prior with a Deep Decoder. The deep decoder is an underparameterized and most importantly unlearned natural signal model similar to the Deep Image Prior. No knowledge of the specific inverse problem is needed in the training of the GAN underlying our method. For compressive sensing and image superresolution, our hybrid model exhibits consistently higher PSNRs than both the GAN priors and Deep Decoder separately, both on in-distribution and out-of-distribution images. This model provides a method for extensibly and cheaply leveraging both the benefits of learned and unlearned image recovery priors in inverse problems.
Authors: Yuexiao Ma, Taisong Jin, Xiawu Zheng, Yan Wang, Huixia Li, Yongjian Wu, Guannan Jiang, Wei Zhang, Rongrong Ji
Abstract: To bridge the ever increasing gap between deep neural networks' complexity and hardware capability, network quantization has attracted more and more research attention. The latest trend of mixed precision quantization takes advantage of hardware's multiple bit-width arithmetic operations to unleash the full potential of network quantization. However, this also results in a difficult integer programming formulation, and forces most existing approaches to use an extremely time-consuming search process even with various relaxations. Instead of solving a problem of the original integer programming, we propose to optimize a proxy metric, the concept of network orthogonality, which is highly correlated with the loss of the integer programming but also easy to optimize with linear programming. This approach reduces the search time and required data amount by orders of magnitude, with little compromise on quantization accuracy. Specifically, we achieve 72.08% Top-1 accuracy on ResNet-18 with 6.7Mb, which does not require any searching iterations. Given the high efficiency and low data dependency of our algorithm, we used it for the post-training quantization, which achieve 71.27% Top-1 accuracy on MobileNetV2 with only 1.5Mb. Our code is available at https://github.com/MAC-AutoML/OMPQ.
Authors: Mara Daniels, Tyler Maunu, Paul Hand
Abstract: We consider the fundamental problem of sampling the optimal transport coupling between given source and target distributions. In certain cases, the optimal transport plan takes the form of a one-to-one mapping from the source support to the target support, but learning or even approximating such a map is computationally challenging for large and high-dimensional datasets due to the high cost of linear programming routines and an intrinsic curse of dimensionality. We study instead the Sinkhorn problem, a regularized form of optimal transport whose solutions are couplings between the source and the target distribution. We introduce a novel framework for learning the Sinkhorn coupling between two distributions in the form of a score-based generative model. Conditioned on source data, our procedure iterates Langevin Dynamics to sample target data according to the regularized optimal coupling. Key to this approach is a neural network parametrization of the Sinkhorn problem, and we prove convergence of gradient descent with respect to network parameters in this formulation. We demonstrate its empirical success on a variety of large scale optimal transport tasks.
Authors: Peyman Morteza
Abstract: We develop a mathematical framework to address a broad class of metric and preference learning problems within a Hilbert space. We obtain a novel representer theorem for the simultaneous task of metric and preference learning. Our key observation is that the representer theorem for this task can be derived by regularizing the problem with respect to the norm inherent in the task structure. For the general task of metric learning, our framework leads to a simple and self-contained representer theorem and offers new geometric insights into the derivation of representer theorems for this task. In the case of Reproducing Kernel Hilbert Spaces (RKHSs), we illustrate how our representer theorem can be used to express the solution of the learning problems in terms of finite kernel terms similar to classical representer theorems. Lastly, our representer theorem leads to a novel nonlinear algorithm for metric and preference learning. We compare our algorithm against challenging baseline methods on real-world rank inference benchmarks, where it achieves competitive performance. Notably, our approach significantly outperforms vanilla ideal point methods and surpasses strong baselines across multiple datasets. Code available at: https://github.com/PeymanMorteza/Metric-Preference-Learning-RKHS
URLs: https://github.com/PeymanMorteza/Metric-Preference-Learning-RKHS
Authors: Quintina L. Campbell, Jonathan Herington, Andrew D. White
Abstract: Machine learning models have dual-use potential, potentially serving both beneficial and malicious purposes. The development of open-source models in chemistry has specifically surfaced dual-use concerns around toxicological data and chemical warfare agents. We discuss a chain risk framework identifying three misuse pathways and corresponding mitigation strategies: inference-level, model-level, and data-level. At the data level, we introduce a model-agnostic noising method to increase prediction error in specific desired regions (sensitive regions). Our results show that selective noise induces variance and attenuation bias, whereas simply omitting sensitive data fails to prevent extrapolation. These findings hold for both molecular feature multilayer perceptrons and graph neural networks. Thus, noising molecular structures can enable open sharing of potential dual-use molecular data.
Authors: Jiajun Fan, Yuzheng Zhuang, Yuecheng Liu, Jianye Hao, Bin Wang, Jiangcheng Zhu, Hao Wang, Shu-Tao Xia
Abstract: The exploration problem is one of the main challenges in deep reinforcement learning (RL). Recent promising works tried to handle the problem with population-based methods, which collect samples with diverse behaviors derived from a population of different exploratory policies. Adaptive policy selection has been adopted for behavior control. However, the behavior selection space is largely limited by the predefined policy population, which further limits behavior diversity. In this paper, we propose a general framework called Learnable Behavioral Control (LBC) to address the limitation, which a) enables a significantly enlarged behavior selection space via formulating a hybrid behavior mapping from all policies; b) constructs a unified learnable process for behavior selection. We introduce LBC into distributed off-policy actor-critic methods and achieve behavior control via optimizing the selection of the behavior mappings with bandit-based meta-controllers. Our agents have achieved 10077.52% mean human normalized score and surpassed 24 human world records within 1B training frames in the Arcade Learning Environment, which demonstrates our significant state-of-the-art (SOTA) performance without degrading the sample efficiency.
Authors: Joseph D. Viviano, Omar G. Younis, Sanghyeok Choi, Victor Schmidt, Yoshua Bengio, Salem Lahlou
Abstract: The growing popularity of generative flow networks (GFlowNets or GFNs) from a range of researchers with diverse backgrounds and areas of expertise necessitates a library that facilitates the testing of new features (e.g., training losses and training policies) against standard benchmark implementations, or on a set of common environments. We present torchgfn, a PyTorch library that aims to address this need. Its core contribution is a modular and decoupled architecture which treats environments, neural network modules, and training objectives as interchangeable components. This provides users with a simple yet powerful API to facilitate rapid prototyping and novel research. Multiple examples are provided, replicating and unifying published results. The library is available on GitHub (https://github.com/GFNOrg/torchgfn) and on pypi (https://pypi.org/project/torchgfn/).
URLs: https://github.com/GFNOrg/torchgfn), https://pypi.org/project/torchgfn/).
Authors: Angela Zhou
Abstract: In consequential domains, it is often impossible to compel individuals to take treatment, so that optimal policy rules are merely suggestions in the presence of human non-adherence to treatment recommendations. Under heterogeneity, covariates may predict take-up of treatment and final outcome, but differently. While optimal treatment rules optimize causal outcomes across the population, access parity constraints or other fairness considerations on who receives treatment can be important. For example, in social services, a persistent puzzle is the gap in take-up of beneficial services among those who may benefit from them the most. We study causal identification and robust estimation of optimal treatment rules, including under potential violations of positivity. We consider fairness constraints such as demographic parity in treatment take-up, and other constraints, via constrained optimization. Our framework can be extended to handle algorithmic recommendations under an often-reasonable covariate-conditional exclusion restriction, using our robustness checks for lack of positivity in the recommendation. We develop a two-stage algorithm for solving over parametrized policy classes under general constraints to obtain variance-sensitive regret bounds. We illustrate the methods in three case studies based on data from reminders of SNAP benefits recertification, randomized encouragement to enroll in insurance, and from pretrial supervised release with electronic monitoring. While the specific remedy to inequities in algorithmic allocation is context-specific, it requires studying both take-up of decisions and downstream outcomes of them.
Authors: Haishuai Wang, Yang Gao, Xin Zheng, Peng Zhang, Jiajun Bu, Philip S. Yu
Abstract: Graph Neural Architecture Search (GNAS) has shown promising results in finding the best graph neural network architecture on a given graph dataset. However, existing GNAS methods still require intensive human labor and rich domain knowledge when designing the search space and search strategy. To this end, we integrate Large Language Models (LLMs) into GNAS and present a new GNAS model based on LLMs (GNAS-LLM for short). The basic idea of GNAS-LLM is to design a new class of GNAS prompts for LLMs to guide LLMs towards understanding the generative task of graph neural architectures. The prompts consist of descriptions of the search space, search strategy, and search feedback of GNAS. By iteratively running LLMs with the prompts, GNAS-LLM generates more accurate graph neural network architectures with fast convergence. Experimental results show that GNAS-LLM outperforms the state-of-the-art GNAS methods on four benchmark graph datasets, with an average improvement of 0.7% on the validation sets and 0.3% on the test sets. Besides, GNAS-LLM achieves an average improvement of 1.0% on the test sets based on the search space from AutoGEL.
Authors: Ngoc N. Tran, Lam Tran, Hoang Phan, Anh Bui, Tung Pham, Toan Tran, Dinh Phung, Trung Le
Abstract: Contrastive Learning first extracts features from unlabeled data, followed by linear probing with labeled data. Adversarial Contrastive Learning (ACL) integrates Adversarial Training into the first phase to enhance feature robustness against attacks in the probing phase. While ACL has shown strong empirical results, its theoretical understanding remains limited. Furthermore, while a fair amount of theoretical works analyze how the unsupervised loss can support the supervised loss in the probing phase, none has examined its role to the robust supervised loss. To fill this gap, our work develops rigorous theories to identify which components in the unsupervised training can help improve the robust supervised loss. Specifically, besides the adversarial contrastive loss, we reveal that the benign one, along with a global divergence between benign and adversarial examples can also improve robustness. Proper experiments are conducted to justify our findings.
Authors: Zexi Li, Zhiqi Li, Jie Lin, Tao Shen, Jun Xiao, Yike Guo, Tao Lin, Chao Wu
Abstract: Model fusion aims to integrate several deep neural network (DNN) models' knowledge into one by fusing parameters, and it has promising applications, such as improving the generalization of foundation models and parameter averaging in federated learning. However, models under different settings (data, hyperparameter, etc.) have diverse neuron permutations; in other words, from the perspective of loss landscape, they reside in different loss basins, thus hindering model fusion performances. To alleviate this issue, previous studies highlighted the role of permutation invariance and have developed methods to find correct network permutations for neuron alignment after training. Orthogonal to previous attempts, this paper studies training-time neuron alignment, improving model fusion without the need for post-matching. Training-time alignment is cheaper than post-alignment and is applicable in various model fusion scenarios. Starting from fundamental hypotheses and theorems, a simple yet lossless algorithm called TNA-PFN is introduced. TNA-PFN utilizes partially fixed neuron weights as anchors to reduce the potential of training-time permutations, and it is empirically validated in reducing the barriers of linear mode connectivity and multi-model fusion. It is also validated that TNA-PFN can improve the fusion of pretrained models under the setting of model soup (vision transformers) and ColD fusion (pretrained language models). Based on TNA-PFN, two federated learning methods, FedPFN and FedPNU, are proposed, showing the prospects of training-time neuron alignment. FedPFN and FedPNU reach state-of-the-art performances in federated learning under heterogeneous settings and can be compatible with the server-side algorithm.
Authors: Ammar Haydari, Dongjie Chen, Zhengfeng Lai, Michael Zhang, Chen-Nee Chuah
Abstract: Generative models have shown promising results in capturing human mobility characteristics and generating synthetic trajectories. However, it remains challenging to ensure that the generated geospatial mobility data is semantically realistic, including consistent location sequences, and reflects real-world characteristics, such as constraining on geospatial limits. We reformat human mobility modeling as an autoregressive generation task to address these issues, leveraging the Generative Pre-trained Transformer (GPT) architecture. To ensure its controllable generation to alleviate the above challenges, we propose a geospatially-aware generative model, MobilityGPT. We propose a gravity-based sampling method to train a transformer for semantic sequence similarity. Then, we constrained the training process via a road connectivity matrix that provides the connectivity of sequences in trajectory generation, thereby keeping generated trajectories in geospatial limits. Lastly, we proposed to construct a preference dataset for fine-tuning MobilityGPT via Reinforcement Learning from Trajectory Feedback (RLTF) mechanism, which minimizes the travel distance between training and the synthetically generated trajectories. Experiments on real-world datasets demonstrate MobilityGPT's superior performance over state-of-the-art methods in generating high-quality mobility trajectories that are closest to real data in terms of origin-destination similarity, trip length, travel radius, link, and gravity distributions. We release the source code and reference links to datasets at https://github.com/ammarhydr/MobilityGPT.
Authors: Imad Aouali
Abstract: Efficient decision-making in contextual bandits with large action spaces is challenging, as methods lacking additional prior information may suffer from computational and statistical inefficiencies. In this work, we leverage pre-trained diffusion models as priors to capture complex action distributions and introduce a diffusion-based decision framework for contextual bandits. We develop practical algorithms to efficiently approximate posteriors under diffusion priors, enabling flexible decision-making strategies. Empirical evaluations demonstrate the effectiveness and versatility of our approach across diverse contextual bandit settings.
Authors: Tianheng Ling, Julian Hoever, Chao Qian, Gregor Schiele
Abstract: In industrial and environmental monitoring, achieving real-time and precise fluid flow measurement remains a critical challenge. This study applies linear quantization in FPGA-based soft sensors for fluid flow estimation, significantly enhancing Neural Network model precision by overcoming the limitations of traditional fixed-point quantization. Our approach achieves up to a 10.10% reduction in Mean Squared Error and a notable 9.39% improvement in inference speed through targeted hardware optimizations. Validated across multiple data sets, our findings demonstrate that the optimized FPGA-based quantized models can provide efficient, accurate real-time inference, offering a viable alternative to cloud-based processing in pervasive autonomous systems.
Authors: Christopher J. Banks, Aeron Sanchez, Vicki Stewart, Kate Bowen, Thomas Doherty, Oliver Tearne, Graham Smith, Rowland R. Kao
Abstract: Diagnostic tests that can detect pre-clinical or sub-clinical infection, are one of the most powerful tools in our armoury of weapons to control infectious diseases. Considerable effort has been paid to improving diagnostic testing for human, plant and animal diseases, including strategies for targeting the use of diagnostic tests towards individuals who are more likely to be infected. We use machine learning to assess the surrounding risk landscape under which a diagnostic test is applied to augment its interpretation. We develop this to predict the occurrence of bovine tuberculosis incidents in cattle herds, exploiting the availability of exceptionally detailed testing records. We show that, without compromising test specificity, test sensitivity can be improved so that the proportion of infected herds detected improves by over 5 percentage points, or 240 additional infected herds detected in one year beyond those detected by the skin test alone. We also use feature importance testing for assessing the weighting of risk factors. While many factors are associated with increased risk of incidents, of note are several factors that suggest that in some herds there is a higher risk of infection going undetected.
Authors: Nils Graef
Abstract: He and Hofmann (arXiv:2311.01906) detailed a skipless transformer without the V and P (post-attention projection) linear layers, which reduces the total number of weights. However, this scheme is only applicable to MHA (multi-head attention), but not for MQA (multi-query attention) and GQA (grouped-query attention). The latter schemes are used by many popular LLMs such as Llama 2, Mistral, Mixtral, PaLM, and Gemma. Therefore, this micro-paper proposes mathematically equivalent versions that are suitable for MQA and GQA. For example, removing Q and P from a skipless version of Mistral-7B would remove 15% of its weights (and thus reduce its compute and memory complexity). Watch our explainer video https://youtu.be/Tx_lMpphd2g and see https://github.com/OpenMachine-ai/transformer-tricks for code and more transformer tricks.
URLs: https://youtu.be/Tx_lMpphd2g, https://github.com/OpenMachine-ai/transformer-tricks
Authors: Dongwei Lyu, Rie Nakata, Pu Ren, Michael W. Mahoney, Arben Pitarka, Nori Nakata, N. Benjamin Erichson
Abstract: We propose a new deep learning model, WaveCastNet, to forecast high-dimensional wavefields. WaveCastNet integrates a convolutional long expressive memory architecture into a sequence-to-sequence forecasting framework, enabling it to model long-term dependencies and multiscale patterns in both space and time. By sharing weights across spatial and temporal dimensions, WaveCastNet requires significantly fewer parameters than more resource-intensive models such as transformers, resulting in faster inference times. Crucially, WaveCastNet also generalizes better than transformers to rare and critical seismic scenarios, such as high-magnitude earthquakes. Here, we show the ability of the model to predict the intensity and timing of destructive ground motions in real time, using simulated data from the San Francisco Bay Area. Furthermore, we demonstrate its zero-shot capabilities by evaluating WaveCastNet on real earthquake data. Our approach does not require estimating earthquake magnitudes and epicenters, steps that are prone to error in conventional methods, nor does it rely on empirical ground-motion models, which often fail to capture strongly heterogeneous wave propagation effects.
Authors: Sungho Jeon, Xinyue Ma, Kwang In Kim, Myeongjae Jeon
Abstract: Recent rehearsal-free continual learning (CL) methods guided by prompts achieve strong performance on vision tasks with non-stationary data but remain resource-intensive, hindering real-world edge deployment. We introduce resource-efficient prompting (REP), which improves the computational and memory efficiency of prompt-based rehearsal-free continual learning methods while minimizing accuracy trade-offs. Our approach employs swift prompt selection to refine input data using a carefully provisioned model and introduces adaptive token merging (AToM) and adaptive layer dropping (ALD) for efficient prompt updates. AToM and ALD selectively skip data and model layers while preserving task-specific features during the learning of new tasks. Extensive experiments on multiple image classification datasets demonstrate REP's superior resource efficiency over state-of-the-art rehearsal-free CL methods.
Authors: Aladin Djuhera, Vlad C. Andrei, Xinyang Li, Ullrich J. M\"onich, Holger Boche, Walid Saad
Abstract: Split federated learning (SFL) is a compute-efficient paradigm in distributed machine learning (ML), where components of large ML models are outsourced to remote servers. A significant challenge in SFL, particularly when deployed over wireless channels, is the susceptibility of transmitted model parameters to adversarial jamming that could jeopardize the learning process. This is particularly pronounced for embedding parameters in large language models (LLMs) and vision language models (VLMs), which are learned feature vectors essential for domain understanding. In this paper, rigorous insights are provided into the influence of jamming embeddings in SFL by deriving an expression for the ML training loss divergence and showing that it is upper-bounded by the mean squared error (MSE). Based on this analysis, a physical layer framework is developed for resilient SFL with LLMs (R-SFLLM) over wireless networks. R-SFLLM leverages wireless sensing data to gather information on the jamming directions-of-arrival (DoAs) for the purpose of devising a novel, sensing-assisted anti-jamming strategy while jointly optimizing beamforming, user scheduling, and resource allocation. Extensive experiments using both LLMs and VLMs demonstrate R-SFLLM's effectiveness, achieving close-to-baseline performance across various natural language processing (NLP) and computer vision (CV) tasks, datasets, and modalities. The proposed methodology further introduces an adversarial training component, where controlled noise exposure significantly enhances the model's resilience to perturbed parameters during training. The results show that more noise-sensitive models, such as RoBERTa, benefit from this feature, especially when resource allocation is unfair. It is also shown that worst-case jamming in particular translates into worst-case model outcomes, thereby necessitating the need for jamming-resilient SFL protocols.
Authors: Shiyuan Zuo, Xingrun Yan, Rongfei Fan, Li Shen, Puning Zhao, Jie Xu, Han Hu
Abstract: Federated Learning (FL) enables multiple clients to collaboratively train models without sharing raw data, but is vulnerable to Byzantine attacks and data heterogeneity, which can severely degrade performance. Existing Byzantine-robust approaches tackle data heterogeneity, but incur high computational overhead during gradient aggregation, thereby slowing down the training process. To address this issue, we propose a simple yet effective Federated Normalized Gradients Algorithm (Fed-NGA), which performs aggregation by merely computing the weighted mean of the normalized gradients from each client. This approach yields a favorable time complexity of $\mathcal{O}(pM)$, where $p$ is the model dimension and $M$ is the number of clients. We rigorously prove that Fed-NGA is robust to both Byzantine faults and data heterogeneity. For non-convex loss functions, Fed-NGA achieves convergence to a neighborhood of stationary points under general assumptions, and further attains zero optimality gap under some mild conditions, which is an outcome rarely achieved in existing literature. In both cases, the convergence rate is $\mathcal{O}(1/T^{\frac{1}{2} - \delta})$, where $T$ denotes the number of iterations and $\delta \in (0, 1/2)$. Experimental results on benchmark datasets confirm the superior time efficiency and convergence performance of Fed-NGA over existing methods.
Authors: Yan Wu, Esther Wershof, Sebastian M Schmon, Marcel Nassar, B{\l}a\.zej Osi\'nski, Ridvan Eksi, Zichao Yan, Rory Stark, Kun Zhang, Thore Graepel
Abstract: We introduce a comprehensive framework for modeling single cell transcriptomic responses to perturbations, aimed at standardizing benchmarking in this rapidly evolving field. Our approach includes a modular and user-friendly model development and evaluation platform, a collection of diverse perturbational datasets, and a set of metrics designed to fairly compare models and dissect their performance. Through extensive evaluation of both published and baseline models across diverse datasets, we highlight the limitations of widely used models, such as mode collapse. We also demonstrate the importance of rank metrics which complement traditional model fit measures, such as RMSE, for validating model effectiveness. Notably, our results show that while no single model architecture clearly outperforms others, simpler architectures are generally competitive and scale well with larger datasets. Overall, this benchmarking exercise sets new standards for model evaluation, supports robust model development, and furthers the use of these models to simulate genetic and chemical screens for therapeutic discovery.
Authors: Haozhe Ma, Zhengding Luo, Thanh Vinh Vo, Kuankuan Sima, Tze-Yun Leong
Abstract: Reward shaping is effective in addressing the sparse-reward challenge in reinforcement learning (RL) by providing immediate feedback through auxiliary, informative rewards. Based on the reward shaping strategy, we propose a novel multi-task reinforcement learning framework that integrates a centralized reward agent (CRA) and multiple distributed policy agents. The CRA functions as a knowledge pool, aimed at distilling knowledge from various tasks and distributing it to individual policy agents to improve learning efficiency. Specifically, the shaped rewards serve as a straightforward metric for encoding knowledge. This framework not only enhances knowledge sharing across established tasks but also adapts to new tasks by transferring meaningful reward signals. We validate the proposed method on both discrete and continuous domains, including the representative Meta-World benchmark, demonstrating its robustness in multi-task sparse-reward settings and its effective transferability to unseen tasks.
Authors: Wang Luo, Haoran Li, Zicheng Zhang, Congying Han, Chi Zhou, Jiayu Lv, Tiande Guo
Abstract: Model-based offline reinforcement learning trains policies using pre-collected datasets and learned environment models, eliminating the need for direct real-world environment interaction. However, this paradigm is inherently challenged by distribution shift~(DS). Existing methods address this issue by leveraging off-policy mechanisms and estimating model uncertainty, but they often result in inconsistent objectives and lack a unified theoretical foundation. This paper offers a comprehensive analysis that disentangles the problem into two fundamental components: model bias and policy shift. Our theoretical and empirical investigations reveal how these factors distort value estimation and restrict policy optimization. To tackle these challenges, we derive a novel shifts-aware reward through a unified probabilistic inference framework, which modifies the vanilla reward to refine value learning and facilitate policy training. Building on this, we develop a practical implementation that leverages classifier-based techniques to approximate the adjusted reward for effective policy optimization. Empirical results across multiple benchmarks demonstrate that the proposed approach mitigates distribution shift and achieves superior or comparable performance, validating our theoretical insights.
Authors: Geetika, Somya Tyagi, Bapi Chatterjee
Abstract: The classical line search for learning rate (LR) tuning in the stochastic gradient descent (SGD) algorithm can tame the convergence slowdown due to data-sampling noise. In a federated setting, wherein the client heterogeneity introduces a slowdown to the global convergence, line search can be relevantly adapted. In this work, we show that a stochastic variant of line search tames the heterogeneity in federated optimization in addition to that due to client-local gradient noise. To this end, we introduce Federated Stochastic Line Search (FedSLS) algorithm and show that it achieves deterministic rates in expectation. Specifically, FedSLS offers linear convergence for strongly convex objectives even with partial client participation. Recently, the extrapolation of the server's LR has shown promise for improved empirical performance for federated learning. To benefit from extrapolation, we extend FedSLS to Federated Extrapolated Stochastic Line Search (FedExpSLS) and prove its convergence. Our extensive empirical results show that the proposed methods perform at par or better than the popular federated learning algorithms across many convex and non-convex problems.
Authors: Biplov Paneru, Bishwash Paneru
Abstract: Ensuring safe water supplies requires effective water quality monitoring, especially in developing countries like Nepal, where contamination risks are high. This paper introduces various hybrid deep learning models to predict on the CCME dataset with multiple water quality parameters from Canada, China, the UK, the USA, and Ireland, with 2.82 million data records feature-engineered and evaluated using them. Models such as CatBoost, XGBoost, and Extra Trees, along with neural networks combining CNN and LSTM layers, are used to capture temporal and spatial patterns in the data. The model demonstrated notable accuracy improvements, aiding proactive water quality control. CatBoost, XGBoost, and Extra Trees Regressor predicted Water Quality Index (WQI) values with an average RMSE of 1.2 and an R squared score of 0.99. Additionally, classifiers achieved 99% accuracy, cross-validated across models. SHAP analysis showed the importance of indicators like F.R.C. and orthophosphate levels in hybrid architectures' classification decisions. The practical application is demonstrated along with a chatbot application for water quality insights.
Authors: I-Chun Chen, Hsu-Shen Liu, Wei-Fang Sun, Chen-Hao Chao, Yen-Chang Hsu, Chun-Yi Lee
Abstract: Sparse Mixture-of-Experts (SMoE) models represent a significant advancement in large language model (LLM) development through their efficient parameter utilization. These models achieve substantial performance improvements at reduced inference costs. However, the deployment of SMoE models faces constraints from extensive memory requirements of expert components in resource-limited environments. To address these limitations, this paper introduces Hierarchical Clustering for Sparsely activated Mixture of Experts (HC-SMoE), a task-agnostic expert merging framework for parameter reduction without retraining. HC-SMoE introduces a novel hierarchical clustering approach based on expert outputs to ensure merging robustness independent of routing decisions. The proposed output-based clustering method enables effective capture of functional relationships between experts for large-scale architectures. We provide theoretical analysis and comprehensive evaluations across multiple zero-shot language tasks to demonstrate HC-SMoE's effectiveness in state-of-the-art models including Qwen and Mixtral. The experimental results validate HC-SMoE's superior performance and practical applicability for real-world deployments.
Authors: Mohammad Hasan Ahmadilivani, Jaan Raik, Masoud Daneshtalab, Maksim Jenihhin
Abstract: The growing exploitation of Machine Learning (ML) in safety-critical applications necessitates rigorous safety analysis. Hardware reliability assessment is a major concern with respect to measuring the level of safety in ML-based systems. Quantifying the reliability of emerging ML models, including Convolutional Neural Networks (CNNs), is highly complex due to their enormous size in terms of the number of parameters and computations. Conventionally, Fault Injection (FI) is applied to perform a reliability measurement. However, performing FI on modern-day CNNs is prohibitively time-consuming if an acceptable confidence level is to be achieved. To speed up FI for large CNNs, statistical FI (SFI) has been proposed, but its runtimes are still considerably long. In this work, we introduce DeepVigor+, a scalable, fast, and accurate semi-analytical method as an efficient alternative for reliability measurement in CNNs. DeepVigor+ implements a fault propagation analysis model and attempts to acquire Vulnerability Factors (VFs) as reliability metrics in an optimal way. The results indicate that DeepVigor+ obtains VFs for CNN models with an error less than $1\%$, i.e., the objective in SFI, but with $14.9$ up to $26.9$ times fewer simulations than the best-known state-of-the-art SFI. DeepVigor+ enables an accurate reliability analysis for large and deep CNNs within a few minutes, rather than achieving the same results in days or weeks.
Authors: Su Hyeong Lee, Sidharth Sharma, Manzil Zaheer, Tian Li
Abstract: Adaptive optimization is critical in federated learning, where enabling adaptivity on both the server and client sides has proven essential for achieving optimal performance. However, the scalability of such jointly adaptive systems is often hindered by resource limitations in communication and memory. In this paper, we introduce a class of efficient adaptive algorithms, named $FedAda^2$ and its enhanced version $FedAda^2$++, designed specifically for large-scale, cross-device federated environments. $FedAda^2$ optimizes communication efficiency by avoiding the transfer of preconditioners between the server and clients. Additionally, $FedAda^2$++ extends this approach by incorporating memory-efficient adaptive optimizers on the client side, further reducing on-device memory usage. Theoretically, we demonstrate that $FedAda^2$ and $FedAda^2$++ achieve the same convergence rates for general, non-convex objectives as its more resource-intensive counterparts that directly integrate joint adaptivity. Extensive empirical evaluations on image and text datasets demonstrate both the advantages of joint adaptivity and the effectiveness of $FedAda^2$/$FedAda^2$++.
Authors: Tuan Ngo, Abid Hassan, Saad Shafiq, Nenad Medvidovic
Abstract: Deep Neural Networks (DNNs) tend to accrue technical debt and suffer from significant retraining costs when adapting to evolving requirements. Modularizing DNNs offers the promise of improving their reusability. Previous work has proposed techniques to decompose DNN models into modules both during and after training. However, these strategies yield several shortcomings, including significant weight overlaps and accuracy losses across modules, restricted focus on convolutional layers only, and added complexity and training time by introducing auxiliary masks to control modularity. In this work, we propose MODA, an activation-driven modular training approach. MODA promotes inherent modularity within a DNN model by directly regulating the activation outputs of its layers based on three modular objectives: intra-class affinity, inter-class dispersion, and compactness. MODA is evaluated using three well-known DNN models and five datasets with varying sizes. This evaluation indicates that, compared to the existing state-of-the-art, using MODA yields several advantages: (1) MODA accomplishes modularization with 22% less training time; (2) the resultant modules generated by MODA comprise up to 24x fewer weights and 37x less weight overlap while (3) preserving the original model's accuracy without additional fine-tuning; in module replacement scenarios, (4) MODA improves the accuracy of a target class by 12% on average while ensuring minimal impact on the accuracy of other classes.
Authors: Saul Santos, Vlad Niculae, Daniel McNamee, Andr\'e F. T. Martins
Abstract: Associative memory models, such as Hopfield networks and their modern variants, have garnered renewed interest due to advancements in memory capacity and connections with self-attention in transformers. In this work, we introduce a unified framework-Hopfield-Fenchel-Young networks-which generalizes these models to a broader family of energy functions. Our energies are formulated as the difference between two Fenchel-Young losses: one, parameterized by a generalized entropy, defines the Hopfield scoring mechanism, while the other applies a post-transformation to the Hopfield output. By utilizing Tsallis and norm entropies, we derive end-to-end differentiable update rules that enable sparse transformations, uncovering new connections between loss margins, sparsity, and exact retrieval of single memory patterns. We further extend this framework to structured Hopfield networks using the SparseMAP transformation, allowing the retrieval of pattern associations rather than a single pattern. Our framework unifies and extends traditional and modern Hopfield networks and provides an energy minimization perspective for widely used post-transformations like $\ell_2$-normalization and layer normalization-all through suitable choices of Fenchel-Young losses and by using convex analysis as a building block. Finally, we validate our Hopfield-Fenchel-Young networks on diverse memory recall tasks, including free and sequential recall. Experiments on simulated data, image retrieval, multiple instance learning, and text rationalization demonstrate the effectiveness of our approach.
Authors: Mert Gurbuzbalaban, Mohammad Rafiqul Islam, Xiaoyu Wang, Lingjiong Zhu
Abstract: Langevin algorithms are popular Markov Chain Monte Carlo methods for Bayesian learning, particularly when the aim is to sample from the posterior distribution of a parametric model, given the input data and the prior distribution over the model parameters. Their stochastic versions such as stochastic gradient Langevin dynamics (SGLD) allow iterative learning based on randomly sampled mini-batches of large datasets and are scalable to large datasets. However, when data is decentralized across a network of agents subject to communication and privacy constraints, standard SGLD algorithms cannot be applied. Instead, we employ decentralized SGLD (DE-SGLD) algorithms, where Bayesian learning is performed collaboratively by a network of agents without sharing individual data. Nonetheless, existing DE-SGLD algorithms induce a bias at every agent that can negatively impact performance; this bias persists even when using full batches and is attributable to network effects. Motivated by the EXTRA algorithm and its generalizations for decentralized optimization, we propose the generalized EXTRA stochastic gradient Langevin dynamics, which eliminates this bias in the full-batch setting. Moreover, we show that, in the mini-batch setting, our algorithm provides performance bounds that significantly improve upon those of standard DE-SGLD algorithms in the literature. Our numerical results also demonstrate the efficiency of the proposed approach.
Authors: Benjamin S. Ruben, William L. Tong, Hamza Tahir Chaudhry, Cengiz Pehlevan
Abstract: Given a fixed budget for total model size, one must choose between training a single large model or combining the predictions of multiple smaller models. We investigate this trade-off for ensembles of random-feature ridge regression models in both the overparameterized and underparameterized regimes. Using deterministic equivalent risk estimates, we prove that when a fixed number of parameters is distributed among $K$ independently trained models, the ridge-optimized test risk increases with $K$. Consequently, a single large model achieves optimal performance. We then ask when ensembles can achieve \textit{near}-optimal performance. In the overparameterized regime, we show that, to leading order, the test error depends on ensemble size and model size only through the total feature count, so that overparameterized ensembles consistently achieve near-optimal performance. To understand underparameterized ensembles, we derive scaling laws for the test risk as a function of total parameter count when the ensemble size and parameters per ensemble member are jointly scaled according to a ``growth exponent'' $\ell$. While the optimal error scaling is always achieved by increasing model size with a fixed ensemble size, our analysis identifies conditions on the kernel and task eigenstructure under which near-optimal scaling laws can be obtained by joint scaling of ensemble size and model size.
Authors: Tobias Lorenz, Marta Kwiatkowska, Mario Fritz
Abstract: Data errors, corruptions, and poisoning attacks during training pose a major threat to the reliability of modern AI systems. While extensive effort has gone into empirical mitigations, the evolving nature of attacks and the complexity of data require a more principled, provable approach to robustly learn on such data - and to understand how perturbations influence the final model. Hence, we introduce MIBP-Cert, a novel certification method based on mixed-integer bilinear programming (MIBP) that computes sound, deterministic bounds to provide provable robustness even under complex threat models. By computing the set of parameters reachable through perturbed or manipulated data, we can predict all possible outcomes and guarantee robustness. To make solving this optimization problem tractable, we propose a novel relaxation scheme that bounds each training step without sacrificing soundness. We demonstrate the applicability of our approach to continuous and discrete data, as well as different threat models - including complex ones that were previously out of reach.
Authors: Yongchao Ye, Xinting Zhu, Xuejin Shen, Xiaoyu Chen, S. Joe Qin, Lishuai Li
Abstract: As Earth's orbital satellite population grows exponentially, effective space situational awareness becomes critical for collision prevention and sustainable operations. Current approaches to monitor satellite behaviors rely on expert knowledge and rule-based systems that scale poorly. Among essential monitoring tasks, satellite pattern-of-life (PoL) identification, analyzing behaviors like station-keeping maneuvers and drift operations, remains underdeveloped due to aerospace system complexity, operational variability, and inconsistent ephemerides sources. We propose a novel generative approach for satellite PoL identification that significantly eliminates the dependence on expert knowledge. The proposed approach leverages orbital elements and positional data to enable automatic pattern discovery directly from observations. Our implementation uses a diffusion model framework for end-to-end identification without manual refinement or domain expertise. The architecture combines a multivariate time-series encoder to capture hidden representations of satellite positional data with a conditional denoising process to generate accurate PoL classifications. Through experiments across diverse real-world satellite operational scenarios, our approach demonstrates superior identification quality and robustness across varying data quality characteristics. A case study using actual satellite data confirms the approach's transformative potential for operational behavior pattern identification, enhanced tracking, and space situational awareness.
Authors: Zaikang Lin, Sei Chang, Aaron Zweig, Minseo Kang, Elham Azizi, David A. Knowles
Abstract: Modern high-throughput biological datasets with thousands of perturbations provide the opportunity for large-scale discovery of causal graphs that represent the regulatory interactions between genes. Differentiable causal graphical models have been proposed to infer a gene regulatory network (GRN) from large scale interventional datasets, capturing the causal gene regulatory relationships from genetic perturbations. However, existing models are limited in their expressivity and scalability while failing to address the dynamic nature of biological processes such as cellular differentiation. We propose PerturbODE, a novel framework that incorporates biologically informative neural ordinary differential equations (neural ODEs) to model cell state trajectories under perturbations and derive the causal GRN from the neural ODE's parameters. We demonstrate PerturbODE's efficacy in trajectory prediction and GRN inference across simulated and real over-expression datasets.
Authors: Somrita Ghosh, Yuelin Xu, Xiao Zhang
Abstract: Achieving high model robustness under adversarial settings is widely recognized as demanding considerable training samples. Recent works propose semi-supervised adversarial training (SSAT) methods with external unlabeled or synthetically generated data, which are the current state-of-the-art. However, SSAT requires substantial extra data to attain high robustness, resulting in prolonged training time and increased memory usage. In this paper, we propose unlabeled data reduction strategies to improve the efficiency of SSAT. Specifically, we design novel latent clustering-based techniques to select or generate a small critical subset of data samples near the model's decision boundary. While focusing on boundary-adjacent points, our methods maintain a balanced ratio between boundary and non-boundary data points to avoid overfitting. Comprehensive experiments on benchmark datasets demonstrate that our methods can significantly reduce SSAT's data requirement and computation costs while preserving its strong robustness advantages. In particular, our latent-space selection scheme based on k-means clustering and our guided DDPM fine-tuning approach with LCG-KM are the most effective, achieving nearly identical robust accuracies with 5x to 10x less unlabeled data and approximately 4x less total runtime.
Authors: Yunyang Li, Lin Huang, Zhihao Ding, Chu Wang, Xinran Wei, Han Yang, Zun Wang, Chang Liu, Yu Shi, Peiran Jin, Tao Qin, Mark Gerstein, Jia Zhang
Abstract: Equivariant Graph Neural Networks (EGNNs) have demonstrated significant success in modeling microscale systems, including those in chemistry, biology and materials science. However, EGNNs face substantial computational challenges due to the high cost of constructing edge features via spherical tensor products, making them impractical for large-scale systems. To address this limitation, we introduce E2Former, an equivariant and efficient transformer architecture that incorporates the Wigner $6j$ convolution (Wigner $6j$ Conv). By shifting the computational burden from edges to nodes, the Wigner $6j$ Conv reduces the complexity from $O(|\mathcal{E}|)$ to $ O(| \mathcal{V}|)$ while preserving both the model's expressive power and rotational equivariance. We show that this approach achieves a 7x-30x speedup compared to conventional $\mathrm{SO}(3)$ convolutions. Furthermore, our empirical results demonstrate that the derived E2Former mitigates the computational challenges of existing approaches without compromising the ability to capture detailed geometric information. This development could suggest a promising direction for scalable and efficient molecular modeling.
Authors: Natalie Maus, Kyurae Kim, Yimeng Zeng, Haydn Thomas Jones, Fangping Wan, Marcelo Der Torossian Torres, Cesar de la Fuente-Nunez, Jacob R. Gardner
Abstract: In multi-objective black-box optimization, the goal is typically to find solutions that optimize a set of $T$ black-box objective functions, $f_1, \ldots f_T$, simultaneously. Traditional approaches often seek a single Pareto-optimal set that balances trade-offs among all objectives. In contrast, we consider a problem setting that departs from this paradigm: finding a small set of $K < T$ solutions, that collectively "cover" the $T$ objectives. A set of solutions is defined as "covering" if, for each objective $f_1, \ldots f_T$, there is at least one good solution. A motivating example for this problem setting occurs in drug design. For example, we may have $T$ pathogens and aim to identify a set of $K < T$ antibiotics such that at least one antibiotic can be used to treat each pathogen. This problem, known as coverage optimization, has yet to be tackled with the Bayesian optimization (BO) framework. To fill this void, we develop Multi-Objective Coverage Bayesian Optimization (MOCOBO), a BO algorithm for solving coverage optimization. Our approach is based on a new acquisition function reminiscent of expected improvement in the vanilla BO setup. We demonstrate the performance of our method on high-dimensional black-box optimization tasks, including applications in peptide and molecular design. Results show that the coverage of the $K < T$ solutions found by MOCOBO matches or nearly matches the coverage of $T$ solutions obtained by optimizing each objective individually. Furthermore, in in vitro experiments, the peptides found by MOCOBO exhibited high potency against drug-resistant pathogens, further demonstrating the potential of MOCOBO for drug discovery. All of our code is publicly available at the following link: https://github.com/nataliemaus/mocobo.
Authors: Shengyu Feng, Yiming Yang
Abstract: This work proposes a simple yet effective sampling framework for combinatorial optimization (CO). Our method builds on discrete Langevin dynamics (LD), an efficient gradient-guided generative paradigm. However, we observe that directly applying LD often leads to limited exploration. To overcome this limitation, we propose the Regularized Langevin Dynamics (RLD), which enforces an expected distance between the sampled and current solutions, effectively avoiding local minima. We develop two CO solvers on top of RLD, one based on simulated annealing (SA), and the other one based on neural network (NN). Empirical results on three classic CO problems demonstrate that both of our methods can achieve comparable or better performance against the previous state-of-the-art (SOTA) SA- and NN-based solvers. In particular, our SA algorithm reduces the runtime of the previous SOTA SA method by up to 80\%, while achieving equal or superior performance. In summary, RLD offers a promising framework for enhancing both traditional heuristics and NN models to solve CO problems. Our code is available at https://github.com/Shengyu-Feng/RLD4CO.
Authors: Dongwon Jo, Jiwon Song, Yulhwa Kim, Jae-Joon Kim
Abstract: While large language models (LLMs) excel at handling long-context sequences, they require substantial prefill computation and key-value (KV) cache, which can heavily burden computational efficiency and memory usage in both prefill and decoding stages. Recent works that compress KV caches with prefill acceleration reduce this cost but inadvertently tie the prefill compute reduction to the decoding KV budget. This coupling arises from overlooking the layer-dependent variation of critical context, often leading to accuracy degradation. To address this issue, we introduce FastKV, a KV cache compression framework designed to reduce latency in both prefill and decoding by leveraging the stabilization of token importance in later layers. FastKV performs full-context computation until a Token-Selective Propagation (TSP) layer, which forwards only the most informative tokens to subsequent layers. From these propagated tokens, FastKV independently selects salient KV entries for caching, thereby decoupling KV budget from the prefill compute reduction based on the TSP decision. This independent control of the TSP rate and KV retention rate enables flexible optimization of efficiency and accuracy. Experimental results show that FastKV achieves speedups of up to 1.82$\times$ in prefill and 2.87$\times$ in decoding compared to the full-context baseline, while matching the accuracy of the baselines that only accelerate the decoding stage. Our code is available at https://github.com/dongwonjo/FastKV.
Authors: Sahar Rajabi, Nayeema Nonta, Sirisha Rambhatla
Abstract: Training large language models (LLMs) is highly resource-intensive due to their massive number of parameters and the overhead of optimizer states. While recent work has aimed to reduce memory consumption, such efforts often entail trade-offs among memory efficiency, training time, and model performance. Yet, true democratization of LLMs requires simultaneous progress across all three dimensions. To this end, we propose SubTrack++ that leverages Grassmannian gradient subspace tracking combined with projection-aware optimizers, enabling Adam's internal statistics to adapt to subspace changes. Additionally, employing recovery scaling, a technique that restores information lost through low-rank projections, further enhances model performance. Our method demonstrates SOTA convergence by exploiting Grassmannian geometry, reducing pre-training wall-time by up to 65% and fine-tuning time by 36% compared to existing SOTA methods, while maintaining the same memory footprint.
Authors: Marco Bertolini, Tuan Le, Djork-Arn\'e Clevert
Abstract: We introduce a novel class of score-based diffusion processes that operate directly in the representation space of Lie groups. Leveraging the framework of Generalized Score Matching, we derive a class of Langevin dynamics that decomposes as a direct sum of Lie algebra representations, enabling the modeling of any target distribution on any (non-Abelian) Lie group. Standard score-matching emerges as a special case of our framework when the Lie group is the translation group. We prove that our generalized generative processes arise as solutions to a new class of paired stochastic differential equations (SDEs), introduced here for the first time. We validate our approach through experiments on diverse data types, demonstrating its effectiveness in real-world applications such as SO(3)-guided molecular conformer generation and modeling ligand-specific global SE(3) transformations for molecular docking, showing improvement in comparison to Riemannian diffusion on the group itself. We show that an appropriate choice of Lie group enhances learning efficiency by reducing the effective dimensionality of the trajectory space and enables the modeling of transitions between complex data distributions.
Authors: Chaofan Lin, Jiaming Tang, Shuo Yang, Hanshuo Wang, Tian Tang, Boyu Tian, Ion Stoica, Mingyu Gao
Abstract: Leveraging attention sparsity to accelerate long-context large language models (LLMs) has been a hot research topic. However, current algorithms such as sparse attention or key-value (KV) cache compression tend to use a fixed budget, which presents a significant challenge during deployment because it fails to account for the dynamic nature of real-world scenarios, where the optimal balance between accuracy and efficiency can vary greatly. In this paper, we find that borrowing top-$p$ sampling (nucleus sampling) to sparse attention can surprisingly achieve adaptive budgeting. Based on this, we propose Twilight, a framework to bring adaptive sparsity to any existing sparse attention algorithm without sacrificing their accuracy. Empirical results show that Twilight can adaptively prune at most 98% of redundant tokens, leading to $15.4\times$ acceleration in self-attention operations and $3.9\times$ acceleration in end-to-end per token latency in long context LLM decoding.
Authors: Reza Shirkavand, Peiran Yu, Qi He, Heng Huang
Abstract: Fine-tuning pre-trained Large Language Models (LLMs) for downstream tasks using First-Order (FO) optimizers presents significant computational challenges. Parameter-Efficient Fine-Tuning (PEFT) methods address these by freezing most model parameters and training only a small subset. However, PEFT often underperforms compared to full fine-tuning when high task-specific accuracy is required. Zeroth-Order (ZO) methods fine-tune the entire pre-trained model without back-propagation, estimating gradients through forward passes only. While memory-efficient, ZO methods suffer from slow convergence and high sensitivity to prompt selection. We bridge these two worlds with Bilevel-ZOFO, a bilevel optimization method that couples fast, local FO-PEFT adaptation at the inner level with stable, memory-efficient ZO updates of the full backbone at the outer level. The FO-PEFT inner loop performs fast, low-memory local adaptation that reduces the variance of ZO estimates and stabilizes the search, guiding the outer ZO updates of the full backbone and reducing prompt sensitivity. In the mean time, the outer ZO provides better generalization ability for PEFT. We provide theoretical convergence guarantees and empirically demonstrate that Bilevel-ZOFO significantly outperforms existing ZO and FO-PEFT methods, achieving 2-4 times faster training while maintaining similar memory efficiency. Additionally, we show by updating the backbone with ZO and adapting only a tiny FO-PEFT block per task, Bilevel-ZOFO combines full-model capacity with few-shot efficiency, making it a very efficient meta-learning algorithm that quickly adapts to new tasks.
Authors: Piersilvio De Bartolomeis, Javier Abad, Guanbo Wang, Konstantin Donhauser, Raymond M. Duch, Fanny Yang, Issa J. Dahabreh
Abstract: Randomized experiments are the preferred approach for evaluating the effects of interventions, but they are costly and often yield estimates with substantial uncertainty. On the other hand, in silico experiments leveraging foundation models offer a cost-effective alternative that can potentially attain higher statistical precision. However, the benefits of in silico experiments come with a significant risk: statistical inferences are not valid if the models fail to accurately predict experimental responses to interventions. In this paper, we propose a novel approach that integrates the predictions from multiple foundation models with experimental data while preserving valid statistical inference. Our estimator is consistent and asymptotically normal, with asymptotic variance no larger than the standard estimator based on experimental data alone. Importantly, these statistical properties hold even when model predictions are arbitrarily biased. Empirical results across several randomized experiments show that our estimator offers substantial precision gains, equivalent to a reduction of up to 20% in the sample size needed to match the same precision as the standard estimator based on experimental data alone.
Authors: Luca Della Libera, Francesco Paissan, Cem Subakan, Mirco Ravanelli
Abstract: Large language models have revolutionized natural language processing through self-supervised pretraining on massive datasets. Inspired by this success, researchers have explored adapting these methods to speech by discretizing continuous audio into tokens using neural audio codecs. However, existing approaches face limitations, including high bitrates, the loss of either semantic or acoustic information, and the reliance on multi-codebook designs when trying to capture both, which increases architectural complexity for downstream tasks. To address these challenges, we introduce FocalCodec, an efficient low-bitrate codec based on focal modulation that utilizes a single binary codebook to compress speech between 0.16 and 0.65 kbps. FocalCodec delivers competitive performance in speech resynthesis and voice conversion at lower bitrates than the current state-of-the-art, while effectively handling multilingual speech and noisy environments. Evaluation on downstream tasks shows that FocalCodec successfully preserves sufficient semantic and acoustic information, while also being well-suited for generative modeling. Demo samples and code are available at https://lucadellalib.github.io/focalcodec-web/.
Authors: Ziheng Cheng, Tianyu Xie, Shiyue Zhang, Cheng Zhang
Abstract: While conditional diffusion models have achieved remarkable success in various applications, they require abundant data to train from scratch, which is often infeasible in practice. To address this issue, transfer learning has emerged as an essential paradigm in small data regimes. Despite its empirical success, the theoretical underpinnings of transfer learning conditional diffusion models remain unexplored. In this paper, we take the first step towards understanding the sample efficiency of transfer learning conditional diffusion models through the lens of representation learning. Inspired by practical training procedures, we assume that there exists a low-dimensional representation of conditions shared across all tasks. Our analysis shows that with a well-learned representation from source tasks, the samplecomplexity of target tasks can be reduced substantially. In addition, we investigate the practical implications of our theoretical results in several real-world applications of conditional diffusion models. Numerical experiments are also conducted to verify our results.
Authors: Taejong Joo, Diego Klabjan
Abstract: Transformers have demonstrated remarkable in-context learning (ICL) capabilities, adapting to new tasks by simply conditioning on demonstrations without parameter updates. Compelling empirical and theoretical evidence suggests that ICL, as a general-purpose learner, could outperform task-specific models. However, it remains unclear to what extent the transformers optimally learn in-context compared to principled learning algorithms. To investigate this, we employ a meta ICL framework in which each prompt defines a distinctive regression task whose target function is drawn from a hierarchical distribution, requiring inference over both the latent model class and task-specific parameters. Within this setup, we benchmark sample complexity of ICL against principled learning algorithms, including the Bayes optimal estimator, under varying performance requirements. Our findings reveal a striking dichotomy: while ICL initially matches the efficiency of a Bayes optimal estimator, its efficiency significantly deteriorates in long context. Through an information-theoretic analysis, we show that the diminishing efficiency is inherent to ICL. These results clarify the trade-offs in adopting ICL as a universal problem solver, motivating a new generation of on-the-fly adaptive methods without the diminishing efficiency.
Authors: Haochen Zhang, Junze Yin, Guanchu Wang, Zirui Liu, Lin F. Yang, Tianyi Zhang, Anshumali Shrivastava, Vladimir Braverman
Abstract: Low-rank optimization has emerged as a promising approach to enabling memory-efficient training of large language models (LLMs). Existing low-rank optimization methods typically project gradients onto a low-rank subspace, reducing the memory cost of storing optimizer states. A key challenge in these methods is selecting suitable subspaces to ensure an effective optimization trajectory. Most existing approaches select the dominant subspace to preserve gradient information, as this intuitively provides the best approximation. However, we find that in practice, the dominant subspace stops changing during pretraining, thereby constraining weight updates to similar subspaces. In this paper, we propose importance sampling for low-rank optimization in LLM pretraining with a provable convergence guarantee, which the dominant subspace approach does not have. Empirically, we demonstrate that our method significantly outperforms previous methods in LLM pretraining tasks.
Authors: Zhaoxian Wu, Quan Xiao, Tayfun Gokmen, Omobayode Fagbohungbe, Tianyi Chen
Abstract: As the economic and environmental costs of training and deploying large vision or language models increase dramatically, analog in-memory computing (AIMC) emerges as a promising energy-efficient solution. However, the training perspective, especially its training dynamic, is underexplored. In AIMC hardware, the trainable weights are represented by the conductance of resistive elements and updated using consecutive electrical pulses. While the conductance changes by a constant in response to each pulse, in reality, the change is scaled by asymmetric and non-linear response functions, leading to a non-ideal training dynamic. This paper provides a theoretical foundation for gradient-based training on AIMC hardware with non-ideal response functions. We demonstrate that asymmetric response functions negatively impact Analog SGD by imposing an implicit penalty on the objective. To overcome the issue, we propose Residual Learning algorithm, which provably converges exactly to a critical point by solving a bilevel optimization problem. We demonstrate that the proposed method can be extended to address other hardware imperfections, such as limited response granularity. As we know, it is the first paper to investigate the impact of a class of generic non-ideal response functions. The conclusion is supported by simulations validating our theoretical insights.
Authors: Yiru Jiao, Sander van Cranenburgh, Simeon Calvert, Hans van Lint
Abstract: The effectiveness of neural network models largely relies on learning meaningful latent patterns from data, where self-supervised learning of informative representations can enhance model performance and generalisability. However, self-supervised representation learning for spatially characterised time series, which are ubiquitous in transportation domain, poses unique challenges due to the necessity of maintaining fine-grained spatio-temporal similarities in the latent space. In this study, we introduce two structure-preserving regularisers for the contrastive learning of spatial time series: one regulariser preserves the topology of similarities between instances, and the other preserves the graph geometry of similarities across spatial and temporal dimensions. To balance the contrastive learning objective and the need for structure preservation, we propose a dynamic weighting mechanism that adaptively manages this trade-off and stabilises training. We validate the proposed method through extensive experiments, including multivariate time series classification to demonstrate its general applicability, as well as macroscopic and microscopic traffic prediction to highlight its particular usefulness in encoding traffic interactions. Across all tasks, our method preserves the similarity structures more effectively and improves state-of-the-art task performances. This method can be integrated with an arbitrary neural network model and is particularly beneficial for time series data with spatial or geographical features. Furthermore, our findings suggest that well-preserved similarity structures in the latent space indicate more informative and useful representations. This provides insights to design more effective neural networks for data-driven transportation research. Our code is made openly accessible with all resulting data at https://github.com/yiru-jiao/spclt
Authors: Annie Marsden, Elad Hazan
Abstract: We study the problem of preconditioning in sequential prediction. From the theoretical lens of linear dynamical systems, we show that convolving the input sequence corresponds to applying a polynomial to the hidden transition matrix. Building on this insight, we propose a universal preconditioning method that convolves the input with coefficients from orthogonal polynomials such as Chebyshev or Legendre. We prove that this approach reduces regret for two distinct prediction algorithms and yields the first ever sublinear and hidden-dimension independent regret bounds (up to logarithmic factors) that hold for systems with marginally stable and asymmetric transition matrices. Finally, extensive synthetic and real-world experiments show that this simple preconditioning strategy improves the performance of a diverse range of algorithms, including recurrent neural networks, and generalizes to signals beyond linear dynamical systems.
Authors: Long-Fei Li, Yu-Yang Qian, Peng Zhao, Zhi-Hua Zhou
Abstract: Reinforcement Learning from Human Feedback (RLHF) has shown remarkable success in aligning Large Language Models (LLMs) with human preferences. Traditional RLHF methods rely on a fixed dataset, which often suffers from limited coverage. To this end, online RLHF has emerged as a promising direction, enabling iterative data collection and refinement. Despite its potential, this paradigm faces a key bottleneck: the requirement to continuously integrate new data into the dataset and re-optimize the model from scratch at each iteration, resulting in computational and storage costs that grow linearly with the number of iterations. In this work, we address this challenge by proposing a one-pass reward modeling method that eliminates the need to store historical data and achieves constant-time updates per iteration. Specifically, we first formalize RLHF as a contextual preference bandit and develop a new algorithm based on online mirror descent with a tailored local norm, replacing the standard maximum likelihood estimation for reward modeling. We then apply it to various online RLHF settings, including passive data collection, active data collection, and deployment-time adaptation. We provide theoretical guarantees showing that our method enhances both statistical and computational efficiency. Finally, we design practical algorithms for LLMs and conduct experiments with the Llama-3-8B-Instruct and Qwen2.5-7B-Instruct models on Ultrafeedback and Mixture2 datasets, validating the effectiveness of our approach.
Authors: Liad Erez, Tomer Koren
Abstract: We study the problem of contextual combinatorial semi-bandits, where input contexts are mapped into subsets of size $m$ of a collection of $K$ possible actions. In each round, the learner observes the realized reward of the predicted actions. Motivated by prototypical applications of contextual bandits, we focus on the $s$-sparse regime where we assume that the sum of rewards is bounded by some value $s\ll K$. For example, in recommendation systems the number of products purchased by any customer is significantly smaller than the total number of available products. Our main result is for the $(\epsilon,\delta)$-PAC variant of the problem for which we design an algorithm that returns an $\epsilon$-optimal policy with high probability using a sample complexity of $\tilde{O}((poly(K/m)+sm/\epsilon^2) \log(|\Pi|/\delta))$ where $\Pi$ is the underlying (finite) class and $s$ is the sparsity parameter. This bound improves upon known bounds for combinatorial semi-bandits whenever $s\ll K$, and in the regime where $s=O(1)$, the leading terms in our bound match the corresponding full-information rates, implying that bandit feedback essentially comes at no cost. Our algorithm is also computationally efficient given access to an ERM oracle for $\Pi$. Our framework generalizes the list multiclass classification problem with bandit feedback, which can be seen as a special case with binary reward vectors. In the special case of single-label classification corresponding to $s=m=1$, we prove an $O((K^7+1/\epsilon^2)\log(|H|/\delta))$ sample complexity bound, which improves upon recent results in this scenario. Additionally, we consider the regret minimization setting where data can be generated adversarially, and establish a regret bound of $\tilde O(|\Pi|+\sqrt{smT\log |\Pi|})$, extending the result of Erez et al. (2024) who consider the simpler single label classification setting.
Authors: Syamantak Kumar, Dheeraj Nagaraj, Purnamrita Sarkar
Abstract: Diffusion models generate samples by estimating the score function of the target distribution at various noise levels. The model is trained using samples drawn from the target distribution by progressively adding noise. Previous sample complexity bounds have polynomial dependence on the dimension $d$, apart from a $\log(|\mathcal{H}|)$ term, where $\mathcal{H}$ is the hypothesis class. In this work, we establish the first (nearly) dimension-free sample complexity bounds, modulo the $\log(|\mathcal{H}|)$ dependence, for learning these score functions, achieving a double exponential improvement in the dimension over prior results. A key aspect of our analysis is the use of a single function approximator to jointly estimate scores across noise levels, a practical feature that enables generalization across time steps. We introduce a martingale-based error decomposition and sharp variance bounds, enabling efficient learning from dependent data generated by Markov processes, which may be of independent interest. Building on these insights, we propose Bootstrapped Score Matching (BSM), a variance reduction technique that leverages previously learned scores to improve accuracy at higher noise levels. These results provide insights into the efficiency and effectiveness of diffusion models for generative modeling.
Authors: \'Alvaro Arroyo, Alessio Gravina, Benjamin Gutteridge, Federico Barbero, Claudio Gallicchio, Xiaowen Dong, Michael Bronstein, Pierre Vandergheynst
Abstract: Graph Neural Networks (GNNs) are models that leverage the graph structure to transmit information between nodes, typically through the message-passing operation. While widely successful, this approach is well known to suffer from the over-smoothing and over-squashing phenomena, which result in representational collapse as the number of layers increases and insensitivity to the information contained at distant and poorly connected nodes, respectively. In this paper, we present a unified view of these problems through the lens of vanishing gradients, using ideas from linear control theory for our analysis. We propose an interpretation of GNNs as recurrent models and empirically demonstrate that a simple state-space formulation of a GNN effectively alleviates over-smoothing and over-squashing at no extra trainable parameter cost. Further, we show theoretically and empirically that (i) GNNs are by design prone to extreme gradient vanishing even after a few layers; (ii) Over-smoothing is directly related to the mechanism causing vanishing gradients; (iii) Over-squashing is most easily alleviated by a combination of graph rewiring and vanishing gradient mitigation. We believe our work will help bridge the gap between the recurrent and graph neural network literature and will unlock the design of new deep and performant GNNs.
Authors: Samuele Bortolotti, Emanuele Marconato, Paolo Morettin, Andrea Passerini, Stefano Teso
Abstract: Concept-based Models are neural networks that learn a concept extractor to map inputs to high-level concepts and an inference layer to translate these into predictions. Ensuring these modules produce interpretable concepts and behave reliably in out-of-distribution is crucial, yet the conditions for achieving this remain unclear. We study this problem by establishing a novel connection between Concept-based Models and reasoning shortcuts (RSs), a common issue where models achieve high accuracy by learning low-quality concepts, even when the inference layer is fixed and provided upfront. Specifically, we extend RSs to the more complex setting of Concept-based Models and derive theoretical conditions for identifying both the concepts and the inference layer. Our empirical results highlight the impact of RSs and show that existing methods, even combined with multiple natural mitigation strategies, often fail to meet these conditions in practice.
Authors: Christian Kl\"otergens, Vijaya Krishna Yalavarthi, Tim Dernedde, Lars Schmidt-Thieme
Abstract: Forecasting Irregular Multivariate Time Series (IMTS) has recently emerged as a distinct research field, necessitating specialized models to address its unique challenges. While most forecasting literature assumes regularly spaced observations without missing values, many real-world datasets - particularly in healthcare, climate research, and biomechanics - violate these assumptions. Time Series (TS)-mixer models have achieved remarkable success in regular multivariate time series forecasting. However, they remain unexplored for IMTS due to their requirement for complete and evenly spaced observations. To bridge this gap, we introduce IMTS-Mixer, a novel forecasting architecture designed specifically for IMTS. Our approach retains the core principles of TS mixer models while introducing innovative methods to transform IMTS into fixed-size matrix representations, enabling their seamless integration with mixer modules. We evaluate IMTS-Mixer on a benchmark of four real-world datasets from various domains. Our results demonstrate that IMTS-Mixer establishes a new state-of-the-art in forecasting accuracy while also improving computational efficiency.
Authors: Sangkyu Lee, Janghoon Han, Hosung Song, Stanley Jungkyu Choi, Honglak Lee, Youngjae Yu
Abstract: Direct Preference Optimization (DPO) demonstrates the advantage of aligning a large language model with human preference using only an offline dataset. However, DPO has the limitation that the KL penalty, which prevents excessive deviation from the reference model, is static throughout the training process. Several methods claim to change this static KL penalty of DPO into a dynamic one, but no approach can adaptively assign different KL penalties for each preference pair. In this paper, we propose $\varepsilon$-Direct Preference Optimization ($\varepsilon$-DPO), which allows adaptive control of the KL penalty strength $\beta$ for each preference pair. Specifically, $\varepsilon$-DPO adaptively controls $\beta$ for each preference pair based on the monotonicity of logits as a preference model under the perturbation of $\beta$ during training. This is equivalent to adjusting the KL penalty by checking whether the change in training-time temperature can lead to better preference confidence as preference models by simply reusing the logit of the current policy and the reference policy. Experimental results show that the simple criterion of $\varepsilon$-DPO for KL penalty relaxation significantly improves DPO compared to most existing direct alignment algorithms on general chatbot benchmarks and reveal that this KL penalty control criterion can reflect confusion as a preference model and provide an efficient KL trade-off, highlighting the significance of instance-level adaptive KL penalty control in DPO.
Authors: Yuxuan Yang, Dalin Zhang, Yuxuan Liang, Hua Lu, Gang Chen, Huan Li
Abstract: Time Series Forecasting (TSF) is a crucial task in various domains, yet existing TSF models rely heavily on high-quality data and insufficiently exploit all available data. This paper explores a novel self-supervised approach to re-label time series datasets by inherently constructing candidate datasets. During the optimization of a simple reconstruction network, intermediates are used as pseudo labels in a self-supervised paradigm, improving generalization for any predictor. We introduce the Self-Correction with Adaptive Mask (SCAM), which discards overfitted components and selectively replaces them with pseudo labels generated from reconstructions. Additionally, we incorporate Spectral Norm Regularization (SNR) to further suppress overfitting from a loss landscape perspective. Our experiments on eleven real-world datasets demonstrate that SCAM consistently improves the performance of various backbone models. This work offers a new perspective on constructing datasets and enhancing the generalization of TSF models through self-supervised learning. The code is available at https://github.com/SuDIS-ZJU/SCAM.
Authors: Yushan Jiang, Wenchao Yu, Geon Lee, Dongjin Song, Kijung Shin, Wei Cheng, Yanchi Liu, Haifeng Chen
Abstract: Time series analysis provides essential insights for real-world system dynamics and informs downstream decision-making, yet most existing methods often overlook the rich contextual signals present in auxiliary modalities. To bridge this gap, we introduce TimeXL, a multi-modal prediction framework that integrates a prototype-based time series encoder with three collaborating Large Language Models (LLMs) to deliver more accurate predictions and interpretable explanations. First, a multi-modal prototype-based encoder processes both time series and textual inputs to generate preliminary forecasts alongside case-based rationales. These outputs then feed into a prediction LLM, which refines the forecasts by reasoning over the encoder's predictions and explanations. Next, a reflection LLM compares the predicted values against the ground truth, identifying textual inconsistencies or noise. Guided by this feedback, a refinement LLM iteratively enhances text quality and triggers encoder retraining. This closed-loop workflow-prediction, critique (reflect), and refinement-continuously boosts the framework's performance and interpretability. Empirical evaluations on four real-world datasets demonstrate that TimeXL achieves up to 8.9% improvement in AUC and produces human-centric, multi-modal explanations, highlighting the power of LLM-driven reasoning for time series prediction.
Authors: Shen Dong, Shaochen Xu, Pengfei He, Yige Li, Jiliang Tang, Tianming Liu, Hui Liu, Zhen Xiang
Abstract: Agents powered by large language models (LLMs) have demonstrated strong capabilities in a wide range of complex, real-world applications. However, LLM agents with a compromised memory bank may easily produce harmful outputs when the past records retrieved for demonstration are malicious. In this paper, we propose a novel Memory INJection Attack, MINJA, without assuming that the attacker can directly modify the memory bank of the agent. The attacker injects malicious records into the memory bank by only interacting with the agent via queries and output observations. These malicious records are designed to elicit a sequence of malicious reasoning steps corresponding to a different target query during the agent's execution of the victim user's query. Specifically, we introduce a sequence of bridging steps to link victim queries to the malicious reasoning steps. During the memory injection, we propose an indication prompt that guides the agent to autonomously generate similar bridging steps, with a progressive shortening strategy that gradually removes the indication prompt, such that the malicious record will be easily retrieved when processing later victim queries. Our extensive experiments across diverse agents demonstrate the effectiveness of MINJA in compromising agent memory. With minimal requirements for execution, MINJA enables any user to influence agent memory, highlighting the risk.
Authors: Luke Guerdan, Solon Barocas, Kenneth Holstein, Hanna Wallach, Zhiwei Steven Wu, Alexandra Chouldechova
Abstract: The LLM-as-a-judge paradigm, in which a judge LLM system replaces human raters in rating the outputs of other generative AI (GenAI) systems, plays a critical role in scaling and standardizing GenAI evaluations. To validate such judge systems, evaluators assess human--judge agreement by first collecting multiple human ratings for each item in a validation corpus, then aggregating the ratings into a single, per-item gold label rating. For many items, however, rating criteria may admit multiple valid interpretations, so a human or LLM rater may deem multiple ratings "reasonable" or "correct." We call this condition rating indeterminacy. Problematically, many rating tasks that contain rating indeterminacy rely on forced-choice elicitation, whereby raters are instructed to select only one rating for each item. In this paper, we introduce a framework for validating LLM-as-a-judge systems under rating indeterminacy. We draw theoretical connections between different measures of judge system performance under different human--judge agreement metrics, and different rating elicitation and aggregation schemes. We demonstrate that differences in how humans and LLMs resolve rating indeterminacy when responding to forced-choice rating instructions can heavily bias LLM-as-a-judge validation. Through extensive experiments involving 11 real-world rating tasks and 9 commercial LLMs, we show that standard validation approaches that rely upon forced-choice ratings select judge systems that are highly suboptimal, performing as much as 31% worse than judge systems selected by our approach that uses multi-label "response set" ratings to account for rating indeterminacy. We conclude with concrete recommendations for more principled approaches to LLM-as-a-judge validation.
Authors: Junkang Wu, Kexin Huang, Xue Wang, Jinyang Gao, Bolin Ding, Jiancan Wu, Xiangnan He, Xiang Wang
Abstract: Aligning large language models (LLMs) with human preferences is critical for real-world deployment, yet existing methods like RLHF face computational and stability challenges. While DPO establishes an offline paradigm with single hyperparameter $\beta$, subsequent methods like SimPO reintroduce complexity through dual parameters ($\beta$, $\gamma$). We propose {ReLU-based Preference Optimization (RePO)}, a streamlined algorithm that eliminates $\beta$ via two advances: (1) retaining SimPO's reference-free margins but removing $\beta$ through gradient analysis, and (2) adopting a ReLU-based max-margin loss that naturally filters trivial pairs. Theoretically, RePO is characterized as SimPO's limiting case ($\beta \to \infty$), where the logistic weighting collapses to binary thresholding, forming a convex envelope of the 0-1 loss. Empirical results on AlpacaEval 2 and Arena-Hard show that RePO outperforms DPO and SimPO across multiple base models, requiring only one hyperparameter to tune.
Authors: Arthur da Cunha, Mikael M{\o}ller H{\o}gsgaard, Andrea Paudice, Yuxin Sun
Abstract: Boosting is a key method in statistical learning, allowing for converting weak learners into strong ones. While well studied in the realizable case, the statistical properties of weak-to-strong learning remain less understood in the agnostic setting, where there are no assumptions on the distribution of the labels. In this work, we propose a new agnostic boosting algorithm with substantially improved sample complexity compared to prior works under very general assumptions. Our approach is based on a reduction to the realizable case, followed by a margin-based filtering of high-quality hypotheses. Furthermore, we show a nearly-matching lower bound, settling the sample complexity of agnostic boosting up to logarithmic factors.
Authors: Xiaobo Xia, Xiaofeng Liu, Jiale Liu, Kuai Fang, Lu Lu, Samet Oymak, William S. Currie, Tongliang Liu
Abstract: Water quality is foundational to environmental sustainability, ecosystem resilience, and public health. Deep learning offers transformative potential for large-scale water quality prediction and scientific insights generation. However, their widespread adoption in high-stakes operational decision-making, such as pollution mitigation and equitable resource allocation, is prevented by unresolved trustworthiness challenges, including performance disparity, robustness, uncertainty, interpretability, generalizability, and reproducibility. In this work, we present a multi-dimensional, quantitative evaluation of trustworthiness benchmarking three state-of-the-art deep learning architectures: recurrent (LSTM), operator-learning (DeepONet), and transformer-based (Informer), trained on 37 years of data from 482 U.S. basins to predict 20 water quality variables. Our investigation reveals systematic performance disparities tied to process complexity, data availability, and basin heterogeneity. Management-critical variables remain the least predictable and most uncertain. Robustness tests reveal pronounced sensitivity to outliers and corrupted targets; notably, the architecture with the strongest baseline performance (LSTM) proves most vulnerable under data corruption. Attribution analyses align for simple variables but diverge for nutrients, underscoring the need for multi-method interpretability. Spatial generalization to ungauged basins remains poor across all models. This work serves as a timely call to action for advancing trustworthy data-driven methods for water resources management and provides a pathway to offering critical insights for researchers, decision-makers, and practitioners seeking to leverage artificial intelligence (AI) responsibly in environmental management.
Authors: Longfei Wei, Fang Sheng, Jianfei Zhang
Abstract: Artificial intelligence (AI) has significantly improved medical screening accuracy, particularly in cancer detection and risk assessment. However, traditional classification metrics often fail to account for imbalanced data, varying performance across cohorts, and patient-level inconsistencies, leading to biased evaluations. We propose the Cohort-Attention Evaluation Metrics (CAT) framework to address these challenges. CAT introduces patient-level assessment, entropy-based distribution weighting, and cohort-weighted sensitivity and specificity. Key metrics like CATSensitivity (CATSen), CATSpecificity (CATSpe), and CATMean ensure balanced and fair evaluation across diverse populations. This approach enhances predictive reliability, fairness, and interpretability, providing a robust evaluation method for AI-driven medical screening models.
Authors: Xin Wu, Fei Teng, Xingwang Li, Ji Zhang, Tianrui Li, Qiang Duan
Abstract: Time series frequently manifest distribution shifts, diverse latent features, and non-stationary learning dynamics, particularly in open and evolving environments. These characteristics pose significant challenges for out-of-distribution (OOD) generalization. While substantial progress has been made, a systematic synthesis of advancements remains lacking. To address this gap, we present the first comprehensive review of OOD generalization methodologies for time series, organized to delineate the field's evolutionary trajectory and contemporary research landscape. We organize our analysis across three foundational dimensions: data distribution, representation learning, and OOD evaluation. For each dimension, we present several popular algorithms in detail. Furthermore, we highlight key application scenarios, emphasizing their real-world impact. Finally, we identify persistent challenges and propose future research directions. A detailed summary of the methods reviewed for the generalization of OOD in time series can be accessed at https://tsood-generalization.com.
Authors: Chenyi Huang, Xinrong Li, Xianchao Xiu
Abstract: Although federated learning has gained prominence as a privacy-preserving framework tailored for distributed Internet of Things (IoT) environments, current federated principal component analysis (PCA) methods lack integration of sparsity, a critical feature for robust anomaly detection. To address this limitation, we propose a novel federated structured sparse PCA (FedSSP) approach for anomaly detection in IoT networks. The proposed model uniquely integrates double sparsity regularization: (1) row-wise sparsity governed by $\ell_{2,p}$-norm with $p\in[0,1)$ to eliminate redundant feature dimensions, and (2) element-wise sparsity via $\ell_{q}$-norm with $q\in[0,1)$ to suppress noise-sensitive components. To efficiently solve this non-convex optimization problem in a distributed setting, we devise a proximal alternating minimization (PAM) algorithm with rigorous theoretical proofs establishing its convergence guarantees. Experiments on real datasets validate that incorporating structured sparsity enhances both model interpretability and detection accuracy.
Authors: Peizhi Niu, Yu-Hsiang Wang, Vishal Rana, Chetan Rupakheti, Abhishek Pandey, Olgica Milenkovic
Abstract: We introduce a new graph diffusion model for small molecule generation, DMol, which outperforms the state-of-the-art DiGress model in terms of validity by roughly 1.5% across all benchmarking datasets while reducing the number of diffusion steps by at least 10-fold, and the running time to roughly one half. The performance improvements are a result of a careful change in the objective function and a graph noise scheduling approach which, at each diffusion step, allows one to only change a subset of nodes of varying size in the molecule graph. Another relevant property of the method is that it can be easily combined with junction-tree-like graph representations that arise by compressing a collection of relevant ring structures into supernodes. Unlike classical junction-tree techniques that involve VAEs and require complicated reconstruction steps, compressed DMol directly performs graph diffusion on a graph that compresses only a carefully selected set of frequent carbon rings into supernodes, which results in straightforward sample generation. This compressed DMol method offers additional validity improvements over generic DMol of roughly 2%, increases the novelty of the method, and further improves the running time due to reductions in the graph size.
Authors: Guowen Li, Xintong Liu, Yang Liu, Mengxuan Chen, Shilei Cao, Xuehe Wang, Juepeng Zheng, Jinxiao Zhang, Haoyuan Liang, Lixian Zhang, Jiuke Wang, Meng Jin, Hong Cheng, Haohuan Fu
Abstract: Accurate Subseasonal-to-Seasonal (S2S) forecasting is vital for decision-making in agriculture, energy production, and emergency management. However, it remains a challenging and underexplored problem due to the chaotic nature of the weather system. Recent data-driven studies have shown promising results, but their performance is limited by the inadequate incorporation of climate states and a model tendency to degrade, progressively losing fine-scale details and yielding over-smoothed forecasts. To overcome these limitations, we propose TianQuan-S2S, a global S2S forecasting model that integrates initial weather states with climatological means via incorporating climatology into patch embedding and enhancing variability capture through an uncertainty-augmented Transformer. Extensive experiments on the Earth Reanalysis 5 (ERA5) reanalysis dataset demonstrate that our model yields a significant improvement in both deterministic and ensemble forecasting over the climatology mean, traditional numerical methods, and data-driven models. Ablation studies empirically show the effectiveness of our model designs. Remarkably, our model outperforms skillful numerical ECMWF-S2S and advanced data-driven Fuxi-S2S in key meteorological variables.
Authors: Shubham Kumar, Narendra Ahuja
Abstract: Post-hoc, unsupervised concept-based explanation methods (U-CBEMs) translate a vision model's internal reasoning into human-understandable concepts, leading to interpretable explanations. However, we find that many state-of-the-art (SOTA) U-CBEMs are not faithful: their concepts seem interpretable but fail to reproduce the model's predictions. We argue that this deficiency has gone unnoticed due to fragmented evaluation - each paper proposes its own faithfulness measure, with no measure-over-measure comparison or broad benchmarking. We close this gap by (i) organizing prior metrics in a unified framework, discussing their limitations, and identifying desiderata for a faithfulness measure; (ii) introducing the Surrogate Faithfulness (SURF) measure, which quantifies faithfulness via the predictive loss of a surrogate that maps explanations to the model's outputs; and (iii) delivering the first comprehensive U-CBEM faithfulness benchmark across diverse tasks and architectures. In a controlled setting, SURF outperforms prior faithfulness measures in measure-over-measure comparisons, and applying SURF to SOTA U-CBEMs reveals that many visually appealing U-CBEMs are surprisingly unfaithful. We demonstrate SURF applicability in two downstream settings - (i) faithfulness versus the number of concepts used in the explanation and (ii) U-CBEM robustness to adversarial attacks - underscoring SURF's value as a reliable faithfulness measure. Code to be released.
Authors: Julius Vetter, Manuel Gloeckler, Daniel Gedon, Jakob H. Macke
Abstract: Simulation-based inference (SBI) offers a flexible and general approach to performing Bayesian inference: In SBI, a neural network is trained on synthetic data simulated from a model and used to rapidly infer posterior distributions for observed data. A key goal for SBI is to achieve accurate inference with as few simulations as possible, especially for expensive simulators. In this work, we address this challenge by repurposing recent probabilistic foundation models for tabular data: We show how tabular foundation models -- specifically TabPFN -- can be used as pre-trained autoregressive conditional density estimators for SBI. We propose Neural Posterior Estimation with Prior-data Fitted Networks (NPE-PFN) and show that it is competitive with current SBI approaches in terms of accuracy for both benchmark tasks and two complex scientific inverse problems. Crucially, it often substantially outperforms them in terms of simulation efficiency, sometimes requiring orders of magnitude fewer simulations. NPE-PFN eliminates the need for inference network selection, training, and hyperparameter tuning. We also show that it exhibits superior robustness to model misspecification and can be scaled to simulation budgets that exceed the context size limit of TabPFN. NPE-PFN provides a new direction for SBI, where training-free, general-purpose inference models offer efficient, easy-to-use, and flexible solutions for a wide range of stochastic inverse problems.
Authors: Selim Mecanna, Aurore Loisy, Christophe Eloy
Abstract: Navigating in a fluid flow while being carried by it, using only information accessible from on-board sensors, is a problem commonly faced by small planktonic organisms. It is also directly relevant to autonomous robots deployed in the oceans. In the last ten years, the fluid mechanics community has widely adopted reinforcement learning, often in the form of its simplest implementations, to address this challenge. But it is unclear how good are the strategies learned by these algorithms. In this paper, we perform a quantitative assessment of reinforcement learning methods applied to navigation in partially observable flows. We first introduce a well-posed problem of directional navigation for which a quasi-optimal policy is known analytically. We then report on the poor performance and robustness of commonly used algorithms (Q-Learning, Advantage Actor Critic) in flows regularly encountered in the literature: Taylor-Green vortices, Arnold-Beltrami-Childress flow, and two-dimensional turbulence. We show that they are vastly surpassed by PPO (Proximal Policy Optimization), a more advanced algorithm that has established dominance across a wide range of benchmarks in the reinforcement learning community. In particular, our custom implementation of PPO matches the theoretical quasi-optimal performance in turbulent flow and does so in a robust manner. Reaching this result required the use of several additional techniques, such as vectorized environments and generalized advantage estimation, as well as hyperparameter optimization. This study demonstrates the importance of algorithm selection, implementation details, and fine-tuning for discovering truly smart autonomous navigation strategies in complex flows.
Authors: Julian B\"uchel, Iason Chalas, Giovanni Acampa, An Chen, Omobayode Fagbohungbe, Sidney Tsai, Kaoutar El Maghraoui, Manuel Le Gallo, Abbas Rahimi, Abu Sebastian
Abstract: Analog in-memory computing (AIMC) is a promising compute paradigm to improve speed and power efficiency of neural network inference beyond the limits of conventional von Neumann-based architectures. However, AIMC introduces fundamental challenges such as noisy computations and strict constraints on input and output quantization. Because of these constraints and imprecisions, off-the-shelf LLMs are not able to achieve 4-bit-level performance when deployed on AIMC-based hardware. While researchers previously investigated recovering this accuracy gap on small, mostly vision-based models, a generic method applicable to LLMs pre-trained on trillions of tokens does not yet exist. In this work, we introduce a general and scalable method to robustly adapt LLMs for execution on noisy, low-precision analog hardware. Our approach enables state-of-the-art models $\unicode{x2013}$ including Phi-3-mini-4k-instruct and Llama-3.2-1B-Instruct $\unicode{x2013}$ to retain performance comparable to 4-bit weight, 8-bit activation baselines, despite the presence of analog noise and quantization constraints. Additionally, we show that as a byproduct of our training methodology, analog foundation models can be quantized for inference on low-precision digital hardware. Finally, we show that our models also benefit from test-time compute scaling, showing better scaling behavior than models trained with 4-bit weight and 8-bit static input quantization. Our work bridges the gap between high-capacity LLMs and efficient analog hardware, offering a path toward energy-efficient foundation models. Code is available at https://github.com/IBM/analog-foundation-models.
Authors: David Grasev
Abstract: Gas turbine engines are complex and highly nonlinear dynamical systems. Deriving their physics-based models can be challenging because it requires performance characteristics that are not always available, often leading to many simplifying assumptions. This paper discusses the limitations of conventional experimental methods used to derive component-level and locally linear parameter-varying models, and addresses these issues by employing identification techniques based on data collected from standard engine operation under closed-loop control. The rotor dynamics are estimated using the sparse identification of nonlinear dynamics. Subsequently, the autonomous part of the dynamics is mapped into an optimally constructed Koopman eigenfunction space. This process involves eigenvalue optimization using metaheuristic algorithms and temporal projection, followed by gradient-based eigenfunction identification. The resulting Koopman model is validated against an in-house reference component-level model. A globally optimal nonlinear feedback controller and a Kalman estimator are then designed within the eigenfunction space and compared to traditional and gain-scheduled proportional-integral controllers, as well as a proposed internal model control approach. The eigenmode structure enables targeting individual modes during optimization, leading to improved performance tuning. Results demonstrate that the Koopman-based controller surpasses other benchmark controllers in both reference tracking and disturbance rejection under sea-level and varying flight conditions, due to its global nature.
Authors: Guoji Fu, Wee Sun Lee
Abstract: This paper studies the approximation and generalization abilities of score-based neural network generative models (SGMs) in estimating an unknown distribution $P_0$ from $n$ i.i.d. observations in $d$ dimensions. Assuming merely that $P_0$ is $\alpha$-sub-Gaussian, we prove that for any time step $t \in [t_0, n^{\mathcal{O}(1)}]$, where $t_0 > \mathcal{O}(\alpha^2n^{-2/d}\log n)$, there exists a deep ReLU neural network with width $\leq \mathcal{O}(n^{\frac{3}{d}}\log_2n)$ and depth $\leq \mathcal{O}(\log^2n)$ that can approximate the scores with $\tilde{\mathcal{O}}(n^{-1})$ mean square error and achieve a nearly optimal rate of $\tilde{\mathcal{O}}(n^{-1}t_0^{-d/2})$ for score estimation, as measured by the score matching loss. Our framework is universal and can be used to establish convergence rates for SGMs under milder assumptions than previous work. For example, assuming further that the target density function $p_0$ lies in Sobolev or Besov classes, with an appropriately early stopping strategy, we demonstrate that neural network-based SGMs can attain nearly minimax convergence rates up to logarithmic factors. Our analysis removes several crucial assumptions, such as Lipschitz continuity of the score function or a strictly positive lower bound on the target density.
Authors: Zhenyi Zhang, Zihan Wang, Yuhao Sun, Tiejun Li, Peijie Zhou
Abstract: Modeling the dynamics from sparsely time-resolved snapshot data is crucial for understanding complex cellular processes and behavior. Existing methods leverage optimal transport, Schr\"odinger bridge theory, or their variants to simultaneously infer stochastic, unbalanced dynamics from snapshot data. However, these approaches remain limited in their ability to account for cell-cell interactions. This integration is essential in real-world scenarios since intercellular communications are fundamental life processes and can influence cell state-transition dynamics. To address this challenge, we formulate the Unbalanced Mean-Field Schr\"odinger Bridge (UMFSB) framework to model unbalanced stochastic interaction dynamics from snapshot data. Inspired by this framework, we further propose CytoBridge, a deep learning algorithm designed to approximate the UMFSB problem. By explicitly modeling cellular transitions, proliferation, and interactions through neural networks, CytoBridge offers the flexibility to learn these processes directly from data. The effectiveness of our method has been extensively validated using both synthetic gene regulatory data and real scRNA-seq datasets. Compared to existing methods, CytoBridge identifies growth, transition, and interaction patterns, eliminates false transitions, and reconstructs the developmental landscape with greater accuracy. Code is available at: https://github.com/zhenyiizhang/CytoBridge-NeurIPS.
Authors: Anders Gj{\o}lbye, Stefan Haufe, Lars Kai Hansen
Abstract: Suppressor variables can influence model predictions without being dependent on the target outcome, and they pose a significant challenge for Explainable AI (XAI) methods. These variables may cause false-positive feature attributions, undermining the utility of explanations. Although effective remedies exist for linear models, their extension to non-linear models and instance-based explanations has remained limited. We introduce PatternLocal, a novel XAI technique that addresses this gap. PatternLocal begins with a locally linear surrogate, e.g., LIME, KernelSHAP, or gradient-based methods, and transforms the resulting discriminative model weights into a generative representation, thereby suppressing the influence of suppressor variables while preserving local fidelity. In extensive hyperparameter optimization on the XAI-TRIS benchmark, PatternLocal consistently outperformed other XAI methods and reduced false-positive attributions when explaining non-linear tasks, thereby enabling more reliable and actionable insights. We further evaluate PatternLocal on an EEG motor imagery dataset, demonstrating physiologically plausible explanations.
Authors: Yuhao Sun, Zhenyi Zhang, Zihan Wang, Tiejun Li, Peijie Zhou
Abstract: Recovering the dynamics from a few snapshots of a high-dimensional system is a challenging task in statistical physics and machine learning, with important applications in computational biology. Many algorithms have been developed to tackle this problem, based on frameworks such as optimal transport and the Schr\"odinger bridge. A notable recent framework is Regularized Unbalanced Optimal Transport (RUOT), which integrates both stochastic dynamics and unnormalized distributions. However, since many existing methods do not explicitly enforce optimality conditions, their solutions often struggle to satisfy the principle of least action and meet challenges to converge in a stable and reliable way. To address these issues, we propose Variational RUOT (Var-RUOT), a new framework to solve the RUOT problem. By incorporating the optimal necessary conditions for the RUOT problem into both the parameterization of the search space and the loss function design, Var-RUOT only needs to learn a scalar field to solve the RUOT problem and can search for solutions with lower action. We also examined the challenge of selecting a growth penalty function in the widely used Wasserstein-Fisher-Rao metric and proposed a solution that better aligns with biological priors in Var-RUOT. We validated the effectiveness of Var-RUOT on both simulated data and real single-cell datasets. Compared with existing algorithms, Var-RUOT can find solutions with lower action while exhibiting faster convergence and improved training stability. Our code is available at https://github.com/ZerooVector/VarRUOT.
Authors: Seanie Lee, Sangwoo Park, Dong Bok Lee, Dominik Wagner, Haebin Seong, Tobias Bocklet, Juho Lee, Sung Ju Hwang
Abstract: Low-Rank Adaptation (LoRA), which introduces a product of two trainable low-rank matrices into frozen pre-trained weights, is widely used for efficient fine-tuning of language models in federated learning (FL). However, when combined with differentially private stochastic gradient descent (DP-SGD), LoRA faces substantial noise amplification: DP-SGD perturbs per-sample gradients, and the matrix multiplication of the LoRA update ($BA$) intensifies this effect. Freezing one matrix (e.g., $A$) reduces the noise but restricts model expressiveness, often resulting in suboptimal adaptation. To address this, we propose $\texttt{FedSVD}$, a simple yet effective method that introduces a global reparameterization based on singular value decomposition (SVD). In our approach, each client optimizes only the $B$ matrix and transmits it to the server. The server aggregates the $B$ matrices, computes the product $BA$ using the previous $A$, and refactorizes the result via SVD. This yields a new adaptive $A$ composed of the orthonormal right singular vectors of $BA$, and an updated $B$ containing the remaining SVD components. This reparameterization avoids quadratic noise amplification, while allowing $A$ to better capture the principal directions of the aggregate updates. Moreover, the orthonormal structure of $A$ bounds the gradient norms of $B$ and preserves more signal under DP-SGD, as confirmed by our theoretical analysis. As a result, $\texttt{FedSVD}$ consistently improves stability and performance across a variety of privacy settings and benchmarks, outperforming relevant baselines under both private and non-private regimes.
Authors: Jie Yan, Jing Liu, Zhong-Yuan Zhang
Abstract: Federated clustering (FC) aims to discover global cluster structures across decentralized clients without sharing raw data, making privacy preservation a fundamental requirement. There are two critical challenges: (1) privacy leakage during collaboration, and (2) robustness degradation due to aggregation of proxy information from non-independent and identically distributed (Non-IID) local data, leading to inaccurate or inconsistent global clustering. Existing solutions typically rely on model-specific local proxies, which are sensitive to data heterogeneity and inherit inductive biases from their centralized counterparts, thus limiting robustness and generality. We propose Omni Federated Clustering (OmniFC), a unified and model-agnostic framework. Leveraging Lagrange coded computing, our method enables clients to share only encoded data, allowing exact reconstruction of the global distance matrix--a fundamental representation of sample relationships--without leaking private information, even under client collusion. This construction is naturally resilient to Non-IID data distributions. This approach decouples FC from model-specific proxies, providing a unified extension mechanism applicable to diverse centralized clustering methods. Theoretical analysis confirms both reconstruction fidelity and privacy guarantees, while comprehensive experiments demonstrate OmniFC's superior robustness, effectiveness, and generality across various benchmarks compared to state-of-the-art methods. Code will be released.
Authors: Song-Lin Lv, Rui Zhu, Tong Wei, Yu-Feng Li, Lan-Zhe Guo
Abstract: Semi-supervised learning (SSL) alleviates the cost of data labeling process by exploiting unlabeled data and has achieved promising results. Meanwhile, with the development of large foundation models, exploiting pre-trained models becomes a promising way to address the label scarcity in the downstream tasks, such as various parameter-efficient fine-tuning techniques. This raises a natural yet critical question: When labeled data is limited, should we rely on unlabeled data or pre-trained models? To investigate this issue, we conduct a fair comparison between SSL methods and pre-trained models (e.g., CLIP) on representative image classification tasks under a controlled supervision budget. Experiments reveal that SSL has met its ``Waterloo" in the era of large models, as pre-trained models show both high efficiency and strong performance on widely adopted SSL benchmarks. This underscores the urgent need for SSL researchers to explore new avenues, such as deeper integration between the SSL and pre-trained models. Furthermore, we investigate the potential of Multi-Modal Large Language Models (MLLMs) in image classification tasks. Results show that, despite their massive parameter scales, MLLMs still face significant performance limitations, highlighting that even a seemingly well-studied task remains highly challenging.
Authors: Jialong Wu, Shaofeng Yin, Ningya Feng, Mingsheng Long
Abstract: World models predict state transitions in response to actions and are increasingly developed across diverse modalities. However, standard training objectives such as maximum likelihood estimation (MLE) often misalign with task-specific goals of world models, i.e., transition prediction metrics like accuracy or perceptual quality. In this paper, we present RLVR-World, a unified framework that leverages reinforcement learning with verifiable rewards (RLVR) to directly optimize world models for such metrics. Despite formulating world modeling as autoregressive prediction of tokenized sequences, RLVR-World evaluates metrics of decoded predictions as verifiable rewards. We demonstrate substantial performance gains on both language- and video-based world models across domains, including text games, web navigation, and robot manipulation. Our work indicates that, beyond recent advances in reasoning language models, RLVR offers a promising post-training paradigm for enhancing the utility of generative models more broadly. Code, datasets, models, and video samples are available at the project website: https://thuml.github.io/RLVR-World.
Authors: Gyubin Lee, Truong Nhat Nguyen Bao, Jaesik Yoon, Dongwoo Lee, Minsu Kim, Yoshua Bengio, Sungjin Ahn
Abstract: Diffusion models have demonstrated strong generative capabilities across domains ranging from image synthesis to complex reasoning tasks. However, most inference-time scaling methods rely on fixed denoising schedules, limiting their ability to allocate computation based on instance difficulty or task-specific demands adaptively. We introduce the challenge of adaptive inference-time scaling-dynamically adjusting computational effort during inference-and propose Adaptive Bi-directional Cyclic Diffusion (ABCD), a flexible, search-based inference framework. ABCD refines outputs through bi-directional diffusion cycles while adaptively controlling exploration depth and termination. It comprises three components: Cyclic Diffusion Search, Automatic Exploration-Exploitation Balancing, and Adaptive Thinking Time. Experiments show that ABCD improves performance across diverse tasks while maintaining computational efficiency.
Authors: Nima Hosseini Dashtbayaz, Hesam Salehipour, Adrian Butscher, Nigel Morris
Abstract: Reduced-order modeling (ROM) of time-dependent and parameterized differential equations aims to accelerate the simulation of complex high-dimensional systems by learning a compact latent manifold representation that captures the characteristics of the solution fields and their time-dependent dynamics. Although high-fidelity numerical solvers generate the training datasets, they have thus far been excluded from the training process, causing the learned latent dynamics to drift away from the discretized governing physics. This mismatch often limits generalization and forecasting capabilities. In this work, we propose Physics-informed ROM ($\Phi$-ROM) by incorporating differentiable PDE solvers into the training procedure. Specifically, the latent space dynamics and its dependence on PDE parameters are shaped directly by the governing physics encoded in the solver, ensuring a strong correspondence between the full and reduced systems. Our model outperforms state-of-the-art data-driven ROMs and other physics-informed strategies by accurately generalizing to new dynamics arising from unseen parameters, enabling long-term forecasting beyond the training horizon, maintaining continuity in both time and space, and reducing the data cost. Furthermore, $\Phi$-ROM learns to recover and forecast the solution fields even when trained or evaluated with sparse and irregular observations of the fields, providing a flexible framework for field reconstruction and data assimilation. We demonstrate the framework's robustness across various PDE solvers and highlight its broad applicability by providing an open-source JAX implementation that is readily extensible to other PDE systems and differentiable solvers, available at https://phi-rom.github.io.
Authors: Martin Carrasco, Guillermo Bernardez, Marco Montagna, Nina Miolane, Lev Telyatnikov
Abstract: While Graph Neural Networks (GNNs) have proven highly effective at modeling relational data, pairwise connections cannot fully capture multi-way relationships naturally present in complex real-world systems. In response to this, Topological Deep Learning (TDL) leverages more general combinatorial representations -- such as simplicial or cellular complexes -- to accommodate higher-order interactions. Existing TDL methods often extend GNNs through Higher-Order Message Passing (HOMP), but face critical \emph{scalability challenges} due to \textit{(i)} a combinatorial explosion of message-passing routes, and \textit{(ii)} significant complexity overhead from the propagation mechanism. This work presents HOPSE (Higher-Order Positional and Structural Encoder), an alternative method to solve tasks involving higher-order interactions \emph{without message passing}. Instead, HOPSE breaks \emph{arbitrary higher-order domains} into their neighborhood relationships using a Hasse graph decomposition. This method shows that decoupling the representation learning of neighborhood topology from that of attributes results in lower computational complexity, casting doubt on the need for HOMP. The experiments on molecular graph tasks and topological benchmarks show that HOPSE matches performance on traditional TDL datasets and outperforms HOMP methods on topological tasks, achieving up to $7\times$ speedups over HOMP-based models, opening a new path for scalable TDL.
Authors: Hossein Zakerinia, Christoph H. Lampert
Abstract: We present new fast-rate PAC-Bayesian generalization bounds for multi-task and meta-learning in the unbalanced setting, i.e. when the tasks have training sets of different sizes, as is typically the case in real-world scenarios. Previously, only standard-rate bounds were known for this situation, while fast-rate bounds were limited to the setting where all training sets are of equal size. Our new bounds are numerically computable as well as interpretable, and we demonstrate their flexibility in handling a number of cases where they give stronger guarantees than previous bounds. Besides the bounds themselves, we also make conceptual contributions: we demonstrate that the unbalanced multi-task setting has different statistical properties than the balanced situation, specifically that proofs from the balanced situation do not carry over to the unbalanced setting. Additionally, we shed light on the fact that the unbalanced situation allows two meaningful definitions of multi-task risk, depending on whether all tasks should be considered equally important or if sample-rich tasks should receive more weight than sample-poor ones.
Authors: Haruki Settai, Naoya Takeishi, Takehisa Yairi
Abstract: For continuous systems modeled by dynamical equations such as ODEs and SDEs, Bellman's Principle of Optimality takes the form of the Hamilton-Jacobi-Bellman (HJB) equation, which provides the theoretical target of reinforcement learning (RL). Although recent advances in RL successfully leverage this formulation, the existing methods typically assume the underlying dynamics are known a priori because they need explicit access to the coefficient functions of dynamical equations to update the value function following the HJB equation. We address this inherent limitation of HJB-based RL; we propose a model-free approach still targeting the HJB equation and propose the corresponding temporal difference method. We establish exponential convergence of the idealized continuous-time dynamics and empirically demonstrate its potential advantages over transition-kernel-based formulations. The proposed formulation paves the way toward bridging stochastic control and model-free reinforcement learning.
Authors: Ronen Gradwohl, Eilam Shapira, Moshe Tennenholtz
Abstract: Algorithmic fairness has emerged as a central issue in ML, and it has become standard practice to adjust ML algorithms so that they will satisfy fairness requirements such as Equal Opportunity. In this paper we consider the effects of adopting such fair classifiers on the overall level of ecosystem fairness. Specifically, we introduce the study of fairness with competing firms, and demonstrate the failure of fair classifiers in yielding fair ecosystems. Our results quantify the loss of fairness in systems, under a variety of conditions, based on classifiers' correlation and the level of their data overlap. We show that even if competing classifiers are individually fair, the ecosystem's outcome may be unfair; and that adjusting biased algorithms to improve their individual fairness may lead to an overall decline in ecosystem fairness. In addition to these theoretical results, we also provide supporting experimental evidence. Together, our model and results provide a novel and essential call for action.
Authors: Parsa Moradi, Hanzaleh Akabrinodehi, Mohammad Ali Maddah-Ali
Abstract: In this paper, we investigate the adversarial robustness of nonparametric regression, a fundamental problem in machine learning, under the setting where an adversary can arbitrarily corrupt a subset of the input data. While the robustness of parametric regression has been extensively studied, its nonparametric counterpart remains largely unexplored. We characterize the adversarial robustness in nonparametric regression, assuming the regression function belongs to the second-order Sobolev space (i.e., it is square integrable up to its second derivative). The contribution of this paper is two-fold: (i) we establish a minimax lower bound on the estimation error, revealing a fundamental limit that no estimator can overcome, and (ii) we show that, perhaps surprisingly, the classical smoothing spline estimator, when properly regularized, exhibits robustness against adversarial corruption. These results imply that if $o(n)$ out of $n$ samples are corrupted, the estimation error of the smoothing spline vanishes as $n \to \infty$. On the other hand, when a constant fraction of the data is corrupted, no estimator can guarantee vanishing estimation error, implying the optimality of the smoothing spline in terms of maximum tolerable number of corrupted samples.
Authors: Jaewon Cheon, Pilsung Kang
Abstract: The growing size of large language models has created significant computational inefficiencies. To address this challenge, sparse activation methods selectively deactivates non-essential parameters during inference, reducing computational costs in FFNN layers. While existing methods focus on non-linear gating mechanisms, we hypothesize that the sparsity of the FFNN layer lies globally in the form of a linear combination over its internal down projection matrix. Based on this insight, we propose two methods: M-COUNTDOWN, leveraging indirect coefficients, and D-COUNTDOWN, utilizing direct coefficients of the linear combination. Experimental results demonstrate that D-COUNTDOWN can omit 90% of computations with performance loss as low as 5.5% ideally, while M-COUNTDOWN provides a predictor-free solution with up to 29.4% better performance preservation compared to existing methods. Our specialized kernel implementations effectively realize these theoretical gains into substantial real-world acceleration.
Authors: Morteza Karimzadeh, Zhongying Wang, James L. Crooks
Abstract: Deep learning has shown strong performance in geospatial prediction tasks, but the role of geolocation information in improving accuracy and generalizability remains underexamined. Recent work has introduced location encoders that aim to represent spatial context in a transferable way, yet most evaluations have focused on static mapping tasks. Here, we study the effect of incorporating geolocation into deep learning for a dynamic and spatially heterogeneous application: estimating daily surface-level PM2.5 across the contiguous United States using satellite and ground-based observations. We compare three strategies for representing location: excluding geolocation, using raw latitude and longitude, and using pretrained location encoders. We evaluate each under within-region and out-of-region generalization settings. Results show that raw coordinates can improve performance within regions by supporting spatial interpolation, but can reduce generalizability across regions. In contrast, pretrained location encoders such as GeoCLIP improve both predictive accuracy and geographic transfer. However, we also observe spatial artifacts linked to encoder characteristics, and performance varies across encoder types (e.g., SatCLIP vs. GeoCLIP). This work provides the first systematic evaluation of location encoders in a dynamic environmental estimation context and offers guidance for incorporating geolocation into deep learning models for geospatial prediction.
Authors: Can Yaras, Alec S. Xu, Pierre Abillama, Changwoo Lee, Laura Balzano
Abstract: Transformers have achieved state-of-the-art performance across various tasks, but suffer from a notable quadratic complexity in sequence length due to the attention mechanism. In this work, we propose MonarchAttention -- a novel approach to sub-quadratic attention approximation via Monarch matrices, an expressive class of structured matrices. Based on the variational form of softmax, we describe an efficient optimization-based algorithm to compute an approximate projection of softmax attention onto the class of Monarch matrices with $\Theta(N\sqrt{N} d)$ computational complexity and $\Theta(Nd)$ memory/IO complexity. Unlike previous approaches, MonarchAttention is both (1) transferable, yielding minimal performance loss with no additional training, even when replacing every attention layer of the Transformer, and (2) hardware-efficient, utilizing the highest-throughput tensor core units on modern GPUs. With optimized kernels, MonarchAttention achieves substantial speed-ups in wall-time over FlashAttention-2: $1.4\times$ for shorter sequences $(N=256)$, $4.5\times$ for medium-length sequences $(N=4K)$, and $8.2\times$ for longer sequences $(N=16K)$. We demonstrate the quality of MonarchAttention on diverse tasks and architectures in vision and language problems, showing that it flexibly and accurately approximates softmax attention in a variety of contexts. Our code is available at https://github.com/cjyaras/monarch-attention.
Authors: Jingyuan Liu, Zeyu Zhang, Xuchuang Wang, Xutong Liu, John C. S. Lui, Mohammad Hajiesmaili, Carlee Joe-Wong
Abstract: Contextual multi-armed bandit is a fundamental learning framework for making a sequence of decisions, e.g., advertising recommendations for a sequence of arriving users. Recent works have shown that clustering these users based on the similarity of their learned preferences can accelerate the learning. However, prior work has primarily focused on the online setting, which requires continually collecting user data, ignoring the offline data widely available in many applications. To tackle these limitations, we study the offline clustering of bandits (Off-ClusBand) problem, which studies how to use the offline dataset to learn cluster properties and improve decision-making. The key challenge in Off-ClusBand arises from data insufficiency for users: unlike the online case where we continually learn from online data, in the offline case, we have a fixed, limited dataset to work from and thus must determine whether we have enough data to confidently cluster users together. To address this challenge, we propose two algorithms: Off-C2LUB, which we show analytically and experimentally outperforms existing methods under limited offline user data, and Off-CLUB, which may incur bias when data is sparse but performs well and nearly matches the lower bound when data is sufficient. We experimentally validate these results on both real and synthetic datasets.
Authors: Yeongmin Kim, Heesun Bae, Byeonghu Na, Il-Chul Moon
Abstract: Direct preference optimization (DPO) is widely used as a simple and stable method for aligning large language models (LLMs) with human preferences. This paper investigates a generalized DPO loss that enables a policy model to match the target policy from a likelihood ratio estimation perspective. The ratio of the target policy provides a unique identification of the policy distribution without relying on reward models or partition functions. This allows the generalized loss to retain both simplicity and theoretical guarantees, which prior work such as $f$-PO fails to achieve simultaneously. We propose Bregman preference optimization (BPO), a generalized framework for ratio matching that provides a family of objective functions achieving target policy optimality. BPO subsumes DPO as a special case and offers tractable forms for all instances, allowing implementation with a few lines of code. We further develop scaled Basu's power divergence (SBA), a gradient scaling method that can be used for BPO instances. The BPO framework complements other DPO variants and is applicable to target policies defined by these variants. In experiments, unlike other probabilistic loss extensions such as $f$-DPO or $f$-PO, which exhibit a trade-off between generation fidelity and diversity, instances of BPO improve both win rate and entropy compared with DPO. When applied to Llama-3-8B-Instruct, BPO achieves state-of-the-art performance among Llama-3-8B backbones, with a 55.9\% length-controlled win rate on AlpacaEval2. Project page: https://github.com/aailab-kaist/BPO.
Authors: Swetha Ganesh, Vaneet Aggarwal
Abstract: Actor-Critic methods are widely used for their scalability, yet existing theoretical guarantees for infinite-horizon average-reward Markov Decision Processes (MDPs) often rely on restrictive ergodicity assumptions. We propose NAC-B, a Natural Actor-Critic with Batching, that achieves order-optimal regret of $\tilde{O}(\sqrt{T})$ in infinite-horizon average-reward MDPs under the unichain assumption, which permits both transient states and periodicity. This assumption is among the weakest under which the classic policy gradient theorem remains valid for average-reward settings. NAC-B employs function approximation for both the actor and the critic, enabling scalability to problems with large state and action spaces. The use of batching in our algorithm helps mitigate potential periodicity in the MDP and reduces stochasticity in gradient estimates, and our analysis formalizes these benefits through the introduction of the constants $C_{\text{hit}}$ and $C_{\text{tar}}$, which characterize the rate at which empirical averages over Markovian samples converge to the stationary distribution.
Authors: Alejandro Hern\'andez-Cano, Dhia Garbaya, Imanol Schlag, Martin Jaggi
Abstract: Despite the significant potential of FP8 data formats for large language model (LLM) pre-training, their adoption has been limited due to challenges in maintaining stability at scale. Existing approaches often rely on suboptimal fine-grained FP8 kernels or fall back to higher-precision matrix multiplications (GEMMs) in sensitive components, such as attention projections, compromising potential throughput gains. We introduce a new class of LLM architectures that, for the first time, support FP8 computation for all GEMMs within transformer blocks during both forward and backward passes. This enables unprecedented throughput gains, particularly at scale, while matching the downstream performance of standard BF16 training. Our architecture design reduces large outlier activations, promoting stable long-term FP8 training. In addition, we identify key metrics to monitor low-precision training and predict potential future divergences.
Authors: Mustafa Hajij, Lennart Bastian, Sarah Osentoski, Hardik Kabaria, John L. Davenport, Sheik Dawood, Balaji Cherukuri, Joseph G. Kocheemoolayil, Nastaran Shahmansouri, Adrian Lew, Theodore Papamarkou, Tolga Birdal
Abstract: We introduce copresheaf topological neural networks (CTNNs), a powerful unifying framework that encapsulates a wide spectrum of deep learning architectures, designed to operate on structured data, including images, point clouds, graphs, meshes, and topological manifolds. While deep learning has profoundly impacted domains ranging from digital assistants to autonomous systems, the principled design of neural architectures tailored to specific tasks and data types remains one of the field's most persistent open challenges. CTNNs address this gap by formulating model design in the language of copresheaves, a concept from algebraic topology that generalizes most practical deep learning models in use today. This abstract yet constructive formulation yields a rich design space from which theoretically sound and practically effective solutions can be derived to tackle core challenges in representation learning, such as long-range dependencies, oversmoothing, heterophily, and non-Euclidean domains. Our empirical results on structured data benchmarks demonstrate that CTNNs consistently outperform conventional baselines, particularly in tasks requiring hierarchical or localized sensitivity. These results establish CTNNs as a principled multi-scale foundation for the next generation of deep learning architectures.
Authors: Ziheng Cheng, Yixiao Huang, Hui Xu, Somayeh Sojoudi, Xuandong Zhao, Dawn Song, Song Mei
Abstract: Text-to-Image (T2I) models have achieved remarkable success in generating visual content from text inputs. Although multiple safety alignment strategies have been proposed to prevent harmful outputs, they often lead to overly cautious behavior -- rejecting even benign prompts -- a phenomenon known as $\textit{over-refusal}$ that reduces the practical utility of T2I models. Despite over-refusal having been observed in practice, there is no large-scale benchmark that systematically evaluates this phenomenon for T2I models. In this paper, we present an automatic workflow to construct synthetic evaluation data, resulting in OVERT ($\textbf{OVE}$r-$\textbf{R}$efusal evaluation on $\textbf{T}$ext-to-image models), the first large-scale benchmark for assessing over-refusal behaviors in T2I models. OVERT includes 4,600 seemingly harmful but benign prompts across nine safety-related categories, along with 1,785 genuinely harmful prompts (OVERT-unsafe) to evaluate the safety-utility trade-off. Using OVERT, we evaluate several leading T2I models and find that over-refusal is a widespread issue across various categories (Figure 1), underscoring the need for further research to enhance the safety alignment of T2I models without compromising their functionality. As a preliminary attempt to reduce over-refusal, we explore prompt rewriting; however, we find it often compromises faithfulness to the meaning of the original prompts. Finally, we demonstrate the flexibility of our generation framework in accommodating diverse safety requirements by generating customized evaluation data adapting to user-defined policies.
Authors: Ruijie Zhang, Ziyue Liu, Zhengyang Wang, Zheng Zhang
Abstract: Training foundation models such as ViTs and LLMs requires tremendous computing cost. Low-rank matrix or tensor factorization offers a parameter-efficient alternative, but often downgrades performance due to the restricted parameter space. In this work, we introduce {\textbf{Latent Crossing (LaX)}} -- a simple yet effective plug-and-play module that enhances the capacity of low-rank models by enabling information flow across low-rank subspaces. We extensively validate the benefits of LaX on pre-training tasks with ViT-Base/Large and LLaMA-like models ranging from 60M to 1B parameters. LaX boosts low-rank model performance to match or exceed the full-rank baselines while using 2-3\(\times\) fewer parameters. When equipped with low-rank adapters (i.e., LoRA) for fine-tuning LLaMA-7/13B, LaX consistently improves performance on arithmetic and common sense reasoning tasks with negligible cost.
Authors: Moritz Haas, Sebastian Bordt, Ulrike von Luxburg, Leena Chennuru Vankadara
Abstract: Scaling limits, such as infinite-width limits, serve as promising theoretical tools to study large-scale models. However, it is widely believed that existing infinite-width theory does not faithfully explain the behavior of practical networks, especially those trained in standard parameterization (SP) meaning He initialization with a global learning rate. For instance, existing theory for SP predicts instability at large learning rates and vanishing feature learning at stable ones. In practice, however, optimal learning rates decay slower than theoretically predicted and networks exhibit both stable training and non-trivial feature learning, even at very large widths. Here, we show that this discrepancy is not fully explained by finite-width phenomena. Instead, we find a resolution through a finer-grained analysis of the regime previously considered unstable and therefore uninteresting. In particular, we show that, under cross-entropy (CE) loss, the unstable regime comprises two distinct sub-regimes: a catastrophically unstable regime and a more benign controlled divergence regime, where logits diverge but gradients and activations remain stable. Moreover, under large learning rates at the edge of the controlled divergence regime, there exists a well-defined infinite width limit where features continue to evolve in all the hidden layers. In experiments across optimizers, architectures, and data modalities, we validate that neural networks operate in this controlled divergence regime under CE loss but not under MSE loss. Our empirical evidence suggests that width-scaling considerations are surprisingly useful for predicting empirically maximal stable learning rate exponents which provide useful guidance on optimal learning rate exponents. Finally, our analysis clarifies the effectiveness and limitations of recently proposed layerwise learning rate scaling for standard initialization.
Authors: Michal Kmicikiewicz, Vincent Fortuin, Ewa Szczurek
Abstract: Designing protein sequences of both high fitness and novelty is a challenging task in data-efficient protein engineering. Exploration beyond wild-type neighborhoods often leads to biologically implausible sequences or relies on surrogate models that lose fidelity in novel regions. Here, we propose ProSpero, an active learning framework in which a frozen pre-trained generative model is guided by a surrogate updated from oracle feedback. By integrating fitness-relevant residue selection with biologically-constrained Sequential Monte Carlo sampling, our approach enables exploration beyond wild-type neighborhoods while preserving biological plausibility. We show that our framework remains effective even when the surrogate is misspecified. ProSpero consistently outperforms or matches existing methods across diverse protein engineering tasks, retrieving sequences of both high fitness and novelty.
Authors: Wenju Sun, Qingyong Li, Wen Wang, Yang Liu, Yangli-ao Geng, Boyang Li
Abstract: Multi-task model merging aims to consolidate knowledge from multiple fine-tuned task-specific experts into a unified model while minimizing performance degradation. Existing methods primarily approach this by minimizing differences between task-specific experts and the unified model, either from a parameter-level or a task-loss perspective. However, parameter-level methods exhibit a significant performance gap compared to the upper bound, while task-loss approaches entail costly secondary training procedures. In contrast, we observe that performance degradation closely correlates with feature drift, i.e., differences in feature representations of the same sample caused by model merging. Motivated by this observation, we propose Layer-wise Optimal Task Vector Merging (LOT Merging), a technique that explicitly minimizes feature drift between task-specific experts and the unified model in a layer-by-layer manner. LOT Merging can be formulated as a convex quadratic optimization problem, enabling us to analytically derive closed-form solutions for the parameters of linear and normalization layers. Consequently, LOT Merging achieves efficient model consolidation through basic matrix operations. Extensive experiments across vision and vision-language benchmarks demonstrate that LOT Merging significantly outperforms baseline methods, achieving improvements of up to 4.4% (ViT-B/32) over state-of-the-art approaches. The source code is available at https://github.com/SunWenJu123/model-merging.
Authors: Yehonathan Refael, Guy Smorodinsky, Tom Tirer, Ofir Lindenbaum
Abstract: Low-rank gradient-based optimization methods have significantly improved memory efficiency during the training of large language models (LLMs), enabling operations within constrained hardware without sacrificing performance. However, these methods primarily emphasize memory savings, often overlooking potential acceleration in convergence due to their reliance on standard isotropic steepest descent techniques, which can perform suboptimally in the highly anisotropic landscapes typical of deep networks, particularly LLMs. In this paper, we propose SUMO (Subspace-Aware Moment-Orthogonalization), an optimizer that employs exact singular value decomposition (SVD) for moment orthogonalization within a dynamically adapted low-dimensional subspace, enabling norm-inducing steepest descent optimization steps. By explicitly aligning optimization steps with the spectral characteristics of the loss landscape, SUMO effectively mitigates approximation errors associated with commonly used methods like Newton-Schulz orthogonalization approximation. We theoretically establish an upper bound on these approximation errors, proving their dependence on the condition numbers of moments, conditions we analytically demonstrate are encountered during LLM training. Furthermore, we both theoretically and empirically illustrate that exact orthogonalization via SVD substantially improves convergence rates while reducing overall complexity. Empirical evaluations confirm that SUMO accelerates convergence, enhances stability, improves performance, and reduces memory requirements by up to 20% compared to state-of-the-art methods.
Authors: Mana Sakai, Ryo Karakida, Masaaki Imaizumi
Abstract: In modern theoretical analyses of neural networks, the infinite-width limit is often invoked to justify Gaussian approximations of neuron preactivations (e.g., via neural network Gaussian processes or Tensor Programs). However, these Gaussian-based asymptotic theories have so far been unable to capture the behavior of attention layers, except under special regimes such as infinitely many heads or tailored scaling schemes. In this paper, leveraging the Tensor Programs framework, we rigorously identify the infinite-width limit distribution of variables within a single attention layer under realistic architectural dimensionality and standard $1/\sqrt{n}$-scaling with $n$ dimensionality. We derive the exact form of this limit law without resorting to infinite-head approximations or tailored scalings, demonstrating that it departs fundamentally from Gaussianity. This limiting distribution exhibits non-Gaussianity from a hierarchical structure, being Gaussian conditional on the random similarity scores. Numerical experiments validate our theoretical predictions, confirming the effectiveness of our theory at finite width and accurate description of finite-head attentions. Beyond characterizing a standalone attention layer, our findings lay the groundwork for developing a unified theory of deep Transformer architectures in the infinite-width regime.
Authors: Ning Zhang, Henry Kenlay, Li Zhang, Mihai Cucuringu, Xiaowen Dong
Abstract: Graph convolutional neural networks (GCNNs) have emerged as powerful tools for analyzing graph-structured data, achieving remarkable success across diverse applications. However, the theoretical understanding of the stability of these models, i.e., their sensitivity to small changes in the graph structure, remains in rather limited settings, hampering the development and deployment of robust and trustworthy models in practice. To fill this gap, we study how perturbations in the graph topology affect GCNN outputs and propose a novel formulation for analyzing model stability. Unlike prior studies that focus only on worst-case perturbations, our distribution-aware formulation characterizes output perturbations across a broad range of input data. This way, our framework enables, for the first time, a probabilistic perspective on the interplay between the statistical properties of the node data and perturbations in the graph topology. We conduct extensive experiments to validate our theoretical findings and demonstrate their benefits over existing baselines, in terms of both representation stability and adversarial attacks on downstream tasks. Our results demonstrate the practical significance of the proposed formulation and highlight the importance of incorporating data distribution into stability analysis.
Authors: Taehoon Yoon, Yunhong Min, Kyeongmin Yeo, Minhyuk Sung
Abstract: We introduce $\Psi$-Sampler, an SMC-based framework incorporating pCNL-based initial particle sampling for effective inference-time reward alignment with a score-based generative model. Inference-time reward alignment with score-based generative models has recently gained significant traction, following a broader paradigm shift from pre-training to post-training optimization. At the core of this trend is the application of Sequential Monte Carlo (SMC) to the denoising process. However, existing methods typically initialize particles from the Gaussian prior, which inadequately captures reward-relevant regions and results in reduced sampling efficiency. We demonstrate that initializing from the reward-aware posterior significantly improves alignment performance. To enable posterior sampling in high-dimensional latent spaces, we introduce the preconditioned Crank-Nicolson Langevin (pCNL) algorithm, which combines dimension-robust proposals with gradient-informed dynamics. This approach enables efficient and scalable posterior sampling and consistently improves performance across various reward alignment tasks, including layout-to-image generation, quantity-aware generation, and aesthetic-preference generation, as demonstrated in our experiments. Project Webpage: https://psi-sampler.github.io/
Authors: Niklas Ro{\ss}berg, Marion Neumeier, Sinan Hasirlioglu, Mohamed Essayed Bouzouraa, Michael Botsch
Abstract: The ability to operate safely in increasingly complex traffic scenarios is a fundamental requirement for Automated Driving Systems (ADS). Ensuring the safe release of ADS functions necessitates a precise understanding of the occurring traffic scenarios. To support this objective, this work introduces a pipeline for traffic scenario clustering and the analysis of scenario category completeness. The Clustering Vector Quantized - Variational Autoencoder (CVQ-VAE) is employed for the clustering of highway traffic scenarios and utilized to create various catalogs with differing numbers of traffic scenario categories. Subsequently, the impact of the number of categories on the completeness considerations of the traffic scenario categories is analyzed. The results show an outperforming clustering performance compared to previous work. The trade-off between cluster quality and the amount of required data to maintain completeness is discussed based on the publicly available highD dataset.
Authors: Wenhao Wu, Fuhong Liu, Haoru Li, Zican Hu, Daoyi Dong, Chunlin Chen, Zhi Wang
Abstract: In-context reinforcement learning (ICRL) has emerged as a promising paradigm for adapting RL agents to downstream tasks through prompt conditioning. However, two notable challenges remain in fully harnessing in-context learning within RL domains: the intrinsic multi-modality of the state-action-reward data and the diverse, heterogeneous nature of decision tasks. To tackle these challenges, we propose T2MIR (Token- and Task-wise MoE for In-context RL), an innovative framework that introduces architectural advances of mixture-of-experts (MoE) into transformer-based decision models. T2MIR substitutes the feedforward layer with two parallel layers: a token-wise MoE that captures distinct semantics of input tokens across multiple modalities, and a task-wise MoE that routes diverse tasks to specialized experts for managing a broad task distribution with alleviated gradient conflicts. To enhance task-wise routing, we introduce a contrastive learning method that maximizes the mutual information between the task and its router representation, enabling more precise capture of task-relevant information. The outputs of two MoE components are concatenated and fed into the next layer. Comprehensive experiments show that T2MIR significantly facilitates in-context learning capacity and outperforms various types of baselines. We bring the potential and promise of MoE to ICRL, offering a simple and scalable architectural enhancement to advance ICRL one step closer toward achievements in language and vision communities. Our code is available at https://github.com/NJU-RL/T2MIR.
Authors: Jes Frellsen, Maher M. Kassem, Tone Bengtsen, Lars Olsen, Kresten Lindorff-Larsen, Jesper Ferkinghoff-Borg, Wouter Boomsma
Abstract: Inverse folding models have proven to be highly effective zero-shot predictors of protein stability. Despite this success, the link between the amino acid preferences of an inverse folding model and the free-energy considerations underlying thermodynamic stability remains incompletely understood. A better understanding would be of interest not only from a theoretical perspective, but also potentially provide the basis for stronger zero-shot stability prediction. In this paper, we take steps to clarify the free-energy foundations of inverse folding models. Our derivation reveals the standard practice of likelihood ratios as a simplistic approximation and suggests several paths towards better estimates of the relative stability. We empirically assess these approaches and demonstrate that considerable gains in zero-shot performance can be achieved with fairly simple means.
Authors: Pavel Dolin, Weizhi Li, Gautam Dasarathy, Visar Berisha
Abstract: This position paper argues that post-deployment monitoring in clinical AI is underdeveloped and proposes statistically valid and label-efficient testing frameworks as a principled foundation for ensuring reliability and safety in real-world deployment. A recent review found that only 9% of FDA-registered AI-based healthcare tools include a post-deployment surveillance plan. Existing monitoring approaches are often manual, sporadic, and reactive, making them ill-suited for the dynamic environments in which clinical models operate. We contend that post-deployment monitoring should be grounded in label-efficient and statistically valid testing frameworks, offering a principled alternative to current practices. We use the term "statistically valid" to refer to methods that provide explicit guarantees on error rates (e.g., Type I/II error), enable formal inference under pre-defined assumptions, and support reproducibility--features that align with regulatory requirements. Specifically, we propose that the detection of changes in the data and model performance degradation should be framed as distinct statistical hypothesis testing problems. Grounding monitoring in statistical rigor ensures a reproducible and scientifically sound basis for maintaining the reliability of clinical AI systems. Importantly, it also opens new research directions for the technical community--spanning theory, methods, and tools for statistically principled detection, attribution, and mitigation of post-deployment model failures in real-world settings.
Authors: Pratiksha Thaker, Neil Kale, Zhiwei Steven Wu, Virginia Smith
Abstract: The state-of-the-art for membership inference attacks on machine learning models is a class of attacks based on shadow models that mimic the behavior of the target model on subsets of held-out nonmember data. However, we find that this class of attacks is fundamentally limited because of a key assumption -- that the shadow models can replicate the target model's behavior on the distribution of interest. As a result, we show that attacks relying on shadow models can fail catastrophically on critical AI safety applications where data access is restricted due to legal, ethical, or logistical constraints, so that the shadow models have no reasonable signal on the query examples. Although this problem seems intractable within the shadow model paradigm, we find that quantile regression attacks are a promising approach in this setting, as these models learn features of member examples that can generalize to unseen classes. We demonstrate this both empirically and theoretically, showing that quantile regression attacks achieve up to 11x the TPR of shadow model-based approaches in practice, and providing a theoretical model that outlines the generalization properties required for this approach to succeed. Our work identifies an important failure mode in existing MIAs and provides a cautionary tale for practitioners that aim to directly use existing tools for real-world applications of AI safety.
Authors: Ran Levinstein, Amit Attia, Matan Schliserman, Uri Sherman, Tomer Koren, Daniel Soudry, Itay Evron
Abstract: We study realizable continual linear regression under random task orderings, a common setting for developing continual learning theory. In this setup, the worst-case expected loss after $k$ learning iterations admits a lower bound of $\Omega(1/k)$. However, prior work using an unregularized scheme has only established an upper bound of $O(1/k^{1/4})$, leaving a significant gap. Our paper proves that this gap can be narrowed, or even closed, using two frequently used regularization schemes: (1) explicit isotropic $\ell_2$ regularization, and (2) implicit regularization via finite step budgets. We show that these approaches, which are used in practice to mitigate forgetting, reduce to stochastic gradient descent (SGD) on carefully defined surrogate losses. Through this lens, we identify a fixed regularization strength that yields a near-optimal rate of $O(\log k / k)$. Moreover, formalizing and analyzing a generalized variant of SGD for time-varying functions, we derive an increasing regularization strength schedule that provably achieves an optimal rate of $O(1/k)$. This suggests that schedules that increase the regularization coefficient or decrease the number of steps per task are beneficial, at least in the worst case.
Authors: Jialin Chen, Ziyu Zhao, Gaukhar Nurbek, Aosong Feng, Ali Maatouk, Leandros Tassiulas, Yifeng Gao, Rex Ying
Abstract: The ubiquity of dynamic data in domains such as weather, healthcare, and energy underscores a growing need for effective interpretation and retrieval of time-series data. These data are inherently tied to domain-specific contexts, such as clinical notes or weather narratives, making cross-modal retrieval essential not only for downstream tasks but also for developing robust time-series foundation models by retrieval-augmented generation (RAG). Despite the increasing demand, time-series retrieval remains largely underexplored. Existing methods often lack semantic grounding, struggle to align heterogeneous modalities, and have limited capacity for handling multi-channel signals. To address this gap, we propose TRACE, a generic multimodal retriever that grounds time-series embeddings in aligned textual context. TRACE enables fine-grained channel-level alignment and employs hard negative mining to facilitate semantically meaningful retrieval. It supports flexible cross-modal retrieval modes, including Text-to-Timeseries and Timeseries-to-Text, effectively linking linguistic descriptions with complex temporal patterns. By retrieving semantically relevant pairs, TRACE enriches downstream models with informative context, leading to improved predictive accuracy and interpretability. Beyond a static retrieval engine, TRACE also serves as a powerful standalone encoder, with lightweight task-specific tuning that refines context-aware representations while maintaining strong cross-modal alignment. These representations achieve state-of-the-art performance on downstream forecasting and classification tasks. Extensive experiments across multiple domains highlight its dual utility, as both an effective encoder for downstream applications and a general-purpose retriever to enhance time-series models.
Authors: Jiaqi Zhao, Miao Zhang, Deng Xiang, Ming Li, Weili Guan, Liqiang Nie
Abstract: Existing post-training quantization methods for large language models (LLMs) offer remarkable success. However, the increasingly marginal performance gains suggest that existing quantization strategies are insufficient to support the development of more compressed models. To inspire new directions for future research, this paper introduces the concept of null space into LLMs quantization. We argue that the quantization error can be effectively alleviated by constraining the post-quantization weight perturbation to lie within the null space of input activations. To prove this idea, we propose a plug-and-play null space projection module for existing milestone PTQ baselines named Q2N. Specifically, we first design an efficient and accurate null space projection approximation method tailored to the characteristics of LLMs. Subsequently, we theoretically derive a closed-form solution for an equivalent vector of the obtained projection matrix, which satisfies practical inference condition while avoiding additional memory overhead. Extensive experiments are conducted on various state-of-the-art LLMs (LLaMA3, DeepSeek, Qwen3) and baselines, demonstrating the effectiveness of both our Q2N and the perspective of null space optimization for LLMs quantization. We view this paper the first step to further alleviate the quantization error based on the insights of null space, hoping it inspiring future researchers to design more advanced quantization methods. Codes are available at https://github.com/zjq0455/q2n.
Authors: Juan Carlos Perdomo, Benjamin Recht
Abstract: This tutorial provides a survey of algorithms for Defensive Forecasting, where predictions are derived not by prognostication but by correcting past mistakes. Pioneered by Vovk, Defensive Forecasting frames the goal of prediction as a sequential game, and derives predictions to minimize metrics no matter what outcomes occur. We present an elementary introduction to this general theory and derive simple, near-optimal algorithms for online learning, calibration, prediction with expert advice, and online conformal prediction.
Authors: Kevin Zhang, Yonghan Jung, Divyat Mahajan, Karthikeyan Shanmugam, Shalmali Joshi
Abstract: Identifying and measuring biases associated with sensitive attributes is a crucial consideration in healthcare to prevent treatment disparities. One prominent issue is inaccurate pulse oximeter readings, which tend to overestimate oxygen saturation for dark-skinned patients and misrepresent supplemental oxygen needs. Most existing research has revealed statistical disparities linking device measurement errors to patient outcomes in intensive care units (ICUs) without causal formalization. This study causally investigates how racial discrepancies in oximetry measurements affect invasive ventilation in ICU settings. We employ a causal inference-based approach using path-specific effects to isolate the impact of bias by race on clinical decision-making. To estimate these effects, we leverage a doubly robust estimator, propose its self-normalized variant for improved sample efficiency, and provide novel finite-sample guarantees. Our methodology is validated on semi-synthetic data and applied to two large real-world health datasets: MIMIC-IV and eICU. Contrary to prior work, our analysis reveals minimal impact of racial discrepancies on invasive ventilation rates. However, path-specific effects mediated by oxygen saturation disparity are more pronounced on ventilation duration, and the severity differs by dataset. Our work provides a novel pipeline for investigating potential disparities in clinical decision-making and, more importantly, highlights the necessity of causal methods to robustly assess fairness in healthcare.
Authors: Lowell Weissman, Michael Krumdick, A. Lynn Abbott
Abstract: Recent work on neural scaling laws demonstrates that model performance scales predictably with compute budget, model size, and dataset size. In this work, we develop scaling laws based on problem complexity. We analyze two fundamental complexity measures: solution space size and representation space size. Using the Traveling Salesman Problem (TSP) as a case study, we show that combinatorial optimization promotes smooth cost trends, and therefore meaningful scaling laws can be obtained even in the absence of an interpretable loss. We then show that suboptimality grows predictably for fixed-size models when scaling the number of TSP nodes or spatial dimensions, independent of whether the model was trained with reinforcement learning or supervised fine-tuning on a static dataset. We conclude with an analogy to problem complexity scaling in local search, showing that a much simpler gradient descent of the cost landscape produces similar trends.
Authors: Bruno Mlodozeniec, Isaac Reid, Sam Power, David Krueger, Murat Erdogdu, Richard E. Turner, Roger Grosse
Abstract: Randomness is an unavoidable part of training deep learning models, yet something that traditional training data attribution algorithms fail to rigorously account for. They ignore the fact that, due to stochasticity in the initialisation and batching, training on the same dataset can yield different models. In this paper, we address this shortcoming through introducing distributional training data attribution (d-TDA), the goal of which is to predict how the distribution of model outputs (over training runs) depends upon the dataset. Intriguingly, we find that influence functions (IFs), a popular data attribution tool, are 'secretly distributional': they emerge from our framework as the limit to unrolled differentiation, without requiring restrictive convexity assumptions. This provides a new perspective on the effectiveness of IFs in deep learning. We demonstrate the practical utility of d-TDA in experiments, including improving data pruning for vision transformers and identifying influential examples with diffusion models.
Authors: Shen Yuan, Yin Zheng, Taifeng Wang, Binbin Liu, Hongteng Xu
Abstract: Adapting large-scale foundation models in multi-task scenarios often suffers from task conflict and oblivion. To mitigate such issues, we propose a novel ''model MoE-ization'' strategy that leads to a conflict- and oblivion-resistant multi-task adaptation method. Given a weight matrix of a pre-trained model, our method applies SVD to it and introduces a learnable router to adjust its singular values based on tasks and samples. Accordingly, the weight matrix becomes a Mixture of Orthogonal Rank-one Experts (MoORE), in which each expert corresponds to the outer product of a left singular vector and the corresponding right one. We can improve the model capacity by imposing a learnable orthogonal transform on the right singular vectors. Unlike low-rank adaptation (LoRA) and its MoE-driven variants, MoORE guarantees the experts' orthogonality and maintains the column space of the original weight matrix. These two properties make the adapted model resistant to the conflicts among the new tasks and the oblivion of its original tasks, respectively. Experiments on various datasets demonstrate that MoORE outperforms existing multi-task adaptation methods consistently, showing its superiority in terms of conflict- and oblivion-resistance. The code of the experiments is available at https://github.com/DaShenZi721/MoORE.
Authors: Junqi Gao, Zhichang Guo, Dazhi Zhang, Dong Li, Runze Liu, Pengfei Li, Kai Tian, Biqing Qi
Abstract: Heterogeneous Large Language Model (LLM) fusion integrates the strengths of multiple source LLMs with different architectures into a target LLM with low computational overhead. While promising, existing methods suffer from two major limitations: 1) reliance on real data from limited domain for knowledge fusion, preventing the target LLM from fully acquiring knowledge across diverse domains, and 2) fixed data allocation proportions across domains, failing to dynamically adjust according to the target LLM's varying capabilities across domains, leading to a capability imbalance. To overcome these limitations, we propose Bohdi, a synthetic-data-only heterogeneous LLM fusion framework. Through the organization of knowledge domains into a hierarchical tree structure, Bohdi enables automatic domain exploration and multi-domain data generation through multi-model collaboration, thereby comprehensively extracting knowledge from source LLMs. By formalizing domain expansion and data sampling proportion allocation on the knowledge tree as a Hierarchical Multi-Armed Bandit problem, Bohdi leverages the designed DynaBranches mechanism to adaptively adjust sampling proportions based on the target LLM's performance feedback across domains. Integrated with our proposed Introspection-Rebirth (IR) mechanism, DynaBranches dynamically tracks capability shifts during target LLM's updates via Sliding Window Binomial Likelihood Ratio Testing (SWBLRT), further enhancing its online adaptation capability. Comparative experimental results on a comprehensive suite of benchmarks demonstrate that Bohdi significantly outperforms existing baselines on multiple target LLMs, exhibits higher data efficiency, and virtually eliminates the imbalance in the target LLM's capabilities. Our code is available at https://github.com/gjq100/Bohdi.git.
Authors: Konstantin Usevich, Ricardo Borsoi, Clara D\'erand, Marianne Clausel
Abstract: Polynomial Neural Networks (PNNs) possess a rich algebraic and geometric structure. However, their identifiability -- a key property for ensuring interpretability -- remains poorly understood. In this work, we present a comprehensive analysis of the identifiability of deep PNNs, including architectures with and without bias terms. Our results reveal an intricate interplay between activation degrees and layer widths in achieving identifiability. As special cases, we show that architectures with non-increasing layer widths are generically identifiable under mild conditions, while encoder-decoder networks are identifiable when the decoder widths do not grow too rapidly compared to the activation degrees. Our proofs are constructive and center on a connection between deep PNNs and low-rank tensor decompositions, and Kruskal-type uniqueness theorems. We also settle an open conjecture on the dimension of PNN's neurovarieties, and provide new bounds on the activation degrees required for it to reach the expected dimension.
Authors: Paul C. Bogdan, Uzay Macar, Neel Nanda, Arthur Conmy
Abstract: Current frontier large-language models rely on reasoning to achieve state-of-the-art performance. Many existing interpretability are limited in this area, as standard methods have been designed to study single forward passes of a model rather than the multi-token computational steps that unfold during reasoning. We argue that analyzing reasoning traces at the sentence level is a promising approach to understanding reasoning processes. We introduce a black-box method that measures each sentence's counterfactual importance by repeatedly sampling replacement sentences from the model, filtering for semantically different ones, and continuing the chain of thought from that point onwards to quantify the sentence's impact on the distribution of final answers. We discover that certain sentences can have an outsized impact on the trajectory of the reasoning trace and final answer. We term these sentences \textit{thought anchors}. These are generally planning or uncertainty management sentences, and specialized attention heads consistently attend from subsequent sentences to thought anchors. We further show that examining sentence-sentence causal links within a reasoning trace gives insight into a model's behavior. Such information can be used to predict a problem's difficulty and the extent different question domains involve sequential or diffuse reasoning. As a proof-of-concept, we demonstrate that our techniques together provide a practical toolkit for analyzing reasoning models by conducting a detailed case study of how the model solves a difficult math problem, finding that our techniques yield a consistent picture of the reasoning trace's structure. We provide an open-source tool (thought-anchors.com) for visualizing the outputs of our methods on further problems. The convergence across our methods shows the potential of sentence-level analysis for a deeper understanding of reasoning models.
Authors: Jing Lu, Xuan Wu, Yizhun Tian, Songhan Fan, Yali Fang
Abstract: Flight trajectory prediction (FTP) and similar time series tasks typically require capturing smooth latent dynamics hidden within noisy signals. However, existing deep learning models face significant challenges of high computational cost and insufficient interpretability due to their complex black-box nature. This paper introduces FlightKooba, a novel modeling approach designed to extract such underlying dynamics analytically. Our framework uniquely integrates HiPPO theory, Koopman operator theory, and control theory. By leveraging Legendre polynomial bases, it constructs Koopman operators analytically, thereby avoiding large-scale parameter training. The method's core strengths lie in its exceptional computational efficiency and inherent interpretability. Experiments on multiple public datasets validate our design philosophy: for signals exhibiting strong periodicity or clear physical laws (e.g., in aviation, meteorology, and traffic flow), FlightKooba delivers competitive prediction accuracy while reducing trainable parameters by several orders of magnitude and achieving the fastest training speed. Furthermore, we analyze the model's theoretical boundaries, clarifying its inherent low-pass filtering characteristics that render it unsuitable for sequences dominated by high-frequency noise. In summary, FlightKooba offers a powerful, efficient, and interpretable new alternative for time series analysis, particularly in resource-constrained environments.
Authors: Torben Berndt, Benjamin Walker, Tiexin Qin, Jan St\"uhmer, Andrey Kormilitzin
Abstract: Dynamic graphs exhibit complex temporal dynamics due to the interplay between evolving node features and changing network structures. Recently, Graph Neural Controlled Differential Equations (Graph Neural CDEs) successfully adapted Neural CDEs from paths on Euclidean domains to paths on graph domains. Building on this foundation, we introduce Permutation Equivariant Neural Graph CDEs, which project Graph Neural CDEs onto permutation equivariant function spaces. This significantly reduces the model's parameter count without compromising representational power, resulting in more efficient training and improved generalisation. We empirically demonstrate the advantages of our approach through experiments on simulated dynamical systems and real-world tasks, showing improved performance in both interpolation and extrapolation scenarios.
Authors: Zhiyu Zhao, Haoxuan Li, Haifeng Zhang, Jun Wang, Francesco Faccio, J\"urgen Schmidhuber, Mengyue Yang
Abstract: When building a world model, a common assumption is that the environment has a single, unchanging underlying causal rule, like applying Newton's laws to every situation. In reality, what appears as a drifting causal mechanism is often the manifestation of a fixed underlying mechanism seen through a narrow observational window. This brings about a problem that, when building a world model, even subtle shifts in policy or environment states can alter the very observed causal mechanisms. In this work, we introduce the \textbf{Meta-Causal Graph} as world models, a minimal unified representation that efficiently encodes the transformation rules governing how causal structures shift across different latent world states. A single Meta-Causal Graph is composed of multiple causal subgraphs, each triggered by meta state, which is in the latent state space. Building on this representation, we introduce a \textbf{Causality-Seeking Agent} whose objectives are to (1) identify the meta states that trigger each subgraph, (2) discover the corresponding causal relationships by agent curiosity-driven intervention policy, and (3) iteratively refine the Meta-Causal Graph through ongoing curiosity-driven exploration and agent experiences. Experiments on both synthetic tasks and a challenging robot arm manipulation task demonstrate that our method robustly captures shifts in causal dynamics and generalizes effectively to previously unseen contexts.
Authors: Taeyoun Kim, Fahim Tajwar, Aditi Raghunathan, Aviral Kumar
Abstract: Reasoning methods that adaptively allocate test-time compute have advanced LLM performance on easy to verify domains such as math and code. In this work, we study how to utilize this approach to train models that exhibit a degree of robustness to safety vulnerabilities, and show that doing so can provide benefits. We build a recipe called $\textit{TARS}$ (Training Adaptive Reasoners for Safety), a reinforcement learning (RL) approach that trains models to reason about safety using chain-of-thought traces and a reward signal that balances safety with task completion. To build TARS, we identify three critical design choices: (1) a ``lightweight'' warmstart SFT stage, (2) a mix of harmful, harmless, and ambiguous prompts to prevent shortcut behaviors such as too many refusals, and (3) a reward function to prevent degeneration of reasoning capabilities during training. Models trained with TARS exhibit adaptive behaviors by spending more compute on ambiguous queries, leading to better safety-refusal trade-offs. They also internally learn to better distinguish between safe and unsafe prompts and attain greater robustness to both white-box (e.g., GCG) and black-box attacks (e.g., PAIR). Overall, our work provides an effective, open recipe for training LLMs against jailbreaks and harmful requests by reasoning per prompt.
Authors: Biplov Paneru
Abstract: This research discusses the critical need for early detection and treatment for early prediction of Freezing of Gaits (FOG) utilizing a wearable sensor technology powered with LoRa communication. The system consisted of an Esp-32 microcontroller, in which the trained model is utilized utilizing the Micromlgen Python library. The research investigates accurate FOG classification based on pertinent clinical data by utilizing machine learning (ML) algorithms like Catboost, XGBoost, and Extra Tree classifiers. The XGBoost could classify with approximately 97% accuracy, along with 96% for the catboost and 90% for the Extra Trees Classifier model. The SHAP analysis interpretability shows that GYR SI degree is the most affecting factor in the prediction of the diseases. These results show the possibility of monitoring and identifying the affected person with tracking location on GPS and providing aid as an assistive technology for aiding the affected. The developed sensor-based technology has great potential for real-world problem solving in the field of healthcare and biomedical technology enhancements.
Authors: Yannis Bendi-Ouis (Mnemosyne), Xavier Hinaut (Mnemosyne)
Abstract: While Large Language Models and their underlying Transformer architecture are remarkably efficient, they do not reflect how our brain processes and learns a diversity of cognitive tasks such as language and working memory. Furthermore, sequential data processing with Transformers encounters a fundamental barrier: quadratic complexity growth with sequence length. Motivated by these limitations, our ambition is to create more efficient models that are less reliant on intensive computations. We introduce Echo State Transformers (EST), a hybrid architecture that elegantly resolves this challenge while demonstrating exceptional performance in classification and detection tasks. EST integrates the Transformer attention mechanisms with principles from Reservoir Computing to create a fixed-size window distributed memory system. Drawing inspiration from Echo State Networks, the most prominent instance of the Reservoir Computing paradigm, our approach leverages reservoirs (random recurrent networks) as a lightweight and efficient memory. Our architecture integrates a new module called ''Working Memory'' based on several reservoirs working in parallel. These reservoirs work as independent working memory units with distinct internal dynamics. A novelty here is that the classical reservoir hyperparameters, controlling the dynamics, are now trained. Thus, the EST dynamically adapts the reservoir memory/non-linearity trade-off. Thanks to these working memory units, EST achieves constant computational complexity at each processing step, effectively breaking the quadratic scaling problem of standard Transformers. We evaluate ESTs on a recent challenging timeseries benchmark: the Time Series Library, which comprises 69 tasks across five categories. Results show that ESTs ranks first overall in two of five categories, outperforming strong state-of-the-art baselines on classification and anomaly detection tasks, while remaining competitive on short-term forecasting. These results position ESTs as a compelling alternative for time-series classification and anomaly detection, and a practical complement to transformer-style models in applications that prioritize robust representations and sensitive event detection.
Authors: Patrik Okanovic, Sameer Deshmukh, Grzegorz Kwasniewski, Yi Zhu, Haruto Fujii, Sakina Fatima, Maciej Besta, Kentaro Katayama, Takumi Honda, Yusuke Nagasaka, Torsten Hoefler
Abstract: The energy consumption of large-scale ML models is dominated by data movement, shuffling billions of parameters across memory hierarchies and data centers. Sparsification offers a principled way to mitigate these costs by pruning redundant weights and activations, thereby reducing data movement. Effective sparsification to prune redundant parameters is still challenging: existing methods incur significant accuracy degradation, performance overhead, or both. We introduce (Bl)ock (a)nd (S)parse (T)ransformers (BLaST), a general, robust, and reliable method for sparsification, applicable to linear layers in all settings. Our method iteratively sparsifies weight matrices into a block sparsity pattern suitable for efficient sparse matrix-matrix (SpMM) multiplication. BLaST achieves up to 95% sparsity in MLP weights with negligible accuracy loss (majority <2.25%). We show a 2.2x inference speedup for Llama 3.2 with 16 GPUs, and up to 4.45x reduction in inference memory footprint resulting in a 2.9x reduction in GPU setup and operating costs.
Authors: Alexander Gr\"afe, Sebastian Trimpe
Abstract: Mixed-precision training has emerged as an indispensable tool for enhancing the efficiency of neural network training in recent years. Concurrently, JAX has grown in popularity as a versatile machine learning toolbox. However, it currently lacks robust support for mixed-precision training. We propose MPX, a mixed-precision training toolbox for JAX that simplifies and accelerates the training of large-scale neural networks while preserving model accuracy. MPX seamlessly integrates with popular toolboxes such as Equinox and Flax, allowing users to convert full-precision pipelines to mixed-precision versions with minimal modifications. By casting both inputs and outputs to half precision, and introducing a dynamic loss-scaling mechanism, MPX alleviates issues like gradient underflow and overflow that commonly arise in half precision computations. Its design inherits critical features from JAX's type-promotion behavior, ensuring that operations take place in the correct precision and allowing for selective enforcement of full precision where needed (e.g., sums, means, or softmax). MPX further provides wrappers for automatic creation and management of mixed-precision gradients and optimizers, enabling straightforward integration into existing JAX training pipelines. MPX's source code, documentation, and usage examples are available at github.com/Data-Science-in-Mechanical-Engineering/mixed_precision_for_JAX .
Authors: Bogdan Kulynych, Juan Felipe Gomez, Georgios Kaissis, Jamie Hayes, Borja Balle, Flavio du Pin Calmon, Jean Louis Raisaro
Abstract: Differentially private (DP) mechanisms are difficult to interpret and calibrate because existing methods for mapping standard privacy parameters to concrete privacy risks -- re-identification, attribute inference, and data reconstruction -- are both overly pessimistic and inconsistent. In this work, we use the hypothesis-testing interpretation of DP ($f$-DP), and determine that bounds on attack success can take the same unified form across re-identification, attribute inference, and data reconstruction risks. Our unified bounds are (1) consistent across a multitude of attack settings, and (2) tunable, enabling practitioners to evaluate risk with respect to arbitrary, including worst-case, levels of baseline risk. Empirically, our results are tighter than prior methods using $\varepsilon$-DP, R\'enyi DP, and concentrated DP. As a result, calibrating noise using our bounds can reduce the required noise by 20% at the same risk level, which yields, e.g., an accuracy increase from 52% to 70% in a text classification task. Overall, this unifying perspective provides a principled framework for interpreting and calibrating the degree of protection in DP against specific levels of re-identification, attribute inference, or data reconstruction risk.
Authors: Lily Hong Zhang, Smitha Milli, Karen Jusko, Jonathan Smith, Brandon Amos, Wassim Bouaziz, Manon Revel, Jack Kussman, Yasha Sheynin, Lisa Titus, Bhaktipriya Radharapu, Jane Yu, Vidya Sarma, Kris Rose, Maximilian Nickel
Abstract: How can large language models (LLMs) serve users with varying preferences that may conflict across cultural, political, or other dimensions? To advance this challenge, this paper establishes four key results. First, we demonstrate, through a large-scale multilingual human study with representative samples from five countries (N=15,000), that humans exhibit significantly more variation in preferences than the responses of 21 state-of-the-art LLMs. Second, we show that existing methods for preference dataset collection are insufficient for learning the diversity of human preferences even along two of the most salient dimensions of variability in global values, due to the underlying homogeneity of candidate responses. Third, we argue that this motivates the need for negatively-correlated sampling when generating candidate sets, and we show that simple prompt-based techniques for doing so significantly enhance the performance of alignment methods in learning heterogeneous preferences. Fourth, based on this novel candidate sampling approach, we collect and open-source Community Alignment, the largest and most representative multilingual and multi-turn preference dataset to date, featuring almost 200,000 comparisons from annotators spanning five countries. We hope that the Community Alignment dataset will be a valuable resource for improving the effectiveness of LLMs for a diverse global population.
Authors: Sara Si-Moussi, Stephan Hennekens, Sander Mucher, Stan Los, Yoann Cartier, Borja Jim\'enez-Alfaro, Fabio Attorre, Jens-Christian Svenning, Wilfried Thuiller
Abstract: Habitats integrate the abiotic conditions, vegetation composition and structure that support biodiversity and sustain nature's contributions to people. Most habitats face mounting pressures from human activities, which requires accurate, high-resolution habitat mapping for effective conservation and restoration. Yet, current habitat maps often fall short in thematic or spatial resolution because they must (1) model several mutually exclusive habitat types that co-occur across landscapes and (2) cope with severe class imbalance that complicates exhaustive multi-class training. Here, we evaluated how high-resolution remote sensing (RS) data and Artificial Intelligence (AI) tools can improve habitat mapping across large geographical extents at fine spatial and thematic resolution. Using vegetation plots from the European Vegetation Archive, we modelled the distribution of Level 3 EUNIS habitat types across Europe and assessed multiple modelling strategies against independent validation datasets. Strategies that exploited the hierarchical nature of habitat classifications resolved classification ambiguities, especially in fragmented habitats. Integrating satellite-borne multispectral and radar imagery, particularly through Earth Observation (EO) Foundation models (EO-FMs), enhanced within-formation discrimination and overall performance. Finally, ensemble machine learning that corrects class imbalance boosted predictive accuracy even further. Our methodological framework is transferable beyond Europe and adaptable to other classification systems. Future research should advance temporal modelling of habitat dynamics, extend to habitat segmentation and quality assessment, and exploit next-generation EO data paired with higher-quality in situ observations.
Authors: Minhak Song, Beomhan Baek, Kwangjun Ahn, Chulhee Yun
Abstract: As both model and dataset sizes continue to scale rapidly, conventional pretraining strategies with fixed compute budgets-such as cosine learning rate schedules-are increasingly inadequate for large-scale training. Recent alternatives, including warmup-stable-decay (WSD) schedules and weight averaging, offer greater flexibility. However, WSD relies on explicit decay phases to track progress, while weight averaging addresses this limitation at the cost of additional memory. In search of a more principled and scalable alternative, we revisit the Schedule-Free (SF) method [Defazio et al., 2024], which has shown strong empirical performance across diverse settings. We show that SF-AdamW effectively navigates the "river" structure of the loss landscape without decay phases or auxiliary averaging, making it particularly suitable for continuously scaling training workloads. To understand this behavior, we conduct a theoretical and empirical analysis of SF dynamics, revealing that it implicitly performs weight averaging without memory overhead. Guided by this analysis, we propose a refined variant of SF that improves robustness to momentum and performs better under large batch sizes, addressing key limitations of the original method. Together, these results establish SF as a practical, scalable, and theoretically grounded approach for language model training.
Authors: Andrew C. Li, Toryn Q. Klassen, Andrew Wang, Parand A. Alamdari, Sheila A. McIlraith
Abstract: Grounding language in perception and action is a key challenge when building situated agents that can interact with humans, or other agents, via language. In the past, addressing this challenge has required manually designing the language grounding or curating massive datasets that associate language with the environment. We propose Ground-Compose-Reinforce, an end-to-end, neurosymbolic framework for training RL agents directly from high-level task specifications--without manually designed reward functions or other domain-specific oracles, and without massive datasets. These task specifications take the form of Reward Machines, automata-based representations that capture high-level task structure and are in some cases autoformalizable from natural language. Critically, we show that Reward Machines can be grounded using limited data by exploiting compositionality. Experiments in a custom Meta-World domain with only 350 labelled pretraining trajectories show that our framework faithfully elicits complex behaviours from high-level specifications--including behaviours that never appear in pretraining--while non-compositional approaches fail.
Authors: Meiliang Liu, Huiwen Dong, Xiaoxiao Yang, Yunfang Xu, Zijin Li, Zhengye Si, Xinyue Yang, Zhiwen Zhao
Abstract: With the advancement of deep learning technologies, various neural network-based Granger causality models have been proposed. Although these models have demonstrated notable improvements, several limitations remain. Most existing approaches adopt the component-wise architecture, necessitating the construction of a separate model for each time series, which results in substantial computational costs. In addition, imposing the sparsity-inducing penalty on the first-layer weights of the neural network to extract causal relationships weakens the model's ability to capture complex interactions. To address these limitations, we propose Gradient Regularization-based Neural Granger Causality (GRNGC), which requires only one time series prediction model and applies $L_{1}$ regularization to the gradient between model's input and output to infer Granger causality. Moreover, GRNGC is not tied to a specific time series forecasting model and can be implemented with diverse architectures such as KAN, MLP, and LSTM, offering enhanced flexibility. Numerical simulations on DREAM, Lorenz-96, fMRI BOLD, and CausalTime show that GRNGC outperforms existing baselines and significantly reduces computational overhead. Meanwhile, experiments on real-world DNA, Yeast, HeLa, and bladder urothelial carcinoma datasets further validate the model's effectiveness in reconstructing gene regulatory networks.
Authors: Yimeng Chen, Piotr Pi\c{e}kos, Mateusz Ostaszewski, Firas Laakom, J\"urgen Schmidhuber
Abstract: Evaluating the scientific discovery capabilities of large language model based agents, particularly how they cope with varying environmental complexity and utilize prior knowledge, requires specialized benchmarks currently lacking in the landscape. To address this gap, we introduce \textsc{PhysGym}, a novel benchmark suite and simulation platform for rigorously assessing LLM-based scientific reasoning in interactive physics environments. \textsc{PhysGym}'s primary contribution lies in its sophisticated control over the level of prior knowledge provided to the agent. This allows researchers to dissect agent performance along axes including the complexity of the problem and the prior knowledge levels. The benchmark comprises a suite of interactive simulations, where agents must actively probe environments, gather data sequentially under constraints and formulate hypotheses about underlying physical laws. \textsc{PhysGym} provides standardized evaluation protocols and metrics for assessing hypothesis accuracy and model fidelity. We demonstrate the benchmark's utility by presenting results from baseline LLMs, showcasing its ability to differentiate capabilities based on varying priors and task complexity.
Authors: Anirudh Satheesh, Anant Khandelwal, Mucong Ding, Radu Balan
Abstract: Neural operators offer a powerful paradigm for solving partial differential equations (PDEs) that cannot be solved analytically by learning mappings between function spaces. However, there are two main bottlenecks in training neural operators: they require a significant amount of training data to learn these mappings, and this data needs to be labeled, which can only be accessed via expensive simulations with numerical solvers. To alleviate both of these issues simultaneously, we propose PICore, an unsupervised coreset selection framework that identifies the most informative training samples without requiring access to ground-truth PDE solutions. PICore leverages a physics-informed loss to select unlabeled inputs by their potential contribution to operator learning. After selecting a compact subset of inputs, only those samples are simulated using numerical solvers to generate labels, reducing annotation costs. We then train the neural operator on the reduced labeled dataset, significantly decreasing training time as well. Across four diverse PDE benchmarks and multiple coreset selection strategies, PICore achieves up to 78% average increase in training efficiency relative to supervised coreset selection methods with minimal changes in accuracy. We provide code at https://github.com/Asatheesh6561/PICore.
Authors: Fabian Akkerman, Julien Ferry, Christian Artigues, Emmanuel Hebrard, Thibaut Vidal
Abstract: Despite their theoretical appeal, totally corrective boosting methods based on linear programming have received limited empirical attention. In this paper, we conduct the first large-scale experimental study of six LP-based boosting formulations, including two novel methods, NM-Boost and QRLP-Boost, across 20 diverse datasets. We evaluate the use of both heuristic and optimal base learners within these formulations, and analyze not only accuracy, but also ensemble sparsity, margin distribution, anytime performance, and hyperparameter sensitivity. We show that totally corrective methods can outperform or match state-of-the-art heuristics like XGBoost and LightGBM when using shallow trees, while producing significantly sparser ensembles. We further show that these methods can thin pre-trained ensembles without sacrificing performance, and we highlight both the strengths and limitations of using optimal decision trees in this context.
Authors: Linh Le Pham Van, Minh Hoang Nguyen, Duc Kieu, Hung Le, Hung The Tran, Sunil Gupta
Abstract: Cross-domain offline reinforcement learning (RL) seeks to enhance sample efficiency in offline RL by utilizing additional offline source datasets. A key challenge is to identify and utilize source samples that are most relevant to the target domain. Existing approaches address this challenge by measuring domain gaps through domain classifiers, target transition dynamics modeling, or mutual information estimation using contrastive loss. However, these methods often require large target datasets, which is impractical in many real-world scenarios. In this work, we address cross-domain offline RL under a limited target data setting, identifying two primary challenges: (1) Dataset imbalance, which is caused by large source and small target datasets and leads to overfitting in neural network-based domain gap estimators, resulting in uninformative measurements; and (2) Partial domain overlap, where only a subset of the source data is closely aligned with the target domain. To overcome these issues, we propose DmC, a novel framework for cross-domain offline RL with limited target samples. Specifically, DmC utilizes $k$-nearest neighbor ($k$-NN) based estimation to measure domain proximity without neural network training, effectively mitigating overfitting. Then, by utilizing this domain proximity, we introduce a nearest-neighbor-guided diffusion model to generate additional source samples that are better aligned with the target domain, thus enhancing policy learning with more effective source samples. Through theoretical analysis and extensive experiments in diverse MuJoCo environments, we demonstrate that DmC significantly outperforms state-of-the-art cross-domain offline RL methods, achieving substantial performance gains.
Authors: Abdelrhman Gaber, Hassan Abd-Eltawab, John Elgallab, Youssif Abuzied, Dineo Mpanya, Turgay Celik, Swarun Kumar, Tamer ElBatt
Abstract: Cardiovascular diseases (CVD) cause over 17 million deaths annually worldwide, highlighting the urgent need for privacy-preserving predictive systems. We introduce FedCVD++, an enhanced federated learning (FL) framework that integrates both parametric models (logistic regression, SVM, neural networks) and non-parametric models (Random Forest, XGBoost) for coronary heart disease risk prediction. To address key FL challenges, we propose: (1) tree-subset sampling that reduces Random Forest communication overhead by 70%, (2) XGBoost-based feature extraction enabling lightweight federated ensembles, and (3) federated SMOTE synchronization for resolving cross-institutional class imbalance. Evaluated on the Framingham dataset (4,238 records), FedCVD++ achieves state-of-the-art results: federated XGBoost (F1 = 0.80) surpasses its centralized counterpart (F1 = 0.78), and federated Random Forest (F1 = 0.81) matches non-federated performance. Additionally, our communication-efficient strategies reduce bandwidth consumption by 3.2X while preserving 95% accuracy. Compared to existing FL frameworks, FedCVD++ delivers up to 15% higher F1-scores and superior scalability for multi-institutional deployment. This work represents the first practical integration of non-parametric models into federated healthcare systems, providing a privacy-preserving solution validated under real-world clinical constraints.
Authors: Adam Darmanin, Vince Vella
Abstract: Algorithmic trading requires short-term tactical decisions consistent with long-term financial objectives. Reinforcement Learning (RL) has been applied to such problems, but adoption is limited by myopic behaviour and opaque policies. Large Language Models (LLMs) offer complementary strategic reasoning and multi-modal signal interpretation when guided by well-structured prompts. This paper proposes a hybrid framework in which LLMs generate high-level trading strategies to guide RL agents. We evaluate (i) the economic rationale of LLM-generated strategies through expert review, and (ii) the performance of LLM-guided agents against unguided RL baselines using Sharpe Ratio (SR) and Maximum Drawdown (MDD). Empirical results indicate that LLM guidance improves both return and risk metrics relative to standard RL.
Authors: Carlo Cena, Mauro Martini, Marcello Chiaberge
Abstract: Attitude control is a fundamental aspect of spacecraft operations. Model Predictive Control (MPC) has emerged as a powerful strategy for these tasks, relying on accurate models of the system dynamics to optimize control actions over a prediction horizon. In scenarios where physics models are incomplete, difficult to derive, or computationally expensive, machine learning offers a flexible alternative by learning the system behavior directly from data. However, purely data-driven models often struggle with generalization and stability, especially when applied to inputs outside their training domain. To address these limitations, we investigate the benefits of incorporating Physics-Informed Neural Networks (PINNs) into the learning of spacecraft attitude dynamics, comparing their performance with that of purely data-driven approaches. Using a Real-valued Non-Volume Preserving (Real NVP) neural network architecture with a self-attention mechanism, we trained several models on simulated data generated with the Basilisk simulator. Two training strategies were considered: a purely data-driven baseline and a physics-informed variant to improve robustness and stability. Our results demonstrate that the inclusion of physics-based information significantly enhances the performance in terms of the mean relative error with the best architectures found by 27.08%. These advantages are particularly evident when the learned models are integrated into an MPC framework, where PINN-based models consistently outperform their purely data-driven counterparts in terms of control accuracy and robustness, and achieve improved settling times when compared to traditional MPC approaches, yielding improvements of up to 62%, when subject to observation noise and RWs friction.
Authors: Zihe Liu, Jiashun Liu, Yancheng He, Weixun Wang, Jiaheng Liu, Ling Pan, Xinyu Hu, Shaopan Xiong, Ju Huang, Jian Hu, Shengyi Huang, Johan Obando-Ceron, Siran Yang, Jiamang Wang, Wenbo Su, Bo Zheng
Abstract: Reinforcement learning for LLM reasoning has rapidly emerged as a prominent research area, marked by a significant surge in related studies on both algorithmic innovations and practical applications. Despite this progress, several critical challenges remain, including the absence of standardized guidelines for employing RL techniques and a fragmented understanding of their underlying mechanisms. Additionally, inconsistent experimental settings, variations in training data, and differences in model initialization have led to conflicting conclusions, obscuring the key characteristics of these techniques and creating confusion among practitioners when selecting appropriate techniques. This paper systematically reviews widely adopted RL techniques through rigorous reproductions and isolated evaluations within a unified open-source framework. We analyze the internal mechanisms, applicable scenarios, and core principles of each technique through fine-grained experiments, including datasets of varying difficulty, model sizes, and architectures. Based on these insights, we present clear guidelines for selecting RL techniques tailored to specific setups, and provide a reliable roadmap for practitioners navigating the RL for the LLM domain. Finally, we reveal that a minimalist combination of two techniques can unlock the learning capability of critic-free policies using vanilla PPO loss. The results demonstrate that our simple combination consistently improves performance, surpassing strategies like GRPO and DAPO.
Authors: Reza Shirkavand, Shangqian Gao, Peiran Yu, Heng Huang
Abstract: We study cost-aware routing for large language models across diverse and dynamic pools of models. Existing approaches often overlook prompt-specific context, rely on expensive model profiling, assume a fixed set of experts, or use inefficient trial-and-error strategies. We introduce Cost-Spectrum Contrastive Routing (CSCR), a lightweight framework that maps both prompts and models into a shared embedding space to enable fast, cost-sensitive selection. CSCR uses compact, fast-to-compute logit footprints for open-source models and perplexity fingerprints for black-box APIs. A contrastive encoder is trained to favor the cheapest accurate expert within adaptive cost bands. At inference time, routing reduces to a single k-NN lookup via a FAISS index, requiring no retraining when the expert pool changes and enabling microsecond latency. Across multiple benchmarks, CSCR consistently outperforms baselines, improving the accuracy-cost tradeoff by up to 25%, while generalizing robustly to unseen LLMs and out-of-distribution prompts.
Authors: Jiyeon Kang, Songseong Kim, Chanhui Lee, Doyeong Hwang, Joanie Hayoun Chung, Yunkyung Ko, Sumin Lee, Sungwoong Kim, Sungbin Lim
Abstract: Ordering-based approaches to causal discovery identify topological orders of causal graphs, providing scalable alternatives to combinatorial search methods. Under the Additive Noise Model (ANM) assumption, recent causal ordering methods based on score matching require an accurate estimation of the Hessian diagonal of the log-densities. In this paper, we aim to improve the approximation of the Hessian diagonal of the log-densities, thereby enhancing the performance of ordering-based causal discovery algorithms. Existing approaches that rely on Stein gradient estimators are computationally expensive and memory-intensive, while diffusion-model-based methods remain unstable due to the second-order derivatives of score models. To alleviate these problems, we propose Score-informed Neural Operator (SciNO), a probabilistic generative model in smooth function spaces designed to stably approximate the Hessian diagonal and to preserve structural information during the score modeling. Empirical results show that SciNO reduces order divergence by 42.7% on synthetic graphs and by 31.5% on real-world datasets on average compared to DiffAN, while maintaining memory efficiency and scalability. Furthermore, we propose a probabilistic control algorithm for causal reasoning with autoregressive models that integrates SciNO's probability estimates with autoregressive model priors, enabling reliable data-driven causal ordering informed by semantic information. Consequently, the proposed method enhances causal reasoning abilities of LLMs without additional fine-tuning or prompt engineering.
Authors: Odin Hoff Gardaa, Nello Blaser
Abstract: Contours or closed planar curves are common in many domains. For example, they appear as object boundaries in computer vision, isolines in meteorology, and the orbits of rotating machinery. In many cases when learning from contour data, planar rotations of the input will result in correspondingly rotated outputs. It is therefore desirable that deep learning models be rotationally equivariant. In addition, contours are typically represented as an ordered sequence of edge points, where the choice of starting point is arbitrary. It is therefore also desirable for deep learning methods to be equivariant under cyclic shifts. We present RotaTouille, a deep learning framework for learning from contour data that achieves both rotation and cyclic shift equivariance through complex-valued circular convolution. We further introduce and characterize equivariant non-linearities, coarsening layers, and global pooling layers to obtain invariant representations for downstream tasks. Finally, we demonstrate the effectiveness of RotaTouille through experiments in shape classification, reconstruction, and contour regression.
Authors: Jonas A. Actor, Anthony Gruber, Eric C. Cyr
Abstract: While attention has been empirically shown to improve model performance, it lacks a rigorous mathematical justification. This short paper establishes a novel connection between attention mechanisms and multinomial regression. Specifically, we show that in a fixed multinomial regression setting, optimizing over latent features yields solutions that align with the dynamics induced on features by attention blocks. In other words, the evolution of representations through a transformer can be interpreted as a trajectory that recovers the optimal features for classification.
Authors: Lucas Correia, Jan-Christoph Goos, Thomas B\"ack, Anna V. Kononova
Abstract: Truly unsupervised approaches for time series anomaly detection are rare in the literature. Those that exist suffer from a poorly set threshold, which hampers detection performance, while others, despite claiming to be unsupervised, need to be calibrated using a labelled data subset, which is often not available in the real world. This work integrates active learning with an existing unsupervised anomaly detection method by selectively querying the labels of multivariate time series, which are then used to refine the threshold selection process. To achieve this, we introduce a novel query strategy called the dissimilarity-based query strategy (DQS). DQS aims to maximise the diversity of queried samples by evaluating the similarity between anomaly scores using dynamic time warping. We assess the detection performance of DQS in comparison to other query strategies and explore the impact of mislabelling, a topic that is underexplored in the literature. Our findings indicate that DQS performs best in small-budget scenarios, though the others appear to be more robust when faced with mislabelling. Therefore, in the real world, the choice of query strategy depends on the expertise of the oracle and the number of samples they are willing to label. Regardless, all query strategies outperform the unsupervised threshold even in the presence of mislabelling. Thus, whenever it is feasible to query an oracle, employing an active learning-based threshold is recommended.
Authors: Xi He
Abstract: In the first paper (part I) of this series of two, we introduce four novel definitions of the ODT problems: three for size-constrained trees and one for depth-constrained trees. These definitions are stated unambiguously through executable recursive programs, satisfying all criteria we propose for a formal specification. In this sense, they resemble the "standard form" used in the study of general-purpose solvers. Grounded in algebraic programming theory-a relational formalism for deriving correct-by-construction algorithms from specifications-we can not only establish the existence or nonexistence of dynamic programming solutions but also derive them constructively whenever they exist. Consequently, the four generic problem definitions yield four novel optimal algorithms for ODT problems with arbitrary splitting rules that satisfy the axioms and objective functions of a given form. These algorithms encompass the known depth-constrained, axis-parallel ODT algorithm as the special case, while providing a unified, efficient, and elegant solution for the general ODT problem. In Part II, we present the first optimal hypersurface decision tree algorithm and provide comprehensive experiments against axis-parallel decision tree algorithms, including heuristic CART and state-of-the-art optimal methods. The results demonstrate the significant potential of decision trees with flexible splitting rules. Moreover, our framework is readily extendable to support algorithms for constructing even more flexible decision trees, including those with mixed splitting rules.
Authors: Viet-Hoang Tran, Van Hoan Trinh, Khanh Vinh Bui, Tan M. Nguyen
Abstract: Linear Mode Connectivity (LMC) is a notable phenomenon in the loss landscapes of neural networks, wherein independently trained models have been observed to be connected--up to permutation symmetries--by linear paths in parameter space along which the loss remains consistently low. This observation challenges classical views of non-convex optimization and has implications for model ensembling, generalization, and our understanding of neural loss geometry. Inspired by recent studies on LMC in standard neural networks, we systematically investigate this phenomenon within Mixture-of-Experts (MoE) architectures--a class of models known for their scalability and computational efficiency, which combine traditional neural networks--referred to as experts--through a learnable gating mechanism. We begin by conducting a comprehensive analysis of both dense and sparse gating regimes, demonstrating that the symmetries inherent to MoE architectures are fully characterized by permutations acting on both the expert components and the gating function. Building on these foundational findings, we propose a matching algorithm that enables alignment between independently trained MoEs, thereby facilitating the discovery of LMC. Finally, we empirically validate the presence of LMC using our proposed algorithm across diverse MoE configurations--including dense, sparse, and shared-expert variants--under a wide range of model settings and datasets of varying scales and modalities. Our results confirm the existence of LMC in MoE architectures and offer fundamental insights into the functional landscape and optimization dynamics of deep learning models.
Authors: Xi He
Abstract: Decision trees are a ubiquitous model for classification and regression tasks due to their interpretability and efficiency. However, solving the optimal decision tree (ODT) problem remains a challenging combinatorial optimization task. Even for the simplest splitting rules--axis-parallel hyperplanes--it is NP-hard to optimize. In Part I of this series, we rigorously defined the proper decision tree model through four axioms and, based on these, introduced four formal definitions of the ODT problem. From these definitions, we derived four generic algorithms capable of solving ODT problems for arbitrary decision trees satisfying the axioms. We also analyzed the combinatorial geometric properties of hypersurfaces, showing that decision trees defined by polynomial hypersurface splitting rules satisfy the proper axioms that we proposed. In this second paper (Part II) of this two-part series, building on the algorithmic and geometric foundations established in Part I, we introduce the first hypersurface decision tree (HODT) algorithm. To the best of our knowledge, existing optimal decision tree methods are, to date, limited to hyperplane splitting rules--a special case of hypersurfaces--and rely on general-purpose solvers. In contrast, our HODT algorithm addresses the general hypersurface decision tree model without requiring external solvers. Using synthetic datasets generated from ground-truth hyperplane decision trees, we vary tree size, data size, dimensionality, and label and feature noise. Results showing that our algorithm recovers the ground truth more accurately than axis-parallel trees and exhibits greater robustness to noise. We also analyzed generalization performance across 30 real-world datasets, showing that HODT can achieve up to 30% higher accuracy than the state-of-the-art optimal axis-parallel decision tree algorithm when tree complexity is properly controlled.
Authors: Yue Song, T. Anderson Keller, Sevan Brodjian, Takeru Miyato, Yisong Yue, Pietro Perona, Max Welling
Abstract: Orientation-rich images, such as fingerprints and textures, often exhibit coherent angular directional patterns that are challenging to model using standard generative approaches based on isotropic Euclidean diffusion. Motivated by the role of phase synchronization in biological systems, we propose a score-based generative model built on periodic domains by leveraging stochastic Kuramoto dynamics in the diffusion process. In neural and physical systems, Kuramoto models capture synchronization phenomena across coupled oscillators -- a behavior that we re-purpose here as an inductive bias for structured image generation. In our framework, the forward process performs \textit{synchronization} among phase variables through globally or locally coupled oscillator interactions and attraction to a global reference phase, gradually collapsing the data into a low-entropy von Mises distribution. The reverse process then performs \textit{desynchronization}, generating diverse patterns by reversing the dynamics with a learned score function. This approach enables structured destruction during forward diffusion and a hierarchical generation process that progressively refines global coherence into fine-scale details. We implement wrapped Gaussian transition kernels and periodicity-aware networks to account for the circular geometry. Our method achieves competitive results on general image benchmarks and significantly improves generation quality on orientation-dense datasets like fingerprints and textures. Ultimately, this work demonstrates the promise of biologically inspired synchronization dynamics as structured priors in generative modeling.
Authors: Rikuto Kotoge, Zheng Chen, Tasuku Kimura, Yasuko Matsubara, Takufumi Yanagisawa, Haruhiko Kishima, Yasushi Sakurai
Abstract: Dynamic GNNs, which integrate temporal and spatial features in Electroencephalography (EEG) data, have shown great potential in automating seizure detection. However, fully capturing the underlying dynamics necessary to represent brain states, such as seizure and non-seizure, remains a non-trivial task and presents two fundamental challenges. First, most existing dynamic GNN methods are built on temporally fixed static graphs, which fail to reflect the evolving nature of brain connectivity during seizure progression. Second, current efforts to jointly model temporal signals and graph structures and, more importantly, their interactions remain nascent, often resulting in inconsistent performance. To address these challenges, we present the first theoretical analysis of these two problems, demonstrating the effectiveness and necessity of explicit dynamic modeling and time-then-graph dynamic GNN method. Building on these insights, we propose EvoBrain, a novel seizure detection model that integrates a two-stream Mamba architecture with a GCN enhanced by Laplacian Positional Encoding, following neurological insights. Moreover, EvoBrain incorporates explicitly dynamic graph structures, allowing both nodes and edges to evolve over time. Our contributions include (a) a theoretical analysis proving the expressivity advantage of explicit dynamic modeling and time-then-graph over other approaches, (b) a novel and efficient model that significantly improves AUROC by 23% and F1 score by 30%, compared with the dynamic GNN baseline, and (c) broad evaluations of our method on the challenging early seizure prediction tasks.
Authors: Jorge-Humberto Urrea-Quintero, David Anton, Laura De Lorenzis, Henning Wessels
Abstract: The automated discovery of constitutive models from data has recently emerged as a promising alternative to the traditional model calibration paradigm. In this work, we present a fully automated framework for constitutive model discovery that systematically pairs three sparse regression algorithms (Least Absolute Shrinkage and Selection Operator (LASSO), Least Angle Regression (LARS), and Orthogonal Matching Pursuit (OMP)) with three model selection criteria: $K$-fold cross-validation (CV), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC). This pairing yields nine distinct algorithms for model discovery and enables a systematic exploration of the trade-off between sparsity, predictive performance, and computational cost. While LARS serves as an efficient path-based solver for the $\ell_1$-constrained problem, OMP is introduced as a tractable heuristic for $\ell_0$-regularized selection. The framework is applied to both isotropic and anisotropic hyperelasticity, utilizing both synthetic and experimental datasets. Results reveal that all nine algorithm-criterion combinations perform consistently well in discovering isotropic and anisotropic materials, yielding highly accurate constitutive models. These findings broaden the range of viable discovery algorithms beyond $\ell_1$-based approaches such as LASSO.
Authors: Haocheng Luo, Mehrtash Harandi, Dinh Phung, Trung Le
Abstract: Sharpness-aware minimization (SAM) has emerged as a highly effective technique for improving model generalization, but its underlying principles are not fully understood. We investigated the phenomenon known as m-sharpness, where the performance of SAM improves monotonically as the micro-batch size for computing perturbations decreases. In practice, the empirical m-sharpness effect underpins the deployment of SAM in distributed training, yet a rigorous theoretical account has remained lacking. To provide a theoretical explanation for m-sharpness, we leverage an extended Stochastic Differential Equation (SDE) framework and analyze the structure of stochastic gradient noise (SGN) to characterize the dynamics of various SAM variants, including n-SAM and m-SAM. Our findings reveal that the stochastic noise introduced during SAM perturbations inherently induces a variance-based sharpness regularization effect. Motivated by our theoretical insights, we introduce Reweighted SAM (RW-SAM), which employs sharpness-weighted sampling to mimic the generalization benefits of m-SAM while remaining parallelizable. Comprehensive experiments validate the effectiveness of our theoretical analysis and proposed method.
Authors: Balazs Pejo, Marcell Frank, Krisztian Varga, Peter Veliczky, Gergely Biczok
Abstract: This paper investigates the fragility of contribution evaluation in federated learning, a critical mechanism for ensuring fairness and incentivizing participation. We argue that contribution scores are susceptible to significant distortions from two fundamental perspectives: architectural sensitivity and intentional manipulation. First, we explore how different model aggregation methods impact these scores. While most research assumes a basic averaging approach, we demonstrate that advanced techniques, including those designed to handle unreliable or diverse clients, can unintentionally yet significantly alter the final scores. Second, we explore vulnerabilities posed by poisoning attacks, where malicious participants strategically manipulate their model updates to inflate their own contribution scores or reduce the importance of other participants. Through extensive experiments across diverse datasets and model architectures, implemented within the Flower framework, we rigorously show that both the choice of aggregation method and the presence of attackers are potent vectors for distorting contribution scores, highlighting a critical need for more robust evaluation schemes.
Authors: Daehee Lee, Dongsu Lee, TaeYoon Kwack, Wonje Choi, Honguk Woo
Abstract: Skill Incremental Learning (SIL) is the process by which an embodied agent expands and refines its skill set over time by leveraging experience gained through interaction with its environment or by the integration of additional data. SIL facilitates efficient acquisition of hierarchical policies grounded in reusable skills for downstream tasks. However, as the skill repertoire evolves, it can disrupt compatibility with existing skill-based policies, limiting their reusability and generalization. In this work, we propose SIL-C, a novel framework that ensures skill-policy compatibility, allowing improvements in incrementally learned skills to enhance the performance of downstream policies without requiring policy re-training or structural adaptation. SIL-C employs a bilateral lazy learning-based mapping technique to dynamically align the subtask space referenced by policies with the skill space decoded into agent behaviors. This enables each subtask, derived from the policy's decomposition of a complex task, to be executed by selecting an appropriate skill based on trajectory distribution similarity. We evaluate SIL-C across diverse SIL scenarios and demonstrate that it maintains compatibility between evolving skills and downstream policies while ensuring efficiency throughout the learning process.
Authors: Chang Deng, Bryon Aragam
Abstract: Existing methods for differentiable structure learning in discrete data typically assume that the data are generated from specific structural equation models. However, these assumptions may not align with the true data-generating process, which limits the general applicability of such methods. Furthermore, current approaches often ignore the complex dependence structure inherent in discrete data and consider only linear effects. We propose a differentiable structure learning framework that is capable of capturing arbitrary dependencies among discrete variables. We show that although general discrete models are unidentifiable from purely observational data, it is possible to characterize the complete set of compatible parameters and structures. Additionally, we establish identifiability up to Markov equivalence under mild assumptions. We formulate the learning problem as a single differentiable optimization task in the most general form, thereby avoiding the unrealistic simplifications adopted by previous methods. Empirical results demonstrate that our approach effectively captures complex relationships in discrete data.
Authors: Pavan Karjol, Vivek V Kashyap, Rohan Kashyap, Prathosh A P
Abstract: We introduce a novel framework for the automatic discovery of one-parameter subgroups ($H_{\gamma}$) of $SO(3)$ and, more generally, $SO(n)$. One-parameter subgroups of $SO(n)$ are crucial in a wide range of applications, including robotics, quantum mechanics, and molecular structure analysis. Our method utilizes the standard Jordan form of skew-symmetric matrices, which define the Lie algebra of $SO(n)$, to establish a canonical form for orbits under the action of $H_{\gamma}$. This canonical form is then employed to derive a standardized representation for $H_{\gamma}$-invariant functions. By learning the appropriate parameters, the framework uncovers the underlying one-parameter subgroup $H_{\gamma}$. The effectiveness of the proposed approach is demonstrated through tasks such as double pendulum modeling, moment of inertia prediction, top quark tagging and invariant polynomial regression, where it successfully recovers meaningful subgroup structure and produces interpretable, symmetry-aware representations.
Authors: Chris Kolb, Laetitia Frost, Bernd Bischl, David R\"ugamer
Abstract: Structured sparsity regularization offers a principled way to compact neural networks, but its non-differentiability breaks compatibility with conventional stochastic gradient descent and requires either specialized optimizers or additional post-hoc pruning without formal guarantees. In this work, we propose $D$-Gating, a fully differentiable structured overparameterization that splits each group of weights into a primary weight vector and multiple scalar gating factors. We prove that any local minimum under $D$-Gating is also a local minimum using non-smooth structured $L_{2,2/D}$ penalization, and further show that the $D$-Gating objective converges at least exponentially fast to the $L_{2,2/D}$-regularized loss in the gradient flow limit. Together, our results show that $D$-Gating is theoretically equivalent to solving the original group sparsity problem, yet induces distinct learning dynamics that evolve from a non-sparse regime into sparse optimization. We validate our theory across vision, language, and tabular tasks, where $D$-Gating consistently delivers strong performance-sparsity tradeoffs and outperforms both direct optimization of structured penalties and conventional pruning baselines.
Authors: Dang Huu-Tien, Minh-Phuong Nguyen, Naoya Inoue
Abstract: Label noise in datasets could significantly damage the performance and robustness of deep neural networks (DNNs) trained on these datasets. As the size of modern DNNs grows, there is a growing demand for automated tools for detecting such errors. In this paper, we propose post-hoc, model-agnostic noise detection and rectification methods utilizing the penultimate feature from a DNN. Our idea is based on the observation that the similarity between the penultimate feature of a mislabeled data point and its true class data points is higher than that for data points from other classes, making the probability of label occurrence within a tight, similar cluster informative for detecting and rectifying errors. Through theoretical and empirical analyses, we demonstrate that our approach achieves high detection performance across diverse, realistic noise scenarios and can automatically rectify these errors to improve dataset quality. Our implementation is available at https://anonymous.4open.science/r/noise-detection-and-rectification-AD8E.
URLs: https://anonymous.4open.science/r/noise-detection-and-rectification-AD8E.
Authors: Jayanta Mandi, Marianne Defresne, Senne Berden, Tias Guns
Abstract: When some parameters of a constrained optimization problem (COP) are uncertain, this gives rise to a predict-then-optimize (PtO) problem, comprising two stages: the prediction of the unknown parameters from contextual information and the subsequent optimization using those predicted parameters. Decision-focused learning (DFL) implements the first stage by training a machine learning (ML) model to optimize the quality of the decisions made using the predicted parameters. When the predicted parameters occur in the constraints, they can lead to infeasible solutions. Therefore, it is important to simultaneously manage both feasibility and decision quality. We develop a DFL framework for predicting constraint parameters in a generic COP. While prior works typically assume that the underlying optimization problem is a linear program (LP) or integer LP (ILP), our approach makes no such assumption. We derive two novel loss functions based on maximum likelihood estimation (MLE): the first one penalizes infeasibility (by penalizing predicted parameters that lead to infeasible solutions), while the second one penalizes suboptimal decisions (by penalizing predicted parameters that make the true optimal solution infeasible). We introduce a single tunable parameter to form a weighted average of the two losses, allowing decision-makers to balance suboptimality and feasibility. We experimentally demonstrate that adjusting this parameter provides decision-makers control over this trade-off. Moreover, across several COP instances, we show that adjusting the tunable parameter allows a decision-maker to prioritize either suboptimality or feasibility, outperforming the performance of existing baselines in either objective.
Authors: Yuyao Wang, Yu-Hung Cheng, Debarghya Mukherjee, Huimin Cheng
Abstract: Graphon models provide a flexible nonparametric framework for estimating latent connectivity probabilities in networks, enabling a range of downstream applications such as link prediction and data augmentation. However, accurate graphon estimation typically requires a large graph, whereas in practice, one often only observes a small-sized network. One approach to addressing this issue is to adopt a transfer learning framework, which aims to improve estimation in a small target graph by leveraging structural information from a larger, related source graph. In this paper, we propose a novel method, namely GTRANS, a transfer learning framework that integrates neighborhood smoothing and Gromov-Wasserstein optimal transport to align and transfer structural patterns between graphs. To prevent negative transfer, GTRANS includes an adaptive debiasing mechanism that identifies and corrects for target-specific deviations via residual smoothing. We provide theoretical guarantees on the stability of the estimated alignment matrix and demonstrate the effectiveness of GTRANS in improving the accuracy of target graph estimation through extensive synthetic and real data experiments. These improvements translate directly to enhanced performance in downstream applications, such as the graph classification task and the link prediction task.
Authors: Binbin Huang, Luo Luo, Yanghua Xiao, Deqing Yang, Baojian Zhou
Abstract: This work proposes a novel framework based on nested evolving set processes to accelerate Personalized PageRank (PPR) computation. At each stage of the process, we employ a localized inexact proximal point iteration to solve a simplified linear system. We show that the time complexity of such localized methods is upper bounded by $\min\{\tilde{\mathcal{O}}(R^2/\epsilon^2), \tilde{\mathcal{O}}(m)\}$ to obtain an $\epsilon$-approximation of the PPR vector, where $m$ denotes the number of edges in the graph and $R$ is a constant defined via nested evolving set processes. Furthermore, the algorithms induced by our framework require solving only $\tilde{\mathcal{O}}(1/\sqrt{\alpha})$ such linear systems, where $\alpha$ is the damping factor. When $1/\epsilon^2\ll m$, this implies the existence of an algorithm that computes an $\ epsilon $-approximation of the PPR vector with an overall time complexity of $\tilde{\mathcal{O}}\left(R^2 / (\sqrt{\alpha}\epsilon^2)\right)$, independent of the underlying graph size. Our result resolves an open conjecture from existing literature. Experimental results on real-world graphs validate the efficiency of our methods, demonstrating significant convergence in the early stages.
Authors: Le-Trung Nguyen, Enzo Tartaglione, Van-Tam Nguyen
Abstract: As AI increasingly shapes daily life, energy consumption and data privacy have become pressing concerns. On-device learning trains models directly on edge devices, cutting energy consumption and safeguarding data privacy. However, the expanding scale of modern neural networks creates a major obstacle for on-device training. Although prior work has concentrated on compact convolutional architectures, we instead apply subspace-based training to transformer models. Motivated by the idea that a model's essential information lies in a fixed subspace, we introduce Weight-Activation Subspace Iteration (WASI), a method that mitigates the memory bottleneck of backpropagation and boosts inference efficiency in transformer models by restricting training to this subspace. Our results demonstrate that WASI maintains accuracy comparable to vanilla training while reducing memory usage by up to $62\times$ and computational cost (FLOPs) by up to $2\times$. On a Raspberry Pi 5, WASI achieves roughly $1.5\times$ faster training and inference than vanilla training.
Authors: Ali Atiah Alzahrani
Abstract: We examine whether regime-conditioned generative scenarios combined with a convex CVaR allocator improve portfolio decisions under regime shifts. We present MARCD, a generative-to-decision framework with: (i) a Gaussian HMM to infer latent regimes; (ii) a diffusion generator that produces regime-conditioned scenarios; (iii) signal extraction via blended, shrunk moments; and (iv) a governed CVaR epigraph quadratic program. Contributions: Within the Scenario stage we introduce a tail-weighted diffusion objective that up-weights low-quantile outcomes relevant for drawdowns and a regime-expert (MoE) denoiser whose gate increases with crisis posteriors; both are evaluated end-to-end through the allocator. Under strict walk-forward on liquid multi-asset ETFs (2005-2025), MARCD exhibits stronger scenario calibration and materially smaller drawdowns: MaxDD 9.3% versus 14.1% for BL (a 34% reduction) over 2020-2025 out-of-sample. The framework provides an auditable pipeline with explicit budget, box, and turnover constraints, demonstrating the value of decision-aware generative modeling in finance.
Authors: Kiattikun Chobtham, Kanoksri Sarinnapakorn, Kritanai Torsri, Prattana Deeprasertkul, Jirawan Kamma
Abstract: Accurate rainfall forecasting, particularly for extreme events, remains a significant challenge in climatology and the Earth system. This paper presents novel physics-informed Graph Neural Networks (GNNs) combined with extreme-value analysis techniques to improve gauge-station rainfall predictions across Thailand. The model leverages a graph-structured representation of gauge stations to capture complex spatiotemporal patterns, and it offers explainability through teleconnections. We preprocess relevant climate indices that potentially influence regional rainfall. The proposed Graph Attention Network with Long Short-Term Memory (Attention-LSTM) applies the attention mechanism using initial edge features derived from simple orographic-precipitation physics formulation. The embeddings are subsequently processed by LSTM layers. To address extremes, we perform Peak-Over-Threshold (POT) mapping using the novel Spatial Season-aware Generalized Pareto Distribution (GPD) method, which overcomes limitations of traditional machine-learning models. Experiments demonstrate that our method outperforms well-established baselines across most regions, including areas prone to extremes, and remains strongly competitive with the state of the art. Compared with the operational forecasting system SEAS5, our real-world application improves extreme-event prediction and offers a practical enhancement to produce high-resolution maps that support decision-making in long-term water management.
Authors: Areej AlOtaibi, Lina Alyahya, Raghad Alshabanah, Shahad Alfawzan, Shuruq Alarefei, Reem Alsabti, Nouf Alsubaie, Abdulaziz Alhuzaymi, Lujain Alkhelb, Majd Alsayari, Waad Alahmed, Omar Talabay, Jalal Alowibdi, Salem Alelyani, Adel Bibi
Abstract: Large Language Models (LLMs) have significantly advanced the field of natural language processing, enhancing capabilities in both language understanding and generation across diverse domains. However, developing LLMs for Arabic presents unique challenges. This paper explores these challenges by focusing on critical aspects such as data curation, tokenizer design, and evaluation. We detail our approach to the collection and filtration of Arabic pre-training datasets, assess the impact of various tokenizer designs on model performance, and examine the limitations of existing Arabic evaluation frameworks, for which we propose a systematic corrective methodology. To promote transparency and facilitate collaborative development, we share our data and methodologies, contributing to the advancement of language modeling, particularly for the Arabic language.
Authors: Shinji Ito, Kevin Jamieson, Haipeng Luo, Arnab Maiti, Taira Tsuchiya
Abstract: We study online learning in finite-horizon episodic Markov decision processes (MDPs) under the challenging aggregate bandit feedback model, where the learner observes only the cumulative loss incurred in each episode, rather than individual losses at each state-action pair. While prior work in this setting has focused exclusively on worst-case analysis, we initiate the study of best-of-both-worlds (BOBW) algorithms that achieve low regret in both stochastic and adversarial environments. We propose the first BOBW algorithms for episodic tabular MDPs with aggregate bandit feedback. In the case of known transitions, our algorithms achieve $O(\log T)$ regret in stochastic settings and ${O}(\sqrt{T})$ regret in adversarial ones. Importantly, we also establish matching lower bounds, showing the optimality of our algorithms in this setting. We further extend our approach to unknown-transition settings by incorporating confidence-based techniques. Our results rely on a combination of FTRL over occupancy measures, self-bounding techniques, and new loss estimators inspired by recent advances in online shortest path problems. Along the way, we also provide the first individual-gap-dependent lower bounds and demonstrate near-optimal BOBW algorithms for shortest path problems with bandit feedback.
Authors: Tal Barami, Nimrod Berman, Ilan Naiman, Amos H. Hason, Rotem Ezra, Omri Azencot
Abstract: Learning disentangled representations in sequential data is a key goal in deep learning, with broad applications in vision, audio, and time series. While real-world data involves multiple interacting semantic factors over time, prior work has mostly focused on simpler two-factor static and dynamic settings, primarily because such settings make data collection easier, thereby overlooking the inherently multi-factor nature of real-world data. We introduce the first standardized benchmark for evaluating multi-factor sequential disentanglement across six diverse datasets spanning video, audio, and time series. Our benchmark includes modular tools for dataset integration, model development, and evaluation metrics tailored to multi-factor analysis. We additionally propose a post-hoc Latent Exploration Stage to automatically align latent dimensions with semantic factors, and introduce a Koopman-inspired model that achieves state-of-the-art results. Moreover, we show that Vision-Language Models can automate dataset annotation and serve as zero-shot disentanglement evaluators, removing the need for manual labels and human intervention. Together, these contributions provide a robust and scalable foundation for advancing multi-factor sequential disentanglement. Our code is available on GitHub, and the datasets and trained models are available on Hugging Face.
Authors: Yunfei Liang
Abstract: Recent advances in deep learning have led to a surge of open-source models across diverse domains. While model merging offers a promising way to combine their strengths, existing approaches often suffer from parameter conflicts that degrade performance on domain-specific tasks. We propose MIN-Merging, a router-based framework that selectively merges the most important neurons to reduce such conflicts. Extensive experiments on Computer Vision(CV) and Natural Language Processing(NLP) benchmarks show that MIN-Merging achieves consistent gains on in-domain tasks while retaining the generalization ability of pretrained models on out-of-domain tasks. These results highlight its effectiveness as a practical solution to the parameter conflict problem in model merging.
Authors: Wang Zixian
Abstract: Direct Preference Optimization (DPO) is an efficient alternative to reinforcement learning from human feedback (RLHF), yet it typically assumes hard binary labels and pairwise comparisons. Such assumptions can be brittle under noisy or distribution-shifted supervision. We present Anchored Direct Preference Optimization (ADPO), which (i) incorporates soft preference probabilities, (ii) aligns policy updates through reference anchoring that induces an implicit trust region, and (iii) extends to listwise learning via Plackett-Luce modeling. In controlled synthetic setups covering 12 scenarios (4 noise types x 3 severities) and 3 model scales, ADPO exhibits relative improvements ranging from 12% to 79% over a standard DPO baseline (10-seed means; 95% CIs in the Appendix). Hard labels tend to fare better under severe noise, whereas soft labels yield better calibration under distribution shift; listwise variants achieve the highest WinMass (expected probability mass on the ground-truth best item) in 9/12 scenarios. Larger models amplify ADPO's benefits (0.718 vs. 0.416 at hidden=256), suggesting that anchoring acts as an effective trust-region regularizer. We release code and configurations to facilitate reproducibility.
Authors: Omar El Mansouri, Mohamed El Amine Seddik, Salem Lahlou
Abstract: Reinforcement learning from human feedback (RLHF) or verifiable rewards (RLVR), the standard paradigm for aligning LLMs or building recent SOTA reasoning models, is highly sensitive to noise from inconsistent or erroneous rewards. Yet, the interaction between such noise and widely used group-based policy optimization methods remains underexplored. We introduce a noise-robust Group Relative Policy Optimization (GRPO) and Done Right GRPO (Dr.GRPO) framework that explicitly models reward corruption as Bernoulli noise. Our method applies noise correction after estimating reward flip probabilities to debias the learning signal, yielding provably unbiased gradient estimates. Theoretical analysis shows that group-based methods inherently mitigate individual-level noise, and our correction strategy amplifies this robustness. Empirically, we observe consistent improvements across math and code tasks when applying our noise correction to standard reward model usage, with particular gains of up to 6.7 percentage points in accuracy on math tasks and 1.5 on code tasks under realistic reward model conditions. This work bridges label-noise correction from supervised learning with modern RLHF, offering both theoretical insights and a practical algorithm for noisy real-world deployment.
Authors: Yaning Jia, Chunhui Zhang, Xingjian Diao, Xiangchi Yuan, Zhongyu Ouyang, Chiyu Ma, Soroush Vosoughi
Abstract: Curriculum learning (CL) - ordering training data from easy to hard - has become a popular strategy for improving reasoning in large language models (LLMs). Yet prior work employs disparate difficulty metrics and training setups, leaving open fundamental questions: When does curriculum help? Which direction - forward or reverse - is better? And does the answer depend on what we measure? We address these questions through a unified offline evaluation framework that decomposes curriculum difficulty into five complementary dimensions: Problem Difficulty, Model Surprisal, Confidence Margin, Predictive Uncertainty, and Decision Variability. Through controlled post-training experiments on mathematical reasoning benchmarks with Llama3.1-8B, Mistral-7B, and Gemma3-4B, we find that (i) no curriculum strategy dominates universally - the relative effectiveness of forward versus reverse CL depends jointly on model capability and task complexity; (ii) even within a single metric, samples at different difficulty levels produce distinct gains depending on task demands; and (iii) task-aligned curricula focus on shaping the model's final representations and generalization, whereas inner-state curricula modulate internal states such as confidence and uncertainty. Our findings challenge the notion of a universal curriculum strategy and offer actionable guidance across model and task regimes, with some metrics indicating that prioritizing decision-uncertain samples can further enhance learning outcomes.
Authors: Runzhe Wu, Ankur Samanta, Ayush Jain, Scott Fujimoto, Jeongyeol Kwon, Ben Kretzu, Youliang Yu, Kaveh Hassani, Boris Vidolov, Yonathan Efroni
Abstract: Multi-task post-training of large language models (LLMs) is typically performed by mixing datasets from different tasks and optimizing them jointly. This approach implicitly assumes that all tasks contribute gradients of similar magnitudes; when this assumption fails, optimization becomes biased toward large-gradient tasks. In this paper, however, we show that this assumption fails in RL post-training: certain tasks produce significantly larger gradients, thus biasing updates toward those tasks. Such gradient imbalance would be justified only if larger gradients implied larger learning gains on the tasks (i.e., larger performance improvements) -- but we find this is not true. Large-gradient tasks can achieve similar or even much lower learning gains than small-gradient ones. Further analyses reveal that these gradient imbalances cannot be explained by typical training statistics such as training rewards or advantages, suggesting that they arise from the inherent differences between tasks. This cautions against naive dataset mixing and calls for future work on principled gradient-level corrections for LLMs.
Authors: Paul Agbaje, Arkajyoti Mitra, Afia Anjum, Pranali Khose, Ebelechukwu Nwafor, Habeeb Olufowobi
Abstract: Knowledge distillation (KD) techniques have emerged as a powerful tool for transferring expertise from complex teacher models to lightweight student models, particularly beneficial for deploying high-performance models in resource-constrained devices. This approach has been successfully applied to graph neural networks (GNNs), harnessing their expressive capabilities to generate node embeddings that capture structural and feature-related information. In this study, we depart from the conventional KD approach by exploring the potential of collaborative learning among GNNs. In the absence of a pre-trained teacher model, we show that relatively simple and shallow GNN architectures can synergetically learn efficient models capable of performing better during inference, particularly in tackling multiple tasks. We propose a collaborative learning framework where ensembles of student GNNs mutually teach each other throughout the training process. We introduce an adaptive logit weighting unit to facilitate efficient knowledge exchange among models and an entropy enhancement technique to improve mutual learning. These components dynamically empower the models to adapt their learning strategies during training, optimizing their performance for downstream tasks. Extensive experiments conducted on three datasets each for node and graph classification demonstrate the effectiveness of our approach.
Authors: Bosong Huang, Ming Jin, Yuxuan Liang, Johan Barthelemy, Debo Cheng, Qingsong Wen, Chenghao Liu, Shirui Pan
Abstract: Explaining time series classification models is crucial, particularly in high-stakes applications such as healthcare and finance, where transparency and trust play a critical role. Although numerous time series classification methods have identified key subsequences, known as shapelets, as core features for achieving state-of-the-art performance and validating their pivotal role in classification outcomes, existing post-hoc time series explanation (PHTSE) methods primarily focus on timestep-level feature attribution. These explanation methods overlook the fundamental prior that classification outcomes are predominantly driven by key shapelets. To bridge this gap, we present ShapeX, an innovative framework that segments time series into meaningful shapelet-driven segments and employs Shapley values to assess their saliency. At the core of ShapeX lies the Shapelet Describe-and-Detect (SDD) framework, which effectively learns a diverse set of shapelets essential for classification. We further demonstrate that ShapeX produces explanations which reveal causal relationships instead of just correlations, owing to the atomicity properties of shapelets. Experimental results on both synthetic and real-world datasets demonstrate that ShapeX outperforms existing methods in identifying the most relevant subsequences, enhancing both the precision and causal fidelity of time series explanations.
Authors: Shumin Li
Abstract: The development of accessible screening tools for early cancer detection in dogs represents a significant challenge in veterinary medicine. Routine laboratory data offer a promising, low-cost source for such tools, but their utility is hampered by the non-specificity of individual biomarkers and the severe class imbalance inherent in screening populations. This study assesses the feasibility of cancer risk classification using the Golden Retriever Lifetime Study (GRLS) cohort under real-world constraints, including the grouping of diverse cancer types and the inclusion of post-diagnosis samples. A comprehensive benchmark evaluation was conducted, systematically comparing 126 analytical pipelines that comprised various machine learning models, feature selection methods, and data balancing techniques. Data were partitioned at the patient level to prevent leakage. The optimal model, a Logistic Regression classifier with class weighting and recursive feature elimination, demonstrated moderate ranking ability (AUROC = 0.815; 95% CI: 0.793-0.836) but poor clinical classification performance (F1-score = 0.25, Positive Predictive Value = 0.15). While a high Negative Predictive Value (0.98) was achieved, insufficient recall (0.79) precludes its use as a reliable rule-out test. Interpretability analysis with SHapley Additive exPlanations (SHAP) revealed that predictions were driven by non-specific features like age and markers of inflammation and anemia. It is concluded that while a statistically detectable cancer signal exists in routine lab data, it is too weak and confounded for clinically reliable discrimination from normal aging or other inflammatory conditions. This work establishes a critical performance ceiling for this data modality in isolation and underscores that meaningful progress in computational veterinary oncology will require integration of multi-modal data sources.
Authors: Saraf Anzum Shreya, MD. Abu Ismail Siddique, Sharaf Tasnim
Abstract: Brain tumors are a challenging problem in neuro-oncology, where early and precise diagnosis is important for successful treatment. Deep learning-based brain tumor classification methods often rely on heavy data augmentation which can limit generalization and trust in clinical applications. In this paper, we propose a double-backbone network integrating VGG16 and Xception with a Frequency-Gated Attention (FGA) Block to capture complementary local and global features. Unlike previous studies, our model achieves state-of-the-art performance without augmentation which demonstrates robustness to variably sized and distributed datasets. For further transparency, Grad-CAM is integrated to visualize the tumor regions based on which the model is giving prediction, bridging the gap between model prediction and clinical interpretability. The proposed framework achieves 99.24\% accuracy on the 7K-DS dataset for the 4-class setting, along with 98.68\% and 99.85\% in the 3-class and 2-class settings, respectively. On the independent 3K-DS dataset, the model generalizes with 95.77\% accuracy, outperforming baseline and state-of-the-art methods. To further support clinical usability, we developed a graphical user interface (GUI) that provides real-time classification and Grad-CAM-based tumor localization. These findings suggest that augmentation-free, interpretable, and deployable deep learning models such as DB-FGA-Net hold strong potential for reliable clinical translation in brain tumor diagnosis.
Authors: Yuta Kawamoto, Hideaki Iiduka
Abstract: Stochastic gradient descent (SGD) is the workhorse of large-scale learning, yet classical analyses rely on assumptions that can be either too strong (bounded variance) or too coarse (uniform noise). The expected smoothness (ES) condition has emerged as a flexible alternative that ties the second moment of stochastic gradients to the objective value and the full gradient. This paper presents a self-contained convergence analysis of SGD under ES. We (i) refine ES with interpretations and sampling-dependent constants; (ii) derive bounds of the expectation of squared full gradient norm; and (iii) prove $O(1/K)$ rates with explicit residual errors for various step-size schedules. All proofs are given in full detail in the appendix. Our treatment unifies and extends recent threads (Khaled and Richt\'arik, 2020; Umeda and Iiduka, 2025).
Authors: Edward Berman, Jacob Ginesin, Marco Pacini, Robin Walters
Abstract: Data-sparse settings such as robotic manipulation, molecular physics, and galaxy morphology classification are some of the hardest domains for deep learning. For these problems, equivariant networks can help improve modeling across undersampled parts of the input space, and uncertainty estimation can guard against overconfidence. However, until now, the relationships between equivariance and model confidence, and more generally equivariance and model calibration, has yet to be studied. Since traditional classification and regression error terms show up in the definitions of calibration error, it is natural to suspect that previous work can be used to help understand the relationship between equivariance and calibration error. In this work, we present a theory relating equivariance to uncertainty estimation. By proving lower and upper bounds on uncertainty calibration errors (ECE and ENCE) under various equivariance conditions, we elucidate the generalization limits of equivariant models and illustrate how symmetry mismatch can result in miscalibration in both classification and regression. We complement our theoretical framework with numerical experiments that clarify the relationship between equivariance and uncertainty using a variety of real and simulated datasets, and we comment on trends with symmetry mismatch, group size, and aleatoric and epistemic uncertainties.
Authors: Kaihua Tang, Yulei Niu, Jianqiang Huang, Jiaxin Shi, Hanwang Zhang
Abstract: Today's scene graph generation (SGG) task is still far from practical, mainly due to the severe training bias, e.g., collapsing diverse "human walk on / sit on / lay on beach" into "human on beach". Given such SGG, the down-stream tasks such as VQA can hardly infer better scene structures than merely a bag of objects. However, debiasing in SGG is not trivial because traditional debiasing methods cannot distinguish between the good and bad bias, e.g., good context prior (e.g., "person read book" rather than "eat") and bad long-tailed bias (e.g., "near" dominating "behind / in front of"). In this paper, we present a novel SGG framework based on causal inference but not the conventional likelihood. We first build a causal graph for SGG, and perform traditional biased training with the graph. Then, we propose to draw the counterfactual causality from the trained graph to infer the effect from the bad bias, which should be removed. In particular, we use Total Direct Effect (TDE) as the proposed final predicate score for unbiased SGG. Note that our framework is agnostic to any SGG model and thus can be widely applied in the community who seeks unbiased predictions. By using the proposed Scene Graph Diagnosis toolkit on the SGG benchmark Visual Genome and several prevailing models, we observed significant improvements over the previous state-of-the-art methods.
Authors: Kaihua Tang, Jianqiang Huang, Hanwang Zhang
Abstract: As the class size grows, maintaining a balanced dataset across many classes is challenging because the data are long-tailed in nature; it is even impossible when the sample-of-interest co-exists with each other in one collectable unit, e.g., multiple visual instances in one image. Therefore, long-tailed classification is the key to deep learning at scale. However, existing methods are mainly based on re-weighting/re-sampling heuristics that lack a fundamental theory. In this paper, we establish a causal inference framework, which not only unravels the whys of previous methods, but also derives a new principled solution. Specifically, our theory shows that the SGD momentum is essentially a confounder in long-tailed classification. On one hand, it has a harmful causal effect that misleads the tail prediction biased towards the head. On the other hand, its induced mediation also benefits the representation learning and head prediction. Our framework elegantly disentangles the paradoxical effects of the momentum, by pursuing the direct causal effect caused by an input sample. In particular, we use causal intervention in training, and counterfactual reasoning in inference, to remove the "bad" while keep the "good". We achieve new state-of-the-arts on three long-tailed visual recognition benchmarks: Long-tailed CIFAR-10/-100, ImageNet-LT for image classification and LVIS for instance segmentation.
Authors: Ziwei Zhang, Xin Wang, Zeyang Zhang, Peng Cui, Wenwu Zhu
Abstract: Deep neural networks have achieved great success in the last decade. When designing neural networks to handle the ubiquitous geometric data such as point clouds and graphs, it is critical that the model can maintain invariance towards various transformations such as translation, rotation, and scaling. Most existing graph neural network (GNN) approaches can only maintain permutation-invariance, failing to guarantee invariance with respect to other transformations. Besides GNNs, other works design sophisticated transformation-invariant layers, which are computationally expensive and difficult to be extended. In this paper, we revisit why general neural networks cannot maintain transformation invariance. Our findings show that transformation-invariant and distance-preserving initial point representations are sufficient to achieve transformation invariance rather than needing sophisticated neural layer designs. Motivated by these findings, we propose Transformation Invariant Neural Networks (TinvNN), a straightforward and general plug-in for geometric data. Specifically, we realize transformation invariant and distance-preserving initial point representations by modifying multi-dimensional scaling and feed the representations into existing neural networks. We prove that TinvNN can strictly guarantee transformation invariance, being general and flexible enough to be combined with the existing neural networks. Extensive experimental results on point cloud analysis and combinatorial optimization demonstrate the effectiveness and general applicability of our method. We also extend our method into equivariance cases. Based on the results, we advocate that TinvNN should be considered as an essential baseline for further studies of transformation-invariant geometric deep learning.
Authors: Yijia Wang, Daniel R. Jiang
Abstract: We study infinite horizon Markov decision processes (MDPs) with "fast-slow" structure, where some state variables evolve rapidly ("fast states") while others change more gradually ("slow states"). This structure commonly arises in practice when decisions must be made at high frequencies over long horizons, and where slowly changing information still plays a critical role in determining optimal actions. Examples include inventory control under slowly changing demand indicators or dynamic pricing with gradually shifting consumer behavior. Modeling the problem at the natural decision frequency leads to MDPs with discount factors close to one, making them computationally challenging. We propose a novel approximation strategy that "freezes" slow states during phases of lower-level planning and subsequently applies value iteration to an auxiliary upper-level MDP that evolves on a slower timescale. Freezing states for short periods of time leads to easier-to-solve lower-level problems, while a slower upper-level timescale allows for a more favorable discount factor. On the theoretical side, we analyze the regret incurred by our frozen-state approach, which leads to simple insights on how to trade off regret versus computational cost. Empirically, we benchmark our new frozen-state methods on three domains, (i) inventory control with fixed order costs, (ii) a gridworld problem with spatial tasks, and (iii) dynamic pricing with reference-price effects. We demonstrate that the new methods produce high-quality policies with significantly less computation, and we show that simply omitting slow states is often a poor heuristic.
Authors: Mahdi Ghafourian, Bilgesu Sumer, Ruben Vera-Rodriguez, Julian Fierrez, Ruben Tolosana, Aythami Moralez, Els Kindt
Abstract: Biometric recognition as an efficient and hard-to-forge way of identification and verification has become an indispensable part of the current digital world. The fast evolution of this technology has been a strong incentive for integration into many applications. Meanwhile, blockchain, the decentralized ledger technology, has been widely received by both research and industry in the past few years, and it is being increasingly deployed today in many different applications, such as money transfer, IoT, healthcare, or logistics. Recently, researchers have started to speculate on the pros and cons and what the best applications would be when these two technologies cross paths. This paper provides a survey of the research literature on the combination of blockchain and biometrics and includes a first legal analysis of this integration based on GDPR to shed light on challenges and potentials. Although the integration of blockchain technology into the biometric sector is still in its infancy, with a growing body of literature discussing specific applications and advanced technological setups, this paper aims to provide a holistic understanding of blockchain applicability in biometrics. Based on published studies, this article discusses, among others, practical examples combining blockchain and biometrics for novel applications in PKI systems, distributed trusted services, and identity management. Challenges and limitations when combining blockchain and biometrics that motivate future work will also be discussed; e.g., blockchain networks at their current stage may not be efficient or economical for some real-time biometric applications. Finally, we also discuss key legal aspects of the EU General Data Protection Regulation (GDPR) related to this combination of technologies (blockchain and biometrics); for example, accountability, immutability, anonymity, and data protection elements.
Authors: Rishabh Dixit, Mert Gurbuzbalaban, Waheed U. Bajwa
Abstract: This paper considers the problem of understanding the behavior of a general class of accelerated gradient methods on smooth nonconvex functions. Motivated by some recent works that have proposed effective algorithms, based on Polyak's heavy ball method and the Nesterov accelerated gradient method, to achieve convergence to a local minimum of nonconvex functions, this work proposes a broad class of Nesterov-type accelerated methods and puts forth a rigorous study of these methods encompassing the escape from saddle points and convergence to local minima through both an asymptotic and a non-asymptotic analysis. In the asymptotic regime, this paper answers an open question of whether Nesterov's accelerated gradient method (NAG) with variable momentum parameter avoids strict saddle points almost surely. This work also develops two metrics of asymptotic rates of convergence and divergence, and evaluates these two metrics for several popular standard accelerated methods such as the NAG and Nesterov's accelerated gradient with constant momentum (NCM) near strict saddle points. In the non-asymptotic regime, this work provides an analysis that leads to the "linear" exit time estimates from strict saddle neighborhoods for trajectories of these accelerated methods as well the necessary conditions for the existence of such trajectories. Finally, this work studies a sub-class of accelerated methods that can converge in convex neighborhoods of nonconvex functions with a near optimal rate to a local minimum and at the same time this sub-class offers superior saddle-escape behavior compared to that of NAG.
Authors: Fredy Vides, Idelfonso B. R. Nogueira, Gabriela Lopez Gutierrez, Lendy Banegas, Evelyn Flores
Abstract: In this document, we present key findings in structured matrix approximation theory, with applications to the regressive representation of dynamic financial processes. Initially, we explore a comprehensive approach involving generic nonlinear time delay embedding for time series data extracted from a financial or economic system under examination. Subsequently, we employ sparse least-squares and structured matrix approximation methods to discern approximate representations of the output coupling matrices. These representations play a pivotal role in establishing the regressive models corresponding to the recursive structures inherent in a given financial system. The document further introduces prototypical algorithms that leverage the aforementioned techniques. These algorithms are demonstrated through applications in approximate identification and predictive simulation of dynamic financial and economic processes, encompassing scenarios that may or may not exhibit chaotic behavior.
Authors: Marcos Matabuena, Juan C. Vidal, Oscar Hernan Madrid Padilla, Jukka-Pekka Onnela
Abstract: We introduce a novel \textit{k}-nearest neighbor (\textit{k}-NN) regression method for joint estimation of the conditional mean and variance. The proposed algorithm preserves the computational efficiency and manifold-learning capabilities of classical non-parametric \textit{k}-NN models, while integrating a data-driven variable selection step that improves empirical performance. By accurately estimating both conditional mean and variance regression functions, the method effectively reconstructs the conditional distribution and density functions for multiple families of scale-and-localization generative models. We show that our estimator can achieve fast convergence rates, and we derive practical rules for selecting the smoothing parameter~$k$ that enhance the precision of the algorithm in finite sample regimes. Extensive simulations for low, moderate and large-dimensional covariate spaces, together with a real-world biomedical application, demonstrate that the proposed method can consistently outperform the conventional \textit{k-NN} regression algorithm while being more interpretable in the model output.
Authors: Yu-Neng Chuang, Guanchu Wang, Chia-Yuan Chang, Ruixiang Tang, Shaochen Zhong, Fan Yang, Mengnan Du, Xuanting Cai, Vladimir Braverman, Xia Hu
Abstract: Large language models (LLMs) increasingly produce natural language explanations, yet these explanations often lack faithfulness, and they do not reliably reflect the evidence the model uses to decide. We introduce FaithLM, a model-agnostic framework that evaluates and improves the faithfulness of LLM explanations without token masking or task-specific heuristics. FaithLM formalizes explanation faithfulness as an intervention property: a faithful explanation should yield a prediction shift when its content is contradicted. Theoretical analysis shows that the resulting contrary-hint score is a sound and discriminative estimator of faithfulness. Building on this principle, FaithLM iteratively refines both the elicitation prompt and the explanation to maximize the measured score. Experiments on three multi-domain datasets and multiple LLM backbones demonstrate that FaithLM consistently increases faithfulness and produces explanations more aligned with human rationales than strong self-explanation baselines. These findings highlight that intervention-based evaluation, coupled with iterative optimization, provides a principled route toward faithful and reliable LLM explanations.
Authors: Payam Abdisarabshali, Kwang Taik Kim, Michael Langberg, Weifeng Su, Seyyedali Hosseinalipour
Abstract: Existing studies on federated learning (FL) are mostly focused on system orchestration for static snapshots of the network and making static control decisions (e.g., spectrum allocation). However, real-world wireless networks are susceptible to temporal variations of wireless channel capacity and users' datasets. In this paper, we incorporate multi-granular system dynamics (MSDs) into FL, including (M1) dynamic wireless channel capacity, captured by a set of discrete-time events, called $\mathscr{D}$-Events, and (M2) dynamic datasets of users. The latter is characterized by (M2-a) modeling the dynamics of user's dataset size via an ordinary differential equation and (M2-b) introducing dynamic model drift}, formulated via a partial differential inequality} drawing concrete analytical connections between the dynamics of users' datasets and FL accuracy. We then conduct FL orchestration under MSDs by introducing dynamic cooperative FL with dedicated MAC schedulers (DCLM), exploiting the unique features of open radio access network (O-RAN). DCLM proposes (i) a hierarchical device-to-device (D2D)-assisted model training, (ii) dynamic control decisions through dedicated O-RAN MAC schedulers, and (iii) asymmetric user selection. We provide extensive theoretical analysis to study the convergence of DCLM. We then optimize the degrees of freedom (e.g., user selection and spectrum allocation) in DCLM through a highly non-convex optimization problem. We develop a systematic approach to obtain the solution for this problem, opening the door to solving a broad variety of network-aware FL optimization problems. We show the efficiency of DCLM via numerical simulations and provide a series of future directions.
Authors: Jeeva Keshav Sattianarayanin, Anil Kumar Yerrapragada, Radha Krishna Ganti
Abstract: Accurate decoding of Uplink Control Information (UCI) on the Physical Uplink Control Channel (PUCCH) is essential for enabling 5G wireless links. This paper explores an AI/ML-based receiver design for PUCCH Format 0. Format 0 signaling encodes the UCI content within the phase of a known base waveform and even supports multiplexing of up to 12 users within the same time-frequency resources. The proposed neural network classifier, which we term UCINet0, is capable of predicting when no user is transmitting on the PUCCH, as well as decoding the UCI content for any number of multiplexed users (up to 12). The test results with simulated, hardware-captured (lab) and field datasets show that the UCINet0 model outperforms conventional correlation-based decoders across all SNR ranges and multiple fading scenarios.
Authors: William Roy Orchard, Nastaran Okati, Sergio Hernan Garrido Mejia, Patrick Bl\"obaum, Dominik Janzing
Abstract: The goal of Root Cause Analysis (RCA) is to explain why an anomaly occurred by identifying where the fault originated. Several recent works model the anomalous event as resulting from a change in the causal mechanism at the root cause, i.e., as a soft intervention. RCA is then the task of identifying which causal mechanism changed. In real-world applications, one often has either few or only a single sample from the post-intervention distribution: a severe limitation for most methods, which assume one knows or can estimate the distribution. However, even those that do not are statistically ill-posed due to the need to probe regression models in regions of low probability density. In this paper, we propose simple, efficient methods to overcome both difficulties in the case where there is a single root cause and the causal graph is a polytree. When one knows the causal graph, we give guarantees for a traversal algorithm that requires only marginal anomaly scores and does not depend on specifying an arbitrary anomaly score cut-off. When one does not know the causal graph, we show that the heuristic of identifying root causes as the variables with the highest marginal anomaly scores is causally justified. To this end, we prove that anomalies with small scores are unlikely to cause those with larger scores in polytrees and give upper bounds for the likelihood of causal pathways with non-monotonic anomaly scores.
Authors: Sandeep Kumar, Tirthankar Ghosal, Vinayak Goyal, Asif Ekbal
Abstract: The widespread adoption of Large Language Models (LLMs) and publicly available ChatGPT have marked a significant turning point in the integration of Artificial Intelligence (AI) into people's everyday lives. This study examines the ability of Large Language Models (LLMs) to generate future research ideas from scientific papers. Unlike tasks such as summarization or translation, idea generation lacks a clearly defined reference set or structure, making manual evaluation the default standard. However, human evaluation in this setting is extremely challenging ie: it requires substantial domain expertise, contextual understanding of the paper, and awareness of the current research landscape. This makes it time-consuming, costly, and fundamentally non-scalable, particularly as new LLMs are being released at a rapid pace. Currently, there is no automated evaluation metric specifically designed for this task. To address this gap, we propose two automated evaluation metrics: Idea Alignment Score (IAScore) and Idea Distinctness Index. We further conducted human evaluation to assess the novelty, relevance, and feasibility of the generated future research ideas. This investigation offers insights into the evolving role of LLMs in idea generation, highlighting both its capability and limitations. Our work contributes to the ongoing efforts in evaluating and utilizing language models for generating future research ideas. We make our datasets and codes publicly available
Authors: Olive Franzese, Ali Shahin Shamsabadi, Carter Luck, Hamed Haddadi
Abstract: The black-box service model enables ML service providers to serve clients while keeping their intellectual property and client data confidential. Confidentiality is critical for delivering ML services legally and responsibly, but makes it difficult for outside parties to verify important model properties such as fairness. Existing methods that assess model fairness confidentially lack either (i) reliability because they certify fairness with respect to a static set of data, and therefore fail to guarantee fairness in the presence of distribution shift or service provider malfeasance; and/or (ii) scalability due to the computational overhead of confidentiality-preserving cryptographic primitives. We address these problems by introducing online fairness certificates, which verify that a model is fair with respect to data received by the service provider online during deployment. We then present OATH, a deployably efficient and scalable zero-knowledge proof protocol for confidential online group fairness certification. OATH exploits statistical properties of group fairness via a cut-and-choose style protocol, enabling scalability improvements over baselines.
Authors: Haiyun He, Yepeng Liu, Ziqiao Wang, Yongyi Mao, Yuheng Bu
Abstract: Watermarking has emerged as a crucial method to distinguish AI-generated text from human-created text. Current watermarking approaches often lack formal optimality guarantees or address the scheme and detector design separately. In this paper, we introduce a novel, unified theoretical framework for watermarking Large Language Models (LLMs) that jointly optimizes both the watermarking scheme and detector. Our approach aims to maximize detection performance while maintaining control over the worst-case false positive rate (FPR) and distortion on text quality. We derive closed-form optimal solutions for this joint design and characterize the fundamental trade-off between watermark detectability and distortion. Notably, we reveal that the optimal watermarking schemes should be adaptive to the LLM's generative distribution. Building on our theoretical insights, we propose a distortion-free, distribution-adaptive watermarking algorithm (DAWA) that leverages a surrogate model for model-agnosticism and efficiency. Experiments on Llama2-13B and Mistral-8$\times$7B models confirm the effectiveness of our approach, particularly at ultra-low FPRs. Our code is available at https://github.com/yepengliu/DAWA.
Authors: Hrishav Bakul Barua, Kalin Stefanov, Lemuel Lai En Che, Abhinav Dhall, KokSheik Wong, Ganesh Krishnasamy
Abstract: Reconstruction of High Dynamic Range (HDR) from Low Dynamic Range (LDR) images is an important computer vision task. There is a significant amount of research utilizing both conventional non-learning methods and modern data-driven approaches, focusing on using both single-exposed and multi-exposed LDR for HDR image reconstruction. However, most current state-of-the-art methods require high-quality paired {LDR;HDR} datasets with limited literature use of unpaired datasets, that is, methods that learn the LDR-HDR mapping between domains. This paper proposes CycleHDR, a method that integrates self-supervision into a modified semantic- and cycle-consistent adversarial architecture that utilizes unpaired LDR and HDR datasets for training. Our method introduces novel artifact- and exposure-aware generators to address visual artifact removal. It also puts forward an encoder and loss to address semantic consistency, another under-explored topic. CycleHDR is the first to use semantic and contextual awareness for the LDR-HDR reconstruction task in a self-supervised setup. The method achieves state-of-the-art performance across several benchmark datasets and reconstructs high-quality HDR images. The official website of this work is available at: https://github.com/HrishavBakulBarua/Cycle-HDR
Authors: Kelvin J. L. Koa, Yunshan Ma, Yi Xu, Ritchie Ng, Huanhuan Zheng, Tat-Seng Chua
Abstract: Stock portfolios are often exposed to rare consequential events (e.g., 2007 global financial crisis, 2020 COVID-19 stock market crash), as they do not have enough historical information to learn from. Large Language Models (LLMs) now present a possible tool to tackle this problem, as they can generalize across their large corpus of training data and perform zero-shot reasoning on new events, allowing them to detect possible portfolio crash events without requiring specific training data. However, detecting portfolio crashes is a complex problem that requires more than reasoning abilities. Investors need to dynamically process the impact of each new piece of information found in news articles, analyze the relational network of impacts across different events and portfolio stocks, as well as understand the temporal context between impacts across time-steps, in order to obtain the aggregated impact on the target portfolio. In this work, we propose an algorithmic framework named Temporal Relational Reasoning (TRR). It seeks to emulate the spectrum of human cognitive capabilities used for complex problem-solving, which include brainstorming, memory, attention and reasoning. Through extensive experiments, we show that TRR is able to outperform state-of-the-art techniques on detecting stock portfolio crashes, and demonstrate how each of the proposed components help to contribute to its performance through an ablation study. Additionally, we further explore the possible applications of TRR by extending it to other related complex problems, such as the detection of possible global crisis events in Macroeconomics.
Authors: Yuwei Du, Jie Feng, Jie Zhao, Yong Li
Abstract: Trajectory modeling, which includes research on trajectory data pattern mining and future prediction, has widespread applications in areas such as life services, urban transportation, and public administration. Numerous methods have been proposed to address specific problems within trajectory modeling. However, the heterogeneity of data and the diversity of trajectory tasks make effective and reliable trajectory modeling an important yet highly challenging endeavor, even for domain experts. \fix In this paper, we propose \textit{TrajAgent}, a agent framework powered by large language models (LLMs), designed to facilitate robust and efficient trajectory modeling through automation modeling. This framework leverages and optimizes diverse specialized models to address various trajectory modeling tasks across different datasets effectively. \unfix~In \textit{TrajAgent}, we first develop \textit{UniEnv}, an execution environment with a unified data and model interface, to support the execution and training of various models. Building on \textit{UniEnv}, we introduce an agentic workflow designed for automatic trajectory modeling across various trajectory tasks and data. Furthermore, we introduce collaborative learning schema between LLM-based agents and small speciallized models, to enhance the performance of the whole framework effectively. Extensive experiments on four tasks using four real-world datasets demonstrate the effectiveness of \textit{TrajAgent} in automated trajectory modeling, achieving a performance improvement of \fix 2.38\%-69.91\% \unfix over baseline methods. The codes and data can be accessed via https://github.com/tsinghua-fib-lab/TrajAgent.
Authors: Marton Szep, Daniel Rueckert, R\"udiger von Eisenhart-Rothe, Florian Hinterwimmer
Abstract: Fine-tuning large language models (LLMs) with limited data poses a practical challenge in low-resource languages, specialized domains, and constrained deployment settings. While pre-trained LLMs provide strong foundations, effective adaptation under data scarcity requires focused and efficient fine-tuning techniques. This paper presents a structured and practical survey of recent methods for fine-tuning LLMs in data-scarce scenarios. We systematically review parameter-efficient fine-tuning techniques that lower training and deployment costs, domain and cross-lingual adaptation methods for both encoder and decoder models, and model specialization strategies. We further examine preference alignment approaches that guide model behavior using limited human or synthetic feedback, emphasizing sample and compute efficiency. Throughout, we highlight empirical trade-offs, selection criteria, and best practices for choosing suitable techniques based on task constraints, including model scaling, data scaling, and the mitigation of catastrophic forgetting. The aim is to equip researchers and practitioners with actionable insights for effectively fine-tuning LLMs when data and resources are limited.
Authors: Guillaume Principato, Gilles Stoltz, Yvenn Amara-Ouali, Yannig Goude, Bachir Hamrouche, Jean-Michel Poggi
Abstract: We consider conformal prediction for multivariate data and focus on hierarchical data, where some components are linear combinations of others. Intuitively, the hierarchical structure can be leveraged to reduce the size of prediction regions for the same coverage level. We implement this intuition by including a projection step (also called a reconciliation step) in the split conformal prediction [SCP] procedure, and prove that the resulting prediction regions are indeed globally smaller. We do so both under the classic objective of joint coverage and under a new and challenging task: component-wise coverage, for which efficiency results are more difficult to obtain. The associated strategies and their analyses are based both on the literature of SCP and of forecast reconciliation, which we connect. We also illustrate the theoretical findings, for different scales of hierarchies on simulated data.
Authors: Ayush Mohanty, Jason Dekarske, Stephen K. Robinson, Sanjay Joshi, Nagi Gebraeel
Abstract: Robotic manipulators are critical in many applications but are known to degrade over time. This degradation is influenced by the nature of the tasks performed by the robot. Tasks with higher severity, such as handling heavy payloads, can accelerate the degradation process. One way this degradation is reflected is in the position accuracy of the robot's end-effector. In this paper, we present a prognostic modeling framework that predicts a robotic manipulator's Remaining Useful Life (RUL) while accounting for the effects of task severity. Our framework represents the robot's position accuracy as a Brownian motion process with a random drift parameter that is influenced by task severity. The dynamic nature of task severity is modeled using a continuous-time Markov chain (CTMC). To evaluate RUL, we discuss two approaches -- (1) a novel closed-form expression for Remaining Lifetime Distribution (RLD), and (2) Monte Carlo simulations, commonly used in prognostics literature. Theoretical results establish the equivalence between these RUL computation approaches. We validate our framework through experiments using two distinct physics-based simulators for planar and spatial robot fleets. Our findings show that robots in both fleets experience shorter RUL when handling a higher proportion of high-severity tasks.
Authors: Philipp Reiser, Paul-Christian B\"urkner, Anneli Guthke
Abstract: Surrogate models are often used as computationally efficient approximations to complex simulation models, enabling tasks such as solving inverse problems, sensitivity analysis, and probabilistic forward predictions, which would otherwise be computationally infeasible. During training, surrogate parameters are fitted such that the surrogate reproduces the simulation model's outputs as closely as possible. However, the simulation model itself is merely a simplification of the real-world system, often missing relevant processes or suffering from misspecifications e.g., in inputs or boundary conditions. Hints about these might be captured in real-world measurement data, and yet, we typically ignore those hints during surrogate building. In this paper, we propose two novel probabilistic approaches to integrate simulation data and real-world measurement data during surrogate training. The first method trains separate surrogate models for each data source and combines their predictive distributions, while the second incorporates both data sources by training a single surrogate. We show the conceptual differences and benefits of the two approaches through both synthetic and real-world case studies. The results demonstrate the potential of these methods to improve predictive accuracy, predictive coverage, and to diagnose problems in the underlying simulation model. These insights can improve system understanding and future model development.
Authors: Mohamed Ndaoud, Peter Radchenko, Bradley Rava
Abstract: Selective classification is a powerful tool for automated decision-making in high-risk scenarios, allowing classifiers to act only when confident and abstain when uncertainty is high. Given a target accuracy, our goal is to minimize indecisions, observations we do not automate. For difficult problems, the target accuracy may be unattainable without abstention. By using indecisions, we can control the misclassification rate to any user-specified level, even below the Bayes optimal error rate, while minimizing overall indecision mass. We provide a complete characterization of the minimax risk in selective classification, establishing continuity and monotonicity properties that enable optimal indecision selection. We revisit selective inference via the Neyman-Pearson testing framework, where indecision enables control of type 2 error given fixed type 1 error probability. For both classification and testing, we propose a finite-sample calibration method with non-asymptotic guarantees, proving plug-in classifiers remain consistent and that accuracy-based calibration effectively controls indecision mass. In the binary Gaussian mixture model, we uncover the first sharp phase transition in selective inference, showing minimal indecision can yield near-optimal accuracy even under poor class separation. Experiments on Gaussian mixtures and real datasets confirm that small indecision proportions yield substantial accuracy gains, making indecision a principled tool for risk control.
Authors: Gregory Kang Ruey Lau, Wenyang Hu, Diwen Liu, Jizhuo Chen, See-Kiong Ng, Bryan Kian Hsiang Low
Abstract: Large Language Models (LLMs), particularly smaller variants, still struggle with complex reasoning tasks. While inference-time prompting can guide reasoning, existing methods often rely on sequential queries. Ensemble approaches offer a promising path to performance gains, especially given recent batch inference speed-ups. This work introduces DIPPER, a novel, training-free framework that transforms a single LLM into an effective inference-time ensemble. By feeding the model an optimized and diverse set of prompts in parallel, DIPPER elicits varied reasoning paths, leading to performance gains. We empirically demonstrate significant improvements on reasoning benchmarks, such as MATH, where a DIPPER ensemble of three Qwen2-MATH-1.5B instances (via parallel prompting of a single model) outperforms a larger 7B model.
Authors: King-kui Sin, Xi Xuan, Chunyu Kit, Clara Ho-yan Chan, Honic Ho-kin Ip
Abstract: This paper addresses the challenges translating case law under Hong Kong's bilingual legal system. It highlights the initial success of translating all written statutes into Chinese before the 1997 handover, a task mandated by the Basic Law. The effort involved significant collaboration among legal, linguistic, and translation experts, resulting in a comprehensive and culturally appropriate bilingual legal system. However, translating case law remains a significant challenge due to the sheer volume and continuous growth of judicial decisions. The paper critiques the governments and judiciarys sporadic and uncoordinated efforts to translate case law, contrasting it with the thorough approach previously taken for statute translation. Although the government acknowledges the importance of legal bilingualism, it lacks a sustainable strategy for translating case law. The Judiciarys position that translating all judgments is unnecessary, unrealistic, and not cost-effectiveis analyzed and critiqued for its impact on legal transparency and public trust. A proposed solution involves leveraging machine translation technology through a human-machine interactive translation platform, which undergoes two major transitions. Initially based on a neural model, the platform transitions to using a large language model for improved translation accuracy. Furthermore, it evolves from a single-agent system to a multi-agent system, incorporating Translator, Annotator, and Proofreader agents. This multi-agent approach, supported by a grant, aims to facilitate efficient, high-quality translation of judicial judgments by integrating advanced artificial intelligence and continuous feedback mechanisms, thus better meeting the needs of a bilingual legal system.
Authors: Ruijia Liu, Ancheng Hou, Xiao Yu, Xiang Yin
Abstract: Signal Temporal Logic (STL) is a powerful specification language for describing complex temporal behaviors of continuous signals, making it well-suited for high-level robotic task descriptions. However, generating executable plans for STL tasks is challenging, as it requires consideration of the coupling between the task specification and the system dynamics. Existing approaches either follow a model-based setting that explicitly requires knowledge of the system dynamics or adopt a task-oriented data-driven approach to learn plans for specific tasks. In this work, we address the problem of generating executable STL plans for systems with unknown dynamics. We propose a hierarchical planning framework that enables zero-shot generalization to new STL tasks by leveraging only task-agnostic trajectory data during offline training. The framework consists of three key components: (i) decomposing the STL specification into several progresses and time constraints, (ii) searching for timed waypoints that satisfy all progresses under time constraints, and (iii) generating trajectory segments using a pre-trained diffusion model and stitching them into complete trajectories. We formally prove that our method guarantees STL satisfaction, and simulation results demonstrate its effectiveness in generating dynamically feasible trajectories across diverse long-horizon STL tasks.
Authors: Jie Liu, Gongye Liu, Jiajun Liang, Ziyang Yuan, Xiaokun Liu, Mingwu Zheng, Xiele Wu, Qiulin Wang, Menghan Xia, Xintao Wang, Xiaohong Liu, Fei Yang, Pengfei Wan, Di Zhang, Kun Gai, Yujiu Yang, Wanli Ouyang
Abstract: Video generation has achieved significant advances through rectified flow techniques, but issues like unsmooth motion and misalignment between videos and prompts persist. In this work, we develop a systematic pipeline that harnesses human feedback to mitigate these problems and refine the video generation model. Specifically, we begin by constructing a large-scale human preference dataset focused on modern video generation models, incorporating pairwise annotations across multi-dimensions. We then introduce VideoReward, a multi-dimensional video reward model, and examine how annotations and various design choices impact its rewarding efficacy. From a unified reinforcement learning perspective aimed at maximizing reward with KL regularization, we introduce three alignment algorithms for flow-based models. These include two training-time strategies: direct preference optimization for flow (Flow-DPO) and reward weighted regression for flow (Flow-RWR), and an inference-time technique, Flow-NRG, which applies reward guidance directly to noisy videos. Experimental results indicate that VideoReward significantly outperforms existing reward models, and Flow-DPO demonstrates superior performance compared to both Flow-RWR and supervised fine-tuning methods. Additionally, Flow-NRG lets users assign custom weights to multiple objectives during inference, meeting personalized video quality needs.
Authors: Md Kaykobad Reza, Ameya Patil, Mashhour Solh, M. Salman Asif
Abstract: Multimodal models often experience a significant performance drop when one or more modalities are missing during inference. To address this challenge, we propose a simple yet effective approach that enhances robustness to missing modalities while maintaining strong performance when all modalities are available. Our method introduces cross-modal proxy tokens (CMPTs), which approximate the class token of a missing modality by attending only to the tokens of the available modality without requiring explicit modality generation or auxiliary networks. To efficiently learn these approximations with minimal computational overhead, we employ low-rank adapters in frozen unimodal encoders and jointly optimize an alignment loss with a task-specific loss. Extensive experiments on five multimodal datasets show that our method outperforms state-of-the-art baselines across various missing rates while achieving competitive results in complete-modality settings. Overall, our method offers a flexible and efficient solution for robust multimodal learning. The code for this paper is available at: https://github.com/CSIPlab/Cross-Modal-Proxy-Tokens.
Authors: Qingyue Yang, Jie Wang, Xing Li, Zhihai Wang, Chen Chen, Lei Chen, Xianzhi Yu, Wulong Liu, Jianye Hao, Mingxuan Yuan, Bin Li
Abstract: With the development of large language models (LLMs), efficient inference through Key-Value (KV) cache compression has attracted considerable attention, especially for long-context generation. To compress the KV cache, recent methods identify critical KV tokens through static modeling of attention scores. However, these methods often struggle to accurately determine critical tokens as they neglect the temporal patterns in attention scores, resulting in a noticeable degradation in LLM performance. To address this challenge, we propose AttentionPredictor, which is the first learning-based method to directly predict attention patterns for KV cache compression and critical token identification. Specifically, AttentionPredictor learns a lightweight, unified convolution model to dynamically capture spatiotemporal patterns and predict the next-token attention scores. An appealing feature of AttentionPredictor is that it accurately predicts the attention score and shares the unified prediction model, which consumes negligible memory, among all transformer layers. Moreover, we propose a cross-token critical cache prefetching framework that hides the token estimation time overhead to accelerate the decoding stage. By retaining most of the attention information, AttentionPredictor achieves 13$\times$ KV cache compression and 5.6$\times$ speedup in a cache offloading scenario with comparable LLM performance, significantly outperforming the state-of-the-arts. The code is available at https://github.com/MIRALab-USTC/LLM-AttentionPredictor.
URLs: https://github.com/MIRALab-USTC/LLM-AttentionPredictor.
Authors: Oliver Biggar, Christos Papadimitriou
Abstract: Characterizing the limit behavior -- that is, the attractors -- of learning dynamics is one of the most fundamental open questions in game theory. In recent work on this front, it was conjectured that the attractors of the replicator dynamic are in one-to-one correspondence with the sink equilibria of the game -- the sink strongly connected components of a game's preference graph -- , and it was established that they do stand in at least one-to-many correspondence with them. Here, we show that the one-to-one conjecture is false. We disprove this conjecture over the course of three theorems: the first disproves a stronger form of the conjecture, while the weaker form is disproved separately in the two-player and $N$-player ($N>2$) cases. By showing how the conjecture fails, we lay out the obstacles that lie ahead for characterizing attractors of the replicator, and introduce new ideas with which to tackle them. All three counterexamples derive from an object called a local source -- a point lying within the sink equilibrium, and yet which is `locally repelling'; we prove that the absence of local sources is necessary, but not sufficient, for the one-to-one property to be true. We complement this with a sufficient condition: we introduce a local property of a sink equilibrium called pseudoconvexity, and establish that when the sink equilibria of a two-player game are pseudoconvex then they precisely define the attractors. Pseudoconvexity generalizes the previous cases -- such as zero-sum games and potential games -- where this conjecture was known to hold, and reformulates these cases in terms of a simple graph property.
Authors: Andrej Leban
Abstract: The Distributional Principal Autoencoder (DPA) combines distributionally correct reconstruction with principal-component-like interpretability of the encodings. In this work, we provide exact theoretical guarantees on both fronts. First, we derive a closed-form relation linking each optimal level-set geometry to the data-distribution score. This result explains DPA's empirical ability to disentangle factors of variation of the data, as well as allows the score to be recovered directly from samples. When the data follows the Boltzmann distribution, we demonstrate that this relation yields an approximation of the minimum free-energy path for the Mueller-Brown potential in a single fit. Second, we prove that if the data lies on a manifold that can be approximated by the encoder, latent components beyond the manifold dimension are conditionally independent of the data distribution - carrying no additional information - and thus reveal the intrinsic dimension. Together, these results show that a single model can learn the data distribution and its intrinsic dimension with exact guarantees simultaneously, unifying two longstanding goals of unsupervised learning.
Authors: Kayhan Behdin, Ata Fatahibaarzi, Qingquan Song, Yun Dai, Aman Gupta, Zhipeng Wang, Shao Tang, Hejian Sang, Gregory Dexter, Sirou Zhu, Siyu Zhu, Tejas Dharamsi, Vignesh Kothapalli, Zhoutong Fu, Yihan Cao, Pin-Lun Hsu, Fedor Borisyuk, Natesh Pillai, Luke Simon, Rahul Mazumder
Abstract: Large language models (LLMs) have demonstrated remarkable performance across a wide range of industrial applications, from search and recommendation systems to generative tasks. Although scaling laws indicate that larger models generally yield better generalization and performance, their substantial computational requirements often render them impractical for many real-world scenarios at scale. In this paper, we present a comprehensive set of insights for training and deploying small language models (SLMs) that deliver high performance for a variety of industry use cases. We focus on two key techniques: (1) knowledge distillation and (2) model compression via structured pruning and quantization. These approaches enable SLMs to retain much of the quality of their larger counterparts while significantly reducing training/serving costs and latency. We detail the impact of these techniques on a variety of use cases in a large professional social network platform and share deployment lessons, including hardware optimization strategies that improve speed and throughput for both predictive and reasoning-based applications in Recommendation Systems.
Authors: Yihong Dong, Ge Li, Xue Jiang, Yongding Tao, Kechi Zhang, Hao Zhu, Huanyu Liu, Jiazheng Ding, Jia Li, Jinliang Deng, Hong Mei
Abstract: Periodicity, as one of the most important basic characteristics, lays the foundation for facilitating structured knowledge acquisition and systematic cognitive processes within human learning paradigms. However, the potential flaws of periodicity modeling in Transformer affect the learning efficiency and establishment of underlying principles from data for large language models (LLMs) built upon it. In this paper, we demonstrate that integrating effective periodicity modeling can improve the learning efficiency and performance of LLMs. We introduce FANformer, which adapts Fourier Analysis Network (FAN) into attention mechanism to achieve efficient periodicity modeling, by modifying the feature projection process of attention mechanism. Extensive experimental results on language modeling show that FANformer consistently outperforms Transformer when scaling up model size and training tokens, underscoring its superior learning efficiency. Our pretrained FANformer-1B exhibits marked improvements on downstream tasks compared to open-source LLMs with similar model parameters or training tokens. Moreover, we reveal that FANformer exhibits superior ability to learn and apply rules for reasoning compared to Transformer. The results position FANformer as an effective and promising architecture for advancing LLMs.
Authors: Shai Bergman, Anne-Marie Kermarrec, Diana Petrescu, Rafael Pires, Mathis Randl, Martijn de Vos, Ji Zhang
Abstract: Retrieval-augmented generation (RAG) improves the reliability of large language model (LLM) answers by integrating external knowledge. However, RAG increases the end-to-end inference time since looking for relevant documents from large vector databases is computationally expensive. To address this, we introduce Proximity, an approximate key-value cache that optimizes the RAG workflow by leveraging similarities in user queries. Instead of treating each query independently, Proximity reuses previously retrieved documents when similar queries appear, substantially reducing the reliance on expensive vector database lookups. To efficiently scale, Proximity employs a locality-sensitive hashing (LSH) scheme that enables fast cache lookups while preserving retrieval accuracy. We evaluate Proximity using the MMLU and MedRAG question-answering benchmarks. Our experiments demonstrate that Proximity with our LSH scheme and a realistically-skewed MedRAG workload reduces database calls by 77.2% while maintaining database recall and test accuracy. We experiment with different similarity tolerances and cache capacities, and show that the time spent within the Proximity cache remains low and constant (4.8 microseconds) even as the cache grows substantially in size. Our results demonstrate that approximate caching is a practical and effective strategy for optimizing RAG-based systems.
Authors: Nils Philipp Walter, Jilles Vreeken, Jonas Fischer
Abstract: Neural networks are regularly employed in high-stakes decision-making, where understanding and transparency is key. Attribution methods have been developed to gain understanding into which input features neural networks use for a specific prediction. Although widely used in computer vision, these methods often result in unspecific saliency maps that fail to identify the relevant information that led to a decision, supported by different benchmarks results. Here, we revisit the common attribution pipeline and identify one cause for the lack of specificity in attributions as the computation of attribution of isolated logits. Instead, we suggest to combine attributions of multiple class logits in analogy to how the softmax combines the information across logits. By computing probability distributions of attributions over classes for each spatial location in the image, we unleash the true capabilities of existing attribution methods, revealing better object- and instance-specificity and uncovering discriminative as well as shared features between classes. On common benchmarks, including the grid-pointing game and randomization-based sanity checks, we show that this reconsideration of how and where we compute attributions across the network improves established attribution methods while staying agnostic to model architectures. We make the code publicly available: https://github.com/nilspwalter/var.
Authors: Zhaoyi Li, Xiaohan Zhao, Dong-Dong Wu, Jiacheng Cui, Zhiqiang Shen
Abstract: Despite promising performance on open-source large vision-language models (LVLMs), transfer-based targeted attacks often fail against closed-source commercial LVLMs. Analyzing failed adversarial perturbations reveals that the learned perturbations typically originate from a uniform distribution and lack clear semantic details, resulting in unintended responses. This critical absence of semantic information leads commercial black-box LVLMs to either ignore the perturbation entirely or misinterpret its embedded semantics, thereby causing the attack to fail. To overcome these issues, we propose to refine semantic clarity by encoding explicit semantic details within local regions, thus ensuring the capture of finer-grained features and inter-model transferability, and by concentrating modifications on semantically rich areas rather than applying them uniformly. To achieve this, we propose a simple yet highly effective baseline: at each optimization step, the adversarial image is cropped randomly by a controlled aspect ratio and scale, resized, and then aligned with the target image in the embedding space. While the naive source-target matching method has been utilized before in the literature, we are the first to provide a tight analysis, which establishes a close connection between perturbation optimization and semantics. Experimental results confirm our hypothesis. Our adversarial examples crafted with local-aggregated perturbations focused on crucial regions exhibit surprisingly good transferability to commercial LVLMs, including GPT-4.5, GPT-4o, Gemini-2.0-flash, Claude-3.5/3.7-sonnet, and even reasoning models like o1, Claude-3.7-thinking and Gemini-2.0-flash-thinking. Our approach achieves success rates exceeding 90% on GPT-4.5, 4o, and o1, significantly outperforming all prior state-of-the-art attack methods with lower $\ell_1/\ell_2$ perturbations.
Authors: Quanhao Li, Zhen Xing, Rui Wang, Hui Zhang, Qi Dai, Zuxuan Wu
Abstract: Recent advances in video generation have led to remarkable improvements in visual quality and temporal coherence. Upon this, trajectory-controllable video generation has emerged to enable precise object motion control through explicitly defined spatial paths. However, existing methods struggle with complex object movements and multi-object motion control, resulting in imprecise trajectory adherence, poor object consistency, and compromised visual quality. Furthermore, these methods only support trajectory control in a single format, limiting their applicability in diverse scenarios. Additionally, there is no publicly available dataset or benchmark specifically tailored for trajectory-controllable video generation, hindering robust training and systematic evaluation. To address these challenges, we introduce MagicMotion, a novel image-to-video generation framework that enables trajectory control through three levels of conditions from dense to sparse: masks, bounding boxes, and sparse boxes. Given an input image and trajectories, MagicMotion seamlessly animates objects along defined trajectories while maintaining object consistency and visual quality. Furthermore, we present MagicData, a large-scale trajectory-controlled video dataset, along with an automated pipeline for annotation and filtering. We also introduce MagicBench, a comprehensive benchmark that assesses both video quality and trajectory control accuracy across different numbers of objects. Extensive experiments demonstrate that MagicMotion outperforms previous methods across various metrics. Our project page are publicly available at https://quanhaol.github.io/magicmotion-site.
Authors: Chau Pham, Juan C. Caicedo, Bryan A. Plummer
Abstract: Prior work using Masked Autoencoders (MAEs) typically relies on random patch masking based on the assumption that images have significant redundancies across different channels, allowing for the reconstruction of masked content using cross-channel correlations. However, this assumption does not hold in Multi-Channel Imaging (MCI), where channels may provide complementary information with minimal feature overlap. Thus, these MAEs primarily learn local structures within individual channels from patch reconstruction, failing to fully leverage cross-channel interactions and limiting their MCI effectiveness. In this paper, we present ChA-MAEViT, an MAE-based method that enhances feature learning across MCI channels via four key strategies: (1) dynamic channel-patch masking, which compels the model to reconstruct missing channels in addition to masked patches, thereby enhancing cross-channel dependencies and improving robustness to varying channel configurations; (2) memory tokens, which serve as long-term memory aids to promote information sharing across channels, addressing the challenges of reconstructing structurally diverse channels; (3) hybrid token fusion module, which merges fine-grained patch tokens with a global class token to capture richer representations; and (4) Channel-Aware Decoder, a lightweight decoder utilizes channel tokens to effectively reconstruct image patches. Experiments on satellite and microscopy datasets, CHAMMI, JUMP-CP, and So2Sat, show that ChA-MAEViT significantly outperforms state-of-the-art MCI-ViTs by 3.0-21.5%, highlighting the importance of cross-channel interactions in MCI. Our code is publicly available at https://github.com/chaudatascience/cha_mae_vit.
Authors: Yunhong Min, Daehyeon Choi, Kyeongmin Yeo, Jihyun Lee, Minhyuk Sung
Abstract: We introduce ORIGEN, the first zero-shot method for 3D orientation grounding in text-to-image generation across multiple objects and diverse categories. While previous work on spatial grounding in image generation has mainly focused on 2D positioning, it lacks control over 3D orientation. To address this, we propose a reward-guided sampling approach using a pretrained discriminative model for 3D orientation estimation and a one-step text-to-image generative flow model. While gradient-ascent-based optimization is a natural choice for reward-based guidance, it struggles to maintain image realism. Instead, we adopt a sampling-based approach using Langevin dynamics, which extends gradient ascent by simply injecting random noise--requiring just a single additional line of code. Additionally, we introduce adaptive time rescaling based on the reward function to accelerate convergence. Our experiments show that ORIGEN outperforms both training-based and test-time guidance methods across quantitative metrics and user studies.
Authors: Belinda Z. Li, Been Kim, Zi Wang
Abstract: Large language models (LLMs) have shown impressive performance on reasoning benchmarks like math and logic. While many works have largely assumed well-defined tasks, real-world queries are often underspecified and only solvable by acquiring missing information. We formalize this information-gathering problem as a constraint satisfaction problem (CSP) with missing variable assignments. Using a special case where only one necessary variable assignment is missing, we can evaluate an LLM's ability to identify the minimal necessary question to ask. We present QuestBench, a set of underspecified reasoning tasks solvable by asking at most one question, which includes: (1) Logic-Q: logical reasoning tasks with one missing proposition, (2) Planning-Q: PDDL planning problems with partially-observed initial states, (3) GSM-Q: human-annotated grade school math problems with one unknown variable, and (4) GSME-Q: equation-based version of GSM-Q. The LLM must select the correct clarification question from multiple options. While current models excel at GSM-Q and GSME-Q, they achieve only 40-50% accuracy on Logic-Q and Planning-Q. Analysis shows that the ability to solve well-specified reasoning problems is not sufficient for success on our benchmark: models struggle to identify the right question even when they can solve the fully specified version. This highlights the need for specifically optimizing models' information acquisition capabilities.
Authors: Afra Amini, Tim Vieira, Ryan Cotterell
Abstract: Estimating the Kullback--Leibler (KL) divergence between language models has many applications, e.g., reinforcement learning from human feedback (RLHF), interpretability, and knowledge distillation. However, computing the exact KL divergence between two arbitrary language models is intractable. Thus, practitioners often resort to sampling-based estimators. While it is easy to fashion a simple Monte Carlo (MC) estimator that provides an unbiased estimate of the KL divergence between language models, this estimator notoriously suffers from high variance and can even result in a negative estimate of the KL divergence, a non-negative quantity. In this paper, we introduce a Rao--Blackwellized estimator that is unbiased and provably has variance less than or equal to that of the standard Monte Carlo estimator. In an empirical study on sentiment-controlled fine-tuning, we show that our estimator provides more stable KL estimates and reduces variance substantially. Additionally, we derive an analogous Rao--Blackwellized estimator of the gradient of the KL divergence, which leads to more stable training and produces models that more frequently appear on the Pareto frontier of reward vs. KL compared to the ones trained with the MC estimator of the gradient.
Authors: Mara Daniels
Abstract: We investigate stochastic interpolation, a recently introduced framework for high dimensional sampling which bears many similarities to diffusion modeling. Stochastic interpolation generates a data sample by first randomly initializing a particle drawn from a simple base distribution, then simulating deterministic or stochastic dynamics such that in finite time the particle's distribution converges to the target. We show that for a Gaussian base distribution and a strongly log-concave target distribution, the stochastic interpolation flow map is Lipschitz with a sharp constant which matches that of Caffarelli's theorem for optimal transport maps. We are further able to construct Lipschitz transport maps between non-Gaussian distributions, generalizing some recent constructions in the literature on transport methods for establishing functional inequalities. We discuss the practical implications of our theorem for the sampling and estimation problems required by stochastic interpolation.
Authors: Diego Hitzges, Suman Ghosh, Guillermo Gallego
Abstract: Event cameras offer a promising avenue for multi-view stereo depth estimation and Simultaneous Localization And Mapping (SLAM) due to their ability to detect blur-free 3D edges at high-speed and over broad illumination conditions. However, traditional deep learning frameworks designed for conventional cameras struggle with the asynchronous, stream-like nature of event data, as their architectures are optimized for discrete, image-like inputs. We propose a scalable, flexible and adaptable framework for pixel-wise depth estimation with event cameras in both monocular and stereo setups. The 3D scene structure is encoded into disparity space images (DSIs), representing spatial densities of rays obtained by back-projecting events into space via known camera poses. Our neural network processes local subregions of the DSIs combining 3D convolutions and a recurrent structure to recognize valuable patterns for depth prediction. Local processing enables fast inference with full parallelization and ensures constant ultra-low model complexity and memory costs, regardless of camera resolution. Experiments on standard benchmarks (MVSEC and DSEC datasets) demonstrate unprecedented effectiveness: (i) using purely monocular data, our method achieves comparable results to existing stereo methods; (ii) when applied to stereo data, it strongly outperforms all state-of-the-art (SOTA) approaches, reducing the mean absolute error by at least 42%; (iii) our method also allows for increases in depth completeness by more than 3-fold while still yielding a reduction in median absolute error of at least 30%. Given its remarkable performance and effective processing of event-data, our framework holds strong potential to become a standard approach for using deep learning for event-based depth estimation and SLAM. Project page: https://github.com/tub-rip/DERD-Net
Authors: Yinlong Dai, Robert Ramirez Sanchez, Ryan Jeronimus, Shahabedin Sagheb, Cara M. Nunez, Heramb Nemlekar, Dylan P. Losey
Abstract: Today's robots attempt to learn new tasks by imitating human examples. These robots watch the human complete the task, and then try to match the actions taken by the human expert. However, this standard approach to visual imitation learning is fundamentally limited: the robot observes what the human does, but not why the human chooses those behaviors. Without understanding which features of the system or environment factor into the human's decisions, robot learners often misinterpret the human's examples. In practice, this results in causal confusion, inefficient learning, and robot policies that fail when the environment changes. We therefore propose a shift in perspective: instead of asking human teachers just to show what actions the robot should take, we also enable humans to intuitively indicate why they made those decisions. Under our paradigm human teachers attach markers to task-relevant objects and use natural language prompts to describe their state representation. Our proposed algorithm, CIVIL, leverages this augmented demonstration data to filter the robot's visual observations and extract a feature representation that aligns with the human teacher. CIVIL then applies these causal features to train a transformer-based policy that -- when tested on the robot -- is able to emulate human behaviors without being confused by visual distractors or irrelevant items. Our simulations and real-world experiments demonstrate that robots trained with CIVIL learn both what actions to take and why to take those actions, resulting in better performance than state-of-the-art baselines. From the human's perspective, our user study reveals that this new training paradigm actually reduces the total time required for the robot to learn the task, and also improves the robot's performance in previously unseen scenarios. See videos at our project website: https://civil2025.github.io
Authors: Joshua Engels, David D. Baek, Subhash Kantamneni, Max Tegmark
Abstract: Scalable oversight, the process by which weaker AI systems supervise stronger ones, has been proposed as a key strategy to control future superintelligent systems. However, it is still unclear how scalable oversight itself scales. To address this gap, we propose a framework that quantifies the probability of successful oversight as a function of the capabilities of the overseer and the system being overseen. Specifically, our framework models oversight as a game between capability-mismatched players; the players have oversight-specific Elo scores that are a piecewise-linear function of their general intelligence, with two plateaus corresponding to task incompetence and task saturation. We validate our framework with a modified version of the game Nim and then apply it to four oversight games: Mafia, Debate, Backdoor Code and Wargames. For each game, we find scaling laws that approximate how domain performance depends on general AI system capability. We then build on our findings in a theoretical study of Nested Scalable Oversight (NSO), a process in which trusted models oversee untrusted stronger models, which then become the trusted models in the next step. We identify conditions under which NSO succeeds and derive numerically (and in some cases analytically) the optimal number of oversight levels to maximize the probability of oversight success. We also apply our theory to our four oversight games, where we find that NSO success rates at a general Elo gap of 400 are 13.5% for Mafia, 51.7% for Debate, 10.0% for Backdoor Code, and 9.4% for Wargames; these rates decline further when overseeing stronger systems.
Authors: Kaichen Zhang, Yuzhong Hong, Junwei Bao, Hongfei Jiang, Yang Song, Dingqian Hong, Hui Xiong
Abstract: Post-training plays a crucial role in refining and aligning large language models to meet specific tasks and human preferences. While recent advancements in post-training techniques, such as Group Relative Policy Optimization (GRPO), leverage increased sampling with relative reward scoring to achieve superior performance, these methods often suffer from training instability that limits their practical adoption. As a next step, we present Group Variance Policy Optimization (GVPO). GVPO incorporates the analytical solution to KL-constrained reward maximization directly into its gradient weights, ensuring alignment with the optimal policy. The method provides intuitive physical interpretations: its gradient mirrors the mean squared error between the central distance of implicit rewards and that of actual rewards. GVPO offers two key advantages: (1) it guarantees a unique optimal solution, exactly the KL-constrained reward maximization objective, (2) it supports flexible sampling distributions that avoids on-policy and importance sampling limitations. By unifying theoretical guarantees with practical adaptability, GVPO establishes a new paradigm for reliable and versatile LLM post-training.
Authors: Maximilian Stupp, P. S. Koutsourelakis
Abstract: Coarse-grained (CG) models provide an effective route to reducing the complexity of molecular simulations (MD), but conventional approaches depend heavily on long all-atom MD trajectories to adequately sample configurational space. This data dependence limits accuracy and generalizability, as unvisited configurations remain excluded from the resulting CG models. We introduce a fully data-free, generative framework for CG that directly targets the all-atom Boltzmann distribution. The model defines a structured latent space comprising slow collective variables, associated with multimodal marginal densities capturing metastable states, and fast variables, represented through simple, unimodal conditional distributions. A learnable, bijective map from latent space to atomistic coordinates enables the automatic and accurate reconstruction of molecular structures. Training relies solely on the interatomic potential and minimizes the reverse Kullback-Leibler (KL) divergence via an energy-based objective. To stabilize optimization and ensure mode coverage, we employ an adaptive tempering scheme that promotes the exploration of diverse configurations. Once trained, the model can generate independent, one-shot equilibrium samples at full atomic resolution. Validation on two synthetic systems, a double-well potential and a Gaussian mixture model, as well as on the benchmark alanine dipeptide, demonstrates that the method captures all relevant modes of the Boltzmann distribution, reconstructs atomic configurations, and automatically learns physically meaningful CG representations. These results suggest a promising, data-free alternative to traditional CG techniques, offering both a principled approach to addressing the long-standing "chicken-and-egg" challenge in coarse-graining and an effective solution to the back-mapping problem by enabling accurate reconstruction of all-atom configurations.
Authors: Zongxia Li, Xiyang Wu, Guangyao Shi, Yubin Qin, Hongyang Du, Fuxiao Liu, Tianyi Zhou, Dinesh Manocha, Jordan Lee Boyd-Graber
Abstract: Vision-Language Models (VLMs) have achieved strong results in video understanding, yet a key question remains: do they truly comprehend visual content or only learn shallow correlations between vision and language? Real visual understanding, especially of physics and common sense, is essential for AI systems that interact with the physical world. Current evaluations mostly use real-world videos similar to training data, so high benchmark scores may not reflect real reasoning ability. To address this, we propose negative-control tests using videos that depict physically impossible or logically inconsistent events. We introduce VideoHallu, a synthetic dataset of physics- and commonsense-violating scenes generated with Veo2, Sora, and Kling. It includes expert-annotated question-answer pairs across four categories of violations. Tests of leading VLMs (Qwen-2.5-VL, Video-R1, VideoChat-R1) show that, despite strong results on benchmarks such as MVBench and MMVU, they often miss these violations, exposing gaps in visual reasoning. Reinforcement learning fine-tuning on VideoHallu improves recognition of such violations without reducing standard benchmark performance. Our data is available at https://github.com/zli12321/VideoHallu.git.
Authors: Peter Chen, Xi Chen, Wotao Yin, Tianyi Lin
Abstract: Direct alignment methods are increasingly used for aligning large language models (LLMs) with human preferences. However, these methods suffer from the issues of verbosity and likelihood displacement, which can be driven by the noisy preference pairs that induce similar likelihood for preferred and dispreferred responses. The contributions of this paper are two-fold. First, we propose a new preference alignment method based on zeroth-order, comparison-based optimization via comparison oracles and provide convergence guarantees for its basic scheme. Second, we improve our method using some heuristics and conduct the experiments to demonstrate the flexibility and compatibility of practical scheme in improving the performance of LLMs using noisy preference pairs. Evaluations are conducted across multiple base and instruction-tuned models (Mistral-7B, Llama-3-8B and Gemma-2-9B) with benchmarks (AlpacaEval 2, MT-Bench and Arena-Hard). Experimental results show the effectiveness of our method as an alternative to addressing the limitations of existing direct alignment methods. A highlight of our work is that we evidence the importance of designing specialized methods for preference pairs with distinct likelihood margin, which complements the recent findings in Razin et al (2025).
Authors: Tobias Schnabel, Kiran Tomlinson, Adith Swaminathan, Jennifer Neville
Abstract: Despite their many successes, transformer-based large language models (LLMs) continue to struggle with tasks that require complex reasoning over large parts of their input. We argue that these failures arise due to capacity limits on the accurate flow of information within LLMs. To formalize this issue, we introduce the bounded attention prefix oracle (BAPO) model, a new computational framework that models bandwidth constraints on attention heads, the mechanism for internal communication in LLMs. We show that several important reasoning problems like graph reachability require high communication bandwidth for BAPOs to solve; we call these problems BAPO-hard. Our experiments corroborate our theoretical predictions: GPT-4o, Claude, and Gemini succeed on BAPO-easy tasks and fail even on relatively small BAPO-hard tasks. BAPOs also reveal another benefit of chain of thought (CoT): we prove that breaking down a task using CoT can turn any BAPO-hard problem into a BAPO-easy one. Our results offer principled explanations for key LLM failures and suggest directions for architectures and inference methods that mitigate bandwidth limits.
Authors: Yifeng Jiao, Yuchen Liu, Yu Zhang, Xin Guo, Yushuai Wu, Chen Jiang, Jiyang Li, Hongwei Zhang, Limei Han, Xin Gao, Yuan Qi, Yuan Cheng
Abstract: The advent of single-cell Assay for Transposase-Accessible Chromatin using sequencing (scATAC-seq) offers an innovative perspective for deciphering regulatory mechanisms by assembling a vast repository of single-cell chromatin accessibility data. While foundation models have achieved significant success in single-cell transcriptomics, there is currently no foundation model for scATAC-seq that supports zero-shot high-quality cell identification and comprehensive multi-omics analysis simultaneously. Key challenges lie in the high dimensionality and sparsity of scATAC-seq data, as well as the lack of a standardized schema for representing open chromatin regions (OCRs). Here, we present ChromFound, a foundation model tailored for scATAC-seq. ChromFound utilizes a hybrid architecture and genome-aware tokenization to effectively capture genome-wide long contexts and regulatory signals from dynamic chromatin landscapes. Pretrained on 1.97 million cells from 30 tissues and 6 disease conditions, ChromFound demonstrates broad applicability across 6 diverse tasks. Notably, it achieves robust zero-shot performance in generating universal cell representations and exhibits excellent transferability in cell type annotation and cross-omics prediction. By uncovering enhancer-gene links undetected by existing computational methods, ChromFound offers a promising framework for understanding disease risk variants in the noncoding genome.
Authors: Zahraa Al Sahili, Ioannis Patras, Matthew Purver
Abstract: Multilingual vision-language models (VLMs) promise universal image-text retrieval, yet their social biases remain underexplored. We perform the first systematic audit of four public multilingual CLIP variants: M-CLIP, NLLB-CLIP, CAPIVARA-CLIP, and the debiased SigLIP-2, covering ten languages that differ in resource availability and morphological gender marking. Using balanced subsets of FairFace and the PATA stereotype suite in a zero-shot setting, we quantify race and gender bias and measure stereotype amplification. Contrary to the intuition that multilinguality mitigates bias, every model exhibits stronger gender skew than its English-only baseline. CAPIVARA-CLIP shows its largest biases precisely in the low-resource languages it targets, while the shared encoder of NLLB-CLIP and SigLIP-2 transfers English gender stereotypes into gender-neutral languages; loosely coupled encoders largely avoid this leakage. Although SigLIP-2 reduces agency and communion skews, it inherits -- and in caption-sparse contexts (e.g., Xhosa) amplifies -- the English anchor's crime associations. Highly gendered languages consistently magnify all bias types, yet gender-neutral languages remain vulnerable whenever cross-lingual weight sharing imports foreign stereotypes. Aggregated metrics thus mask language-specific hot spots, underscoring the need for fine-grained, language-aware bias evaluation in future multilingual VLM research.
Authors: Minglu Zhao, Dehong Xu, Deqian Kong, Wen-Hao Zhang, Ying Nian Wu
Abstract: The hippocampus supports spatial navigation by encoding cognitive maps through collective place cell activity. We model the place cell population as non-negative spatial embeddings derived from the spectral decomposition of multi-step random walk transition kernels. In this framework, inner product or equivalently Euclidean distance between embeddings encode similarity between locations in terms of their transition probability across multiple scales, forming a cognitive map of adjacency. The combination of non-negativity and inner-product structure naturally induces sparsity, providing a principled explanation for the localized firing fields of place cells without imposing explicit constraints. The temporal parameter that defines the diffusion scale also determines field size, aligning with the hippocampal dorsoventral hierarchy. Our approach constructs global representations efficiently through recursive composition of local transitions, enabling smooth, trap-free navigation and preplay-like trajectory generation. Moreover, theta phase arises intrinsically as the angular relation between embeddings, linking spatial and temporal coding within a single representational geometry.
Authors: Patrick Kahardipraja, Reduan Achtibat, Thomas Wiegand, Wojciech Samek, Sebastian Lapuschkin
Abstract: Large language models are able to exploit in-context learning to access external knowledge beyond their training data through retrieval-augmentation. While promising, its inner workings remain unclear. In this work, we shed light on the mechanism of in-context retrieval augmentation for question answering by viewing a prompt as a composition of informational components. We propose an attribution-based method to identify specialized attention heads, revealing in-context heads that comprehend instructions and retrieve relevant contextual information, and parametric heads that store entities' relational knowledge. To better understand their roles, we extract function vectors and modify their attention weights to show how they can influence the answer generation process. Finally, we leverage the gained insights to trace the sources of knowledge used during inference, paving the way towards more safe and transparent language models.
Authors: Zhaomin Wu, Ziyang Wang, Bingsheng He
Abstract: Relational databases are often fragmented across organizations, creating data silos that hinder distributed data management and mining. Collaborative learning (CL) -- techniques that enable multiple parties to train models jointly without sharing raw data -- offers a principled approach to this challenge. However, existing CL frameworks (e.g., federated and split learning) remain limited in real-world deployments. Current CL benchmarks and algorithms primarily target the learning step under assumptions of isolated, aligned, and joinable databases, and they typically neglect the end-to-end data management pipeline, especially preprocessing steps such as table joins and data alignment. In contrast, our analysis of the real-world corpus WikiDBs shows that databases are interconnected, unaligned, and sometimes unjoinable, exposing a significant gap between CL algorithm design and practical deployment. To close this evaluation gap, we build WikiDBGraph, a large-scale dataset constructed from 100{,}000 real-world relational databases linked by 17 million weighted edges. Each node (database) and edge (relationship) is annotated with 13 and 12 properties, respectively, capturing a hybrid of instance- and feature-level overlap across databases. Experiments on WikiDBGraph demonstrate both the effectiveness and limitations of existing CL methods under realistic conditions, highlighting previously overlooked gaps in managing real-world data silos and pointing to concrete directions for practical deployment of collaborative learning systems.
Authors: Hossein Adeli, Sun Minni, Nikolaus Kriegeskorte
Abstract: A major goal of neuroscience is to understand brain computations during visual processing in naturalistic settings. A dominant approach is to use image-computable deep neural networks trained with different task objectives as a basis for linear encoding models. However, in addition to requiring estimation of a large number of linear encoding parameters, this approach ignores the structure of the feature maps both in the brain and the models. Recently proposed alternatives factor the linear mapping into separate sets of spatial and feature weights, thus finding static receptive fields for units, which is appropriate only for early visual areas. In this work, we employ the attention mechanism used in the transformer architecture to study how retinotopic visual features can be dynamically routed to category-selective areas in high-level visual processing. We show that this computational motif is significantly more powerful than alternative methods in predicting brain activity during natural scene viewing, across different feature basis models and modalities. We also show that this approach is inherently more interpretable as the attention-routing signals for different high-level categorical areas can be easily visualized for any input image. Given its high performance at predicting brain responses to novel images, the model deserves consideration as a candidate mechanistic model of how visual information from retinotopic maps is routed in the human brain based on the relevance of the input content to different category-selective regions.
Authors: Dan A. Calian, Gregory Farquhar, Iurii Kemaev, Luisa M. Zintgraf, Matteo Hessel, Jeremy Shar, Junhyuk Oh, Andr\'as Gy\"orgy, Tom Schaul, Jeffrey Dean, Hado van Hasselt, David Silver
Abstract: The quality of foundation models depends heavily on their training data. Consequently, great efforts have been put into dataset curation. Yet most approaches rely on manual tuning of coarse-grained mixtures of large buckets of data, or filtering by hand-crafted heuristics. An approach that is ultimately more scalable (let alone more satisfying) is to \emph{learn} which data is actually valuable for training. This type of meta-learning could allow more sophisticated, fine-grained, and effective curation. Our proposed \emph{DataRater} is an instance of this idea. It estimates the value of training on any particular data point. This is done by meta-learning using `meta-gradients', with the objective of improving training efficiency on held out data. In extensive experiments across a range of model scales and datasets, we find that using our DataRater to filter data is highly effective, resulting in significantly improved compute efficiency.
Authors: Yuezhou Ma, Haixu Wu, Hang Zhou, Huikun Weng, Jianmin Wang, Mingsheng Long
Abstract: Physics sensing plays a central role in many scientific and engineering domains, which inherently involves two coupled tasks: reconstructing dense physical fields from sparse observations and optimizing scattered sensor placements to observe maximum information. While deep learning has made rapid advances in sparse-data reconstruction, existing methods generally omit optimization of sensor placements, leaving the mutual enhancement between reconstruction and placement on the shelf. To change this suboptimal practice, we propose PhySense, a synergistic two-stage framework that learns to jointly reconstruct physical fields and to optimize sensor placements, both aiming for accurate physics sensing. The first stage involves a flow-based generative model enhanced by cross-attention to adaptively fuse sparse observations. Leveraging the reconstruction feedback, the second stage performs sensor placement via projected gradient descent to satisfy spatial constraints. We further prove that the learning objectives of the two stages are consistent with classical variance-minimization principles, providing theoretical guarantees. Extensive experiments across three challenging benchmarks, especially a 3D geometry dataset, indicate PhySense achieves state-of-the-art physics sensing accuracy and discovers informative sensor placements previously unconsidered. Code is available at this repository: https://github.com/thuml/PhySense.
Authors: Xiang Li, Yunai Li, Huiying Zhong, Lihua Lei, Zhun Deng
Abstract: Performativity of predictions refers to the phenomenon where prediction-informed decisions influence the very targets they aim to predict -- a dynamic commonly observed in policy-making, social sciences, and economics. In this paper, we initiate an end-to-end framework of statistical inference under performativity. Our contributions are twofold. First, we establish a central limit theorem for estimation and inference in the performative setting, enabling standard inferential tasks such as constructing confidence intervals and conducting hypothesis tests in policy-making contexts. Second, we leverage this central limit theorem to study prediction-powered inference (PPI) under performativity. This approach yields more precise estimates and tighter confidence regions for the model parameters (i.e., policies) of interest in performative prediction. We validate the effectiveness of our framework through numerical experiments. To the best of our knowledge, this is the first work to establish a complete statistical inference under performativity, introducing new challenges and inference settings that we believe will provide substantial value to policy-making, statistics, and machine learning.
Authors: Filippo Bigi, Sanggyu Chong, Agustinus Kristiadi, Michele Ceriotti
Abstract: Molecular dynamics (MD) provides insights into atomic-scale processes by integrating over time the equations that describe the motion of atoms under the action of interatomic forces. Machine learning models have substantially accelerated MD by providing inexpensive predictions of the forces, but they remain constrained to minuscule time integration steps, which are required by the fast time scale of atomic motion. In this work, we propose FlashMD, a method to predict the evolution of positions and momenta over strides that are between one and two orders of magnitude longer than typical MD time steps. We incorporate considerations on the mathematical and physical properties of Hamiltonian dynamics in the architecture, generalize the approach to allow the simulation of any thermodynamic ensemble, and carefully assess the possible failure modes of such a long-stride MD approach. We validate FlashMD's accuracy in reproducing equilibrium and time-dependent properties, using both system-specific and general-purpose models, extending the ability of MD simulation to reach the long time scales needed to model microscopic processes of high scientific and technological relevance.
Authors: Lai Wei, Yuting Li, Chen Wang, Yue Wang, Linghe Kong, Weiran Huang, Lichao Sun
Abstract: Improving Multi-modal Large Language Models (MLLMs) in the post-training stage typically relies on supervised fine-tuning (SFT) or reinforcement learning (RL), which require expensive and manually annotated multi-modal data--an ultimately unsustainable resource. This limitation has motivated a growing interest in unsupervised paradigms as a third stage of post-training after SFT and RL. While recent efforts have explored this direction, their methods are complex and difficult to iterate. To address this, we propose MM-UPT, a simple yet effective framework for unsupervised post-training of MLLMs, enabling continual self-improvement without any external supervision. The training method of MM-UPT builds upon GRPO, replacing traditional reward signals with a self-rewarding mechanism based on majority voting over multiple sampled responses. Our experiments demonstrate that such training method effectively improves the reasoning ability of Qwen2.5-VL-7B (e.g., 66.3\%$\rightarrow$72.9\% on MathVista, 62.9\%$\rightarrow$68.7\% on We-Math), using standard dataset without ground truth labels. To further explore scalability, we extend our framework to a data self-generation setting, designing two strategies that prompt the MLLM to synthesize new training samples on its own. Additional experiments show that combining these synthetic data with the unsupervised training method can also boost performance, highlighting a promising approach for scalable self-improvement. Overall, MM-UPT offers a new paradigm for autonomous enhancement of MLLMs, serving as a critical third step after initial SFT and RL in the absence of external supervision. Our code is available at https://github.com/waltonfuture/MM-UPT.
Authors: Agnideep Aich
Abstract: Classical estimators, the cornerstones of statistical inference, face insurmountable challenges when applied to important emerging classes of Archimedean copulas. These models exhibit pathological properties, including numerically unstable densities, a restrictive lower bound on Kendall's tau, and vanishingly small likelihood gradients, making MLE brittle and limiting MoM's applicability to datasets with sufficiently strong dependence (i.e., only when the empirical Kendall's $\tau$ exceeds the family's lower bound $\approx 0.545$). We introduce \textbf{IGNIS}, a unified neural estimation framework that sidesteps these barriers by learning a direct, robust mapping from data-driven dependency measures to the underlying copula parameter $\theta$. IGNIS utilizes a multi-input architecture and a theory-guided output layer ($\mathrm{softplus}(z) + 1$) to automatically enforce the domain constraint $\hat{\theta} \geq 1$. Trained and validated on four families (Gumbel, Joe, and the numerically challenging A1/A2), IGNIS delivers accurate and stable estimates for real-world financial and health datasets, demonstrating its necessity for reliable inference in modern, complex dependence models where traditional methods fail. To our knowledge, IGNIS is the first \emph{standalone, general-purpose} neural estimator for Archimedean copulas (not a generative model or likelihood optimizer), delivering direct, constraint-aware $\hat{\theta}$ and readily extensible to additional families via retraining or minor output-layer adaptations.
Authors: Agnideep Aich, Ashit Baran Aich
Abstract: We present the Deep Copula Classifier (DCC), a class-conditional generative model that separates marginal estimation from dependence modeling using neural copula densities. DCC is interpretable, Bayes-consistent, and achieves excess-risk $O(n^{-r/(2r+d)})$ for $r$-smooth copulas. In a controlled two-class study with strong dependence ($|\rho|=0.995$), DCC learns Bayes-aligned decision regions. With oracle or pooled marginals, it nearly reaches the best possible performance (accuracy $\approx 0.971$; ROC-AUC $\approx 0.998$). As expected, per-class KDE marginals perform less well (accuracy $0.873$; ROC-AUC $0.957$; PR-AUC $0.966$). On the Pima Indians Diabetes dataset, calibrated DCC ($\tau=1$) achieves accuracy $0.879$, ROC-AUC $0.936$, and PR-AUC $0.870$, outperforming Logistic Regression, SVM (RBF), and Naive Bayes, and matching Logistic Regression on the lowest Expected Calibration Error (ECE). Random Forest is also competitive (accuracy $0.892$; ROC-AUC $0.933$; PR-AUC $0.880$). Directly modeling feature dependence yields strong, well-calibrated performance with a clear probabilistic interpretation, making DCC a practical, theoretically grounded alternative to independence-based classifiers.
Authors: Manish Shetty, Naman Jain, Jinjian Liu, Vijay Kethanaboyina, Koushik Sen, Ion Stoica
Abstract: Developing high-performance software is a complex task that requires specialized expertise. We introduce GSO, a benchmark for evaluating language models' capabilities in developing high-performance software. We develop an automated pipeline that generates and executes performance tests to analyze repository commit histories to identify 102 challenging optimization tasks across 10 codebases, spanning diverse domains and programming languages. An agent is provided with a codebase and performance test as a precise specification, and tasked to improve the runtime efficiency, which is measured against the expert developer optimization. Our quantitative evaluation reveals that leading SWE-Agents struggle significantly, achieving less than 5% success rate, with limited improvements even with inference-time scaling. Our qualitative analysis identifies key failure modes, including difficulties with low-level languages, practicing lazy optimization strategies, and challenges in accurately localizing bottlenecks. We release the code and artifacts of our benchmark along with agent trajectories to enable future research.
Authors: Xiaoyuan Wu, Weiran Lin, Omer Akgul, Lujo Bauer
Abstract: Large language models (LLMs) are prone to hallucinations and sensitiveto prompt perturbations, often resulting in inconsistent or unreliablegenerated text. Different methods have been proposed to mitigate suchhallucinations and fragility, one of which is to measure theconsistency of LLM responses -- the model's confidence in the responseor likelihood of generating a similar response when resampled. Inprevious work, measuring LLM response consistency often relied oncalculating the probability of a response appearing within a pool of resampledresponses, analyzing internal states, or evaluating logits of resopnses.However, it was not clear how well theseapproaches approximated users' perceptions of consistency of LLMresponses. To find out, we performed a user study ($n=2,976$)demonstrating that current methods for measuring LLM responseconsistency typically do not align well with humans' perceptions of LLMconsistency. We propose a logit-based ensemble method for estimatingLLM consistency and show that our method matches the performance of thebest-performing existing metric in estimating human ratings of LLMconsistency. Our results suggest that methods for estimating LLMconsistency without human evaluation are sufficiently imperfect towarrant broader use of evaluation with human input; this would avoidmisjudging the adequacy of models because of the imperfections ofautomated consistency metrics.
Authors: Neehar Kondapaneni, Oisin Mac Aodha, Pietro Perona
Abstract: We propose a method for discovering and visualizing the differences between two learned representations, enabling more direct and interpretable model comparisons. We validate our method, which we call Representational Differences Explanations (RDX), by using it to compare models with known conceptual differences and demonstrate that it recovers meaningful distinctions where existing explainable AI (XAI) techniques fail. Applied to state-of-the-art models on challenging subsets of the ImageNet and iNaturalist datasets, RDX reveals both insightful representational differences and subtle patterns in the data. Although comparison is a cornerstone of scientific analysis, current tools in machine learning, namely post hoc XAI methods, struggle to support model comparison effectively. Our work addresses this gap by introducing an effective and explainable tool for contrasting model representations.
Authors: Aladin Djuhera, Swanand Ravindra Kadhe, Farhan Ahmed, Syed Zawad, Fernando Koch, Walid Saad, Holger Boche
Abstract: Fine-tuning large language models (LLMs) on telecom datasets is a common practice to adapt general-purpose models to the telecom domain. However, little attention has been paid to how this process may compromise model safety. Recent research has shown that even benign fine-tuning can degrade the safety alignment of LLMs, causing them to respond to harmful or unethical user queries. In this paper, we investigate this issue by fine-tuning LLMs on three representative telecom datasets and show that safety degrades even for light telecom domain adaptation. To this end, we introduce TeleHarm, the first telecom-specific red-teaming benchmark, which we use alongside established Direct-Harm and HexPhi datasets to systematically assess harmful behavior. We further extend our analysis to publicly available TeleLLMs that were continually pre-trained on large telecom corpora, revealing that safety alignment is severely lacking, primarily due to the omission of safety-focused instruction tuning. To address these issues, we evaluate three realignment defenses: SafeInstruct, SafeLoRA, SafeMERGE. We show that, across all settings, the proposed defenses can effectively restore safety without compromising telecom task performance, leading to Safe teleCOMMunication (SafeCOMM) models. Our work serves as both a diagnostic study and practical guide for safety realignment in telecom-tuned LLMs, underscoring the need for safety-aware instruction and fine-tuning in the telecom domain.
Authors: Sarthak Kumar Maharana, Saksham Singh Kushwaha, Baoming Zhang, Adrian Rodriguez, Songtao Wei, Yapeng Tian, Yunhui Guo
Abstract: While recent audio-visual models have demonstrated impressive performance, their robustness to distributional shifts at test-time remains not fully understood. Existing robustness benchmarks mainly focus on single modalities, making them insufficient for thoroughly assessing the robustness of audio-visual models. Motivated by real-world scenarios where shifts can occur $\textit{simultaneously}$ in both audio and visual modalities, we introduce $\texttt{AVROBUSTBENCH}$, a comprehensive benchmark designed to evaluate the test-time robustness of audio-visual recognition models. $\texttt{AVROBUSTBENCH}$ comprises four audio-visual benchmark datasets, $\texttt{AUDIOSET-2C}$, $\texttt{VGGSOUND-2C}$, $\texttt{KINETICS-2C}$, and $\texttt{EPICKITCHENS-2C}$, each incorporating 75 bimodal audio-visual corruptions that are $\textit{co-occurring}$ and $\textit{correlated}$. Through extensive evaluations, we observe that state-of-the-art supervised and self-supervised audio-visual models exhibit declining robustness as corruption severity increases. Furthermore, online test-time adaptation (TTA) methods, on $\texttt{VGGSOUND-2C}$ and $\texttt{KINETICS-2C}$, offer minimal improvements in performance under bimodal corruptions. We further propose $\texttt{AV2C}$, a simple TTA approach enabling on-the-fly cross-modal fusion by penalizing high-entropy samples, which achieves improvements on $\texttt{VGGSOUND-2C}$. We hope that $\texttt{AVROBUSTBENCH}$ will steer the development of more effective and robust audio-visual TTA approaches. Our code is available $\href{https://github.com/sarthaxxxxx/AV-C-Robustness-Benchmark}{here}$.
URLs: https://github.com/sarthaxxxxx/AV-C-Robustness-Benchmark
Authors: Zhihao Xu, Saisai Ding, Zhikun Zhang, Xiangjun Wang
Abstract: Integral equations are widely used in fields such as applied modeling, medical imaging, and system identification, providing a powerful framework for solving deterministic problems. While parameter identification for differential equations has been extensively studied, the focus on integral equations, particularly stochastic Volterra integral equations, remains limited. This research addresses the parameter identification problem, also known as the equation reconstruction problem, in Volterra integral equations driven by Gaussian noise. We propose an improved deep neural networks framework for estimating unknown parameters in the drift term of these equations. The network represents the primary variables and their integrals, enhancing parameter estimation accuracy by incorporating inter-output relationships into the loss function. Additionally, the framework extends beyond parameter identification to predict the system's behavior outside the integration interval. Prediction accuracy is validated by comparing predicted and true trajectories using a 95% confidence interval. Numerical experiments demonstrate the effectiveness of the proposed deep neural networks framework in both parameter identification and prediction tasks, showing robust performance under varying noise levels and providing accurate solutions for modeling stochastic systems.
Authors: Benjamin Matthias Ruppik, Julius von Rohrscheidt, Carel van Niekerk, Michael Heck, Renato Vukovic, Shutong Feng, Hsien-chin Lin, Nurul Lubis, Bastian Rieck, Marcus Zibrowius, Milica Ga\v{s}i\'c
Abstract: Understanding the internal mechanisms of large language models (LLMs) remains a challenging and complex endeavor. Even fundamental questions, such as how fine-tuning affects model behavior, often require extensive empirical evaluation. In this paper, we introduce a novel perspective based on the geometric properties of contextual latent embeddings to study the effects of training and fine-tuning. To that end, we measure the local dimensions of a contextual language model's latent space and analyze their shifts during training and fine-tuning. We show that the local dimensions provide insights into the model's training dynamics and generalization ability. Specifically, the mean of the local dimensions predicts when the model's training capabilities are exhausted, as exemplified in a dialogue state tracking task, overfitting, as demonstrated in an emotion recognition task, and grokking, as illustrated with an arithmetic task. Furthermore, our experiments suggest a practical heuristic: reductions in the mean local dimension tend to accompany and predict subsequent performance gains. Through this exploration, we aim to provide practitioners with a deeper understanding of the implications of fine-tuning on embedding spaces, facilitating informed decisions when configuring models for specific applications. The results of this work contribute to the ongoing discourse on the interpretability, adaptability, and generalizability of LLMs by bridging the gap between intrinsic model mechanisms and geometric properties in the respective embeddings.
Authors: Xinyu Zhu, Mengzhou Xia, Zhepei Wei, Wei-Lin Chen, Danqi Chen, Yu Meng
Abstract: Reinforcement learning with verifiable rewards (RLVR) is a promising approach for training language models (LMs) on reasoning tasks that elicit emergent long chains of thought (CoTs). Unlike supervised learning, it updates the model using both correct and incorrect samples via policy gradients. To better understand its mechanism, we decompose the learning signal into reinforcing correct responses and penalizing incorrect ones, referred to as Positive and Negative Sample Reinforcement (PSR and NSR), respectively. We train Qwen2.5-Math-7B, Qwen3-4B and Llama-3.1-8B-Instruct on a mathematical reasoning dataset and uncover a surprising result: training with only negative samples -- without reinforcing correct responses -- can be highly effective: it consistently improves performance over the base model across the entire Pass@$k$ spectrum $k$ up to 256), often matching or surpassing PPO and GRPO. In contrast, reinforcing only correct responses improves Pass@1 but degrades performance at higher $k$, due to reduced diversity. These inference-scaling trends highlight that solely penalizing incorrect responses may contribute more to performance than previously recognized. Through gradient analysis, we show that NSR works by suppressing incorrect generations and redistributing probability mass toward other plausible candidates, guided by the model's prior beliefs. It refines the model's existing knowledge rather than introducing entirely new behaviors. Building on this insight, we propose a simple variant of the RL objective that upweights NSR, and show that it consistently improves overall Pass@$k$ performance on MATH, AIME 2025, and AMC23. Our code is available at https://github.com/TianHongZXY/RLVR-Decomposed.
Authors: Daniel Szelogowski
Abstract: Despite substantial research into the biological basis of memory, the precise mechanisms by which experiences are encoded, stored, and retrieved in the brain remain incompletely understood. A growing body of evidence supports the engram theory, which posits that sparse populations of neurons undergo lasting physical and biochemical changes to support long-term memory. Yet, a comprehensive computational framework that integrates biological findings with mechanistic models remains elusive. This work synthesizes insights from cellular neuroscience and computational modeling to address key challenges in engram research: how engram neurons are identified and manipulated; how synaptic plasticity mechanisms contribute to stable memory traces; and how sparsity promotes efficient, interference-resistant representations. Relevant computational approaches -- such as sparse regularization, engram gating, and biologically inspired architectures like Sparse Distributed Memory and spiking neural networks -- are also examined. Together, these findings suggest that memory efficiency, capacity, and stability emerge from the interaction of plasticity and sparsity constraints. By integrating neurobiological and computational perspectives, this paper provides a comprehensive theoretical foundation for engram research and proposes a roadmap for future inquiry into the mechanisms underlying memory, with implications for the diagnosis and treatment of memory-related disorders.
Authors: Taha Entesari, Arman Hatami, Rinat Khaziev, Anil Ramakrishna, Mahyar Fazlyab
Abstract: Large Language Models (LLMs) deployed in real-world settings increasingly face the need to unlearn sensitive, outdated, or proprietary information. Existing unlearning methods typically formulate forgetting and retention as a regularized trade-off, combining both objectives into a single scalarized loss. This often leads to unstable optimization and degraded performance on retained data, especially under aggressive forgetting. We propose a new formulation of LLM unlearning as a constrained optimization problem: forgetting is enforced via a novel logit-margin flattening loss that explicitly drives the output distribution toward uniformity on a designated forget set, while retention is preserved through a hard constraint on a separate retain set. Compared to entropy-based objectives, our loss is softmax-free, numerically stable, and maintains non-vanishing gradients, enabling more efficient and robust optimization. We solve the constrained problem using a scalable primal-dual algorithm that exposes the trade-off between forgetting and retention through the dynamics of the dual variable, all without any extra computational overhead. Evaluations on the TOFU and MUSE benchmarks across diverse LLM architectures demonstrate that our approach consistently matches or exceeds state-of-the-art baselines, effectively removing targeted information while preserving downstream utility.
Authors: Elim Kwan, Rehman Qureshi, Liam Fletcher, Colin Laganier, Victoria Nockles, Richard Walters
Abstract: Cooperative autonomous robotic systems have significant potential for executing complex multi-task missions across space, air, ground, and maritime domains. But they commonly operate in remote, dynamic and hazardous environments, requiring rapid in-mission adaptation without reliance on fragile or slow communication links to centralised compute. Fast, on-board replanning algorithms are therefore needed to enhance resilience. Reinforcement Learning shows strong promise for efficiently solving mission planning tasks when formulated as Travelling Salesperson Problems (TSPs), but existing methods: 1) are unsuitable for replanning, where agents do not start at a single location; 2) do not allow cooperation between agents; 3) are unable to model tasks with variable durations; or 4) lack practical considerations for on-board deployment. Here we define the Cooperative Mission Replanning Problem as a novel variant of multiple TSP with adaptations to overcome these issues, and develop a new encoder/decoder-based model using Graph Attention Networks and Attention Models to solve it effectively and efficiently. Using a simple example of cooperative drones, we show our replanner consistently (90% of the time) maintains performance within 10% of the state-of-the-art LKH3 heuristic solver, whilst running 85-370 times faster on a Raspberry Pi. This work paves the way for increased resilience in autonomous multi-agent systems.
Authors: Yixiao Huang, Hanlin Zhu, Tianyu Guo, Jiantao Jiao, Somayeh Sojoudi, Michael I. Jordan, Stuart Russell, Song Mei
Abstract: Large language models (LLMs) can acquire new knowledge through fine-tuning, but this process exhibits a puzzling duality: models can generalize remarkably from new facts, yet are also prone to hallucinating incorrect information. However, the reasons for this phenomenon remain poorly understood. In this work, we argue that both behaviors stem from a single mechanism known as out-of-context reasoning (OCR): the ability to deduce implications by associating concepts, even those without a causal link. Our experiments across five prominent LLMs confirm that OCR indeed drives both generalization and hallucination, depending on whether the associated concepts are causally related. To build a rigorous theoretical understanding of this phenomenon, we then formalize OCR as a synthetic factual recall task. We empirically show that a one-layer single-head attention-only transformer with factorized output and value matrices can learn to solve this task, while a model with combined weights cannot, highlighting the crucial role of matrix factorization. Our theoretical analysis shows that the OCR capability can be attributed to the implicit bias of gradient descent, which favors solutions that minimize the nuclear norm of the combined output-value matrix. This mathematical structure explains why the model learns to associate facts and implications with high sample efficiency, regardless of whether the correlation is causal or merely spurious. Ultimately, our work provides a theoretical foundation for understanding the OCR phenomenon, offering a new lens for analyzing and mitigating undesirable behaviors from knowledge injection.
Authors: Avrajit Ghosh, Bai Cong, Rio Yokota, Saiprasad Ravishankar, Rongrong Wang, Molei Tao, Mohammad Emtiyaz Khan, Thomas M\"ollenhoff
Abstract: Variational Learning (VL) has recently gained popularity for training deep neural networks. Part of its empirical success can be explained by theories such as PAC-Bayes bounds, minimum description length and marginal likelihood, but little has been done to unravel the implicit regularization in play. Here, we analyze the implicit regularization of VL through the Edge of Stability (EoS) framework. EoS has previously been used to show that gradient descent can find flat solutions and we extend this result to show that VL can find even flatter solutions. This result is obtained by controlling the shape of the variational posterior as well as the number of posterior samples used during training. The derivation follows in a similar fashion as in the standard EoS literature for deep learning, by first deriving a result for a quadratic problem and then extending it to deep neural networks. We empirically validate these findings on a wide variety of large networks, such as ResNet and ViT, to find that the theoretical results closely match the empirical ones. Ours is the first work to analyze the EoS dynamics of VL.
Authors: Georgios Androutsopoulos, Antonio Bianchi
Abstract: Although Rust ensures memory safety by default, it also permits the use of unsafe code, which can introduce memory safety vulnerabilities if misused. Unfortunately, existing tools for detecting memory bugs in Rust typically exhibit limited detection capabilities, inadequately handle Rust-specific types, or rely heavily on manual intervention. To address these limitations, we present deepSURF, a tool that integrates static analysis with Large Language Model (LLM)-guided fuzzing harness generation to effectively identify memory safety vulnerabilities in Rust libraries, specifically targeting unsafe code. deepSURF introduces a novel approach for handling generics by substituting them with custom types and generating tailored implementations for the required traits, enabling the fuzzer to simulate user-defined behaviors within the fuzzed library. Additionally, deepSURF employs LLMs to augment fuzzing harnesses dynamically, facilitating exploration of complex API interactions and significantly increasing the likelihood of exposing memory safety vulnerabilities. We evaluated deepSURF on 63 real-world Rust crates, successfully rediscovering 30 known memory safety bugs and uncovering 12 previously-unknown vulnerabilities (out of which 11 have been assigned RustSec IDs and 3 have been patched), demonstrating clear improvements over state-of-the-art tools.
Authors: Chenlang Yi, Zizhan Xiong, Qi Qi, Xiyuan Wei, Girish Bathla, Ching-Long Lin, Bobak Jack Mortazavi, Tianbao Yang
Abstract: Contrastive Language-Image Pre-training (CLIP) models have demonstrated superior performance across various visual tasks including medical image classification. However, fairness concerns, including demographic biases, have received limited attention for CLIP models. This oversight leads to critical issues, particularly those related to race and gender, resulting in disparities in diagnostic outcomes and reduced reliability for underrepresented groups. To address these challenges, we introduce AdFair-CLIP, a novel framework employing adversarial feature intervention to suppress sensitive attributes, thereby mitigating spurious correlations and improving prediction fairness. We conduct comprehensive experiments on chest X-ray (CXR) datasets, and show that AdFair-CLIP significantly enhances both fairness and diagnostic accuracy, while maintaining robust generalization in zero-shot and few-shot scenarios. These results establish new benchmarks for fairness-aware learning in CLIP-based medical diagnostic models, particularly for CXR analysis.
Authors: Reihaneh Zohrabi, Hosein Hasani, Mahdieh Soleymani Baghshah, Anna Rohrbach, Marcus Rohrbach, Mohammad Hossein Rohban
Abstract: Out-of-distribution (OOD) detection is crucial for ensuring the reliability and safety of machine learning models in real-world applications, where they frequently face data distributions unseen during training. Despite progress, existing methods are often vulnerable to spurious correlations that mislead models and compromise robustness. To address this, we propose SPROD, a novel prototype-based OOD detection approach that explicitly addresses the challenge posed by unknown spurious correlations. Our post-hoc method refines class prototypes to mitigate bias from spurious features without additional data or hyperparameter tuning, and is broadly applicable across diverse backbones and OOD detection settings. We conduct a comprehensive spurious correlation OOD detection benchmarking, comparing our method against existing approaches and demonstrating its superior performance across challenging OOD datasets, such as CelebA, Waterbirds, UrbanCars, Spurious Imagenet, and the newly introduced Animals MetaCoCo. On average, SPROD improves AUROC by 4.8% and FPR@95 by 9.4% over the second best.
Authors: Hu Fu, Tao Lin
Abstract: In non-truthful auctions such as first-price and all-pay auctions, the independent strategic behaviors of bidders, with the corresponding Bayes-Nash equilibrium notion, are notoriously difficult to characterize and can cause undesirable outcomes. An alternative approach to achieve better outcomes in non-truthful auctions is to coordinate the bidders: let a mediator make incentive-compatible recommendations of correlated bidding strategies to the bidders, namely, implementing a Bayes correlated equilibrium (BCE). The implementation of BCE, however, requires knowledge of the distributions of bidders' private valuations, which is often unavailable. We initiate the study of the sample complexity of learning Bayes correlated equilibria in non-truthful auctions. We prove that the set of strategic-form BCEs in a large class of non-truthful auctions, including first-price and all-pay auctions, can be learned with a polynomial number $\tilde O(\frac{n}{\varepsilon^2})$ of samples of bidders' values. This moderate number of samples demonstrates the statistical feasibility of learning to coordinate bidders. Our technique is a reduction to the problem of estimating bidders' expected utility from samples, combined with an analysis of the pseudo-dimension of the class of all monotone bidding strategies.
Authors: Zilu Meng, Gregory J. Hakim, Wenchang Yang, Gabriel A. Vecchi
Abstract: Deep learning (DL)-based general circulation models (GCMs) are emerging as fast simulators, yet their ability to replicate extreme events outside their training range remains unknown. Here, we evaluate two such models -- the hybrid Neural General Circulation Model (NGCM) and purely data-driven Deep Learning Earth System Model (DL\textit{ESy}M) -- against a conventional high-resolution land-atmosphere model (HiRAM) in simulating land heatwaves and coldwaves. All models are forced with observed sea surface temperatures and sea ice over 1900-2020, focusing on the out-of-sample early-20th-century period (1900-1960). Both DL models generalize successfully to unseen climate conditions, broadly reproducing the frequency and spatial patterns of heatwave and cold wave events during 1900-1960 with skill comparable to HiRAM. An exception is over portions of North Asia and North America, where all models perform poorly during 1940-1960. Due to excessive temperature autocorrelation, DL\textit{ESy}M tends to overestimate heatwave and cold wave frequencies, whereas the physics-DL hybrid NGCM exhibits persistence more similar to HiRAM.
Authors: Liliang Ren, Congcong Chen, Haoran Xu, Young Jin Kim, Adam Atkinson, Zheng Zhan, Jiankai Sun, Baolin Peng, Liyuan Liu, Shuohang Wang, Hao Cheng, Jianfeng Gao, Weizhu Chen, Yelong Shen
Abstract: Recent advances in language modeling have demonstrated the effectiveness of State Space Models (SSMs) for efficient sequence modeling. While hybrid architectures such as Samba and the decoder-decoder architecture, YOCO, have shown promising performance gains over Transformers, prior works have not investigated the efficiency potential of representation sharing between SSM layers. In this paper, we introduce the Gated Memory Unit (GMU), a simple yet effective mechanism for efficient memory sharing across layers. We apply it to create SambaY, a decoder-hybrid-decoder architecture that incorporates GMUs in the cross-decoder to share memory readout states from a Samba-based self-decoder. SambaY significantly enhances decoding efficiency, preserves linear pre-filling time complexity, and boosts long-context performance, all while eliminating the need for explicit positional encoding. Through extensive scaling experiments, we demonstrate that our model exhibits a significantly lower irreducible loss compared to a strong YOCO baseline, indicating superior performance scalability under large-scale compute regimes. Our largest model enhanced with Differential Attention, Phi4-mini-Flash-Reasoning, achieves significantly better performance than Phi4-mini-Reasoning on reasoning tasks such as Math500, AIME24/25, and GPQA Diamond without any reinforcement learning, while delivering up to 10x higher decoding throughput on 2K-length prompts with 32K generation length under the vLLM inference framework. We release our training codebase on open-source data at https://github.com/microsoft/ArchScale.
Authors: Fang Chen, Alex Villa, Gongbo Liang, Xiaoyi Lu, Meng Tang
Abstract: Training data for class-conditional image synthesis often exhibit a long-tailed distribution with limited images for tail classes. Such an imbalance causes mode collapse and reduces the diversity of synthesized images for tail classes. For class-conditional diffusion models trained on imbalanced data, we aim to improve the diversity and fidelity of tail class images without compromising the quality of head class images. We achieve this by introducing two simple but highly effective loss functions. Firstly, we employ an Unsupervised Contrastive Loss (UCL) utilizing negative samples to increase the distance/dissimilarity among synthetic images. Such regularization is coupled with a standard trick of batch resampling to further diversify tail-class images. Our second loss is an Alignment Loss (AL) that aligns class-conditional generation with unconditional generation at large timesteps. This second loss makes the denoising process insensitive to class conditions for the initial steps, which enriches tail classes through knowledge sharing from head classes. We successfully leverage contrastive learning and conditional-unconditional alignment for class-imbalanced diffusion models. Our framework is easy to implement as demonstrated on both U-Net based architecture and Diffusion Transformer. Our method outperforms vanilla denoising diffusion probabilistic models, score-based diffusion model, and alternative methods for class-imbalanced image generation across various datasets, in particular ImageNet-LT with 256x256 resolution.
Authors: Sangmin Bae, Yujin Kim, Reza Bayat, Sungnyun Kim, Jiyoun Ha, Tal Schuster, Adam Fisch, Hrayr Harutyunyan, Ziwei Ji, Aaron Courville, Se-Young Yun
Abstract: Scaling language models unlocks impressive capabilities, but the accompanying computational and memory demands make both training and deployment expensive. Existing efficiency efforts typically target either parameter sharing or adaptive computation, leaving open the question of how to attain both simultaneously. We introduce Mixture-of-Recursions (MoR), a unified framework that combines the two axes of efficiency inside a single Recursive Transformer. MoR reuses a shared stack of layers across recursion steps to achieve parameter efficiency, while lightweight routers enable adaptive token-level thinking by dynamically assigning different recursion depths to individual tokens. This allows MoR to focus quadratic attention computation only among tokens still active at a given recursion depth, further improving memory access efficiency by selectively caching only their key-value pairs. Beyond these core mechanisms, we also propose a KV sharing variant that reuses KV pairs from the first recursion, specifically designed to further decrease memory footprint. Across model scales ranging from 135M to 1.7B parameters, MoR forms a new Pareto frontier: at equal training FLOPs and smaller model sizes, it significantly lowers validation perplexity and improves few-shot accuracy, while delivering higher throughput compared with vanilla and existing recursive baselines. These gains demonstrate that MoR is an effective path towards large-model quality without incurring large-model cost.
Authors: Xiaohao Liu, Xiaobo Xia, See-Kiong Ng, Tat-Seng Chua
Abstract: Multimodal representation learning seeks to create a unified representation space by integrating diverse data modalities to improve multimodal understanding. Traditional methods often depend on pairwise contrastive learning, which relies on a predefined anchor modality, restricting alignment across all modalities. Recent advances have investigated the simultaneous alignment of multiple modalities, yet several challenges remain, such as limitations imposed by fixed anchor points and instability arising from optimizing the product of singular values. To address the challenges, in this paper, we propose Principled Multimodal Representation Learning (PMRL), a novel framework that achieves simultaneous alignment of multiple modalities without anchor dependency in a more stable manner. Specifically, grounded in the theoretical insight that full alignment corresponds to a rank-1 Gram matrix, PMRL optimizes the dominant singular value of the representation matrix to align modalities along a shared leading direction. We propose a softmax-based loss function that treats singular values as logits to prioritize the largest singular value. Besides, instance-wise contrastive regularization on the leading eigenvectors maintains inter-instance separability and prevents representation collapse. Extensive experiments across diverse tasks demonstrate PMRL's superiority compared to baseline methods. The source code will be publicly available.
Authors: Trevor Harris, Yan Liu
Abstract: Operator models are regression algorithms between Banach spaces of functions. They have become an increasingly critical tool for spatiotemporal forecasting and physics emulation, especially in high-stakes scenarios where robust, calibrated uncertainty quantification is required. We introduce Local Sliced Conformal Inference (LSCI), a distribution-free framework for generating function-valued, locally adaptive prediction sets for operator models. We prove finite-sample validity and derive a data-dependent upper bound on the coverage gap under local exchangeability. On synthetic Gaussian-process tasks and real applications (air quality monitoring, energy demand forecasting, and weather prediction), LSCI yields tighter sets with stronger adaptivity compared to conformal baselines. We also empirically demonstrate robustness against biased predictions and certain out-of-distribution noise regimes.
Authors: Woojae Jeong, Aditya Kommineni, Kleanthis Avramidis, Colin McDaniel, Donald Berry, Myzelle Hughes, Thomas McGee, Elsi Kaiser, Dani Byrd, Assal Habibi, B. Rael Cahn, Idan A. Blank, Kristina Lerman, Dimitrios Pantazis, Sudarsana R. Kadiri, Takfarinas Medani, Shrikanth Narayanan, Richard M. Leahy
Abstract: Depression and suicidality affect cognitive and emotional processes, yet objective, task-evoked neural readouts of mental health remain limited. We investigated the spatiotemporal dynamics of affective semantic processing using multivariate decoding of time-resolved, 64-channel electroencephalography (EEG). Participants (N=137) performed a sentence-evaluation task with emotionally salient, self-referential statements. We identified robust neural signatures of semantic processing, with peak decoding accuracy between 300-600 ms -- a window associated with rapid, stimulus-driven semantic evaluation and conflict monitoring. Relative to healthy controls, individuals with depression and suicidal ideation showed earlier onset, longer duration, and greater amplitude decoding responses, along with broader cross-temporal generalization and enhanced contributions from frontocentral and parietotemporal components. These findings suggest altered sensitivity and impaired disengagement from emotionally salient content in the clinical groups, advancing our understanding of the neurocognitive basis of mental health and establishing a compact and interpretable EEG-based index of semantic-evaluation dynamics with potential diagnostic relevance.
Authors: Lyle Regenwetter, Yazan Abu Obaideh, Fabien Chiotti, Ioanna Lykourentzou, Faez Ahmed
Abstract: We introduce BikeBench, an engineering design benchmark for evaluating generative models on problems with multiple real-world objectives and constraints. As generative AI's reach continues to grow, evaluating its capability to understand physical laws, human guidelines, and hard constraints grows increasingly important. Engineering product design lies at the intersection of these difficult tasks, providing new challenges for AI capabilities. BikeBench evaluates AI models' capabilities to generate bicycle designs that not only resemble the dataset, but meet specific performance objectives and constraints. To do so, BikeBench quantifies a variety of human-centered and multiphysics performance characteristics, such as aerodynamics, ergonomics, structural mechanics, human-rated usability, and similarity to subjective text or image prompts. Supporting the benchmark are several datasets of simulation results, a dataset of 10,000 human-rated bicycle assessments, and a synthetically generated dataset of 1.6M designs, each with a parametric, CAD/XML, SVG, and PNG representation. BikeBench is uniquely configured to evaluate tabular generative models, large language models (LLMs), design optimization, and hybrid algorithms side-by-side. Our experiments indicate that LLMs and tabular generative models fall short of hybrid GenAI+optimization algorithms in design quality, constraint satisfaction, and similarity scores, suggesting significant room for improvement. We hope that BikeBench, a first-of-its-kind benchmark, will help catalyze progress in generative AI for constrained multi-objective engineering design problems. We provide code, data, an interactive leaderboard, and other resources at https://github.com/Lyleregenwetter/BikeBench.
Authors: Alain Riou, Bernardo Torres, Ben Hayes, Stefan Lattner, Ga\"etan Hadjeres, Ga\"el Richard, Geoffroy Peeters
Abstract: In this paper, we introduce PESTO, a self-supervised learning approach for single-pitch estimation using a Siamese architecture. Our model processes individual frames of a Variable-$Q$ Transform (VQT) and predicts pitch distributions. The neural network is designed to be equivariant to translations, notably thanks to a Toeplitz fully-connected layer. In addition, we construct pitch-shifted pairs by translating and cropping the VQT frames and train our model with a novel class-based transposition-equivariant objective, eliminating the need for annotated data. Thanks to this architecture and training objective, our model achieves remarkable performances while being very lightweight ($130$k parameters). Evaluations on music and speech datasets (MIR-1K, MDB-stem-synth, and PTDB) demonstrate that PESTO not only outperforms self-supervised baselines but also competes with supervised methods, exhibiting superior cross-dataset generalization. Finally, we enhance PESTO's practical utility by developing a streamable VQT implementation using cached convolutions. Combined with our model's low latency (less than 10 ms) and minimal parameter count, this makes PESTO particularly suitable for real-time applications.
Authors: Sami Sadat, Mohammad Irtiza Hossain, Junaid Ahmed Sifat, Suhail Haque Rafi, Md. Waseq Alauddin Alvi, Md. Khalilur Rhaman
Abstract: A deep learning real-time smoking detection system for CCTV surveillance of fire exit areas is proposed due to critical safety requirements. The dataset contains 8,124 images from 20 different scenarios along with 2,708 raw samples demonstrating low-light areas. We evaluated three advanced object detection models: YOLOv8, YOLOv11, and YOLOv12, followed by development of a custom model derived from YOLOv8 with added structures for challenging surveillance contexts. The proposed model outperformed the others, achieving a recall of 78.90 percent and mAP at 50 of 83.70 percent, delivering optimal object detection across varied environments. Performance evaluation on multiple edge devices using multithreaded operations showed the Jetson Xavier NX processed data at 52 to 97 milliseconds per inference, establishing its suitability for time-sensitive operations. This system offers a robust and adaptable platform for monitoring public safety and enabling automatic regulatory compliance.
Authors: Woojin Chung, Jeonghoon Kim
Abstract: Large language models are trained with tokenizers, and the resulting token distribution is highly imbalanced: a few words dominate the stream while most occur rarely. Recent practice favors ever-larger vocabularies, but it is unclear where the benefit comes from. To this end, we perform a controlled study that scales the vocabulary of the language model from 24K to 196K while holding data, computation, and optimization unchanged. We begin by quantifying the complexity of tokenized text -- formalized via Kolmogorov complexity -- and show that larger vocabularies reduce this complexity. Above 24K, every common word is already tokenized as a single token, so enlarging vocabulary only deepens the relative token-frequency imbalance. Word-level loss decomposition shows that larger vocabularies reduce cross-entropy loss almost exclusively by lowering uncertainty on the 2,500 most frequent words, even though loss on the rare tail rises. The same frequent words cover roughly 75% of tokens in downstream benchmarks, so this training advantage transfers intact. We further show that enlarging model parameters with a fixed vocabulary yields the same frequent-word benefit. Our results recast "bigger vocabularies help" as "lowering complexity of tokenized text helps," offering a simple, principled knob for tokenizer--model co-design and clarifying the loss dynamics that govern language model scaling in pre-training.
Authors: Daniel M. Yellin
Abstract: In this paper we evaluate the capabilities of LLM Agents in generating code for real-world problems. Specifically, we explore code synthesis for microservice-based applications, a widely used architectural pattern for building applications. We define a standard template for specifying these applications, and we propose a metric for scoring the difficulty of a specification. The higher the score, the more difficult it is to generate code for the specification. Our experimental results show that agents using strong LLMs (like GPT-3o-mini) do fairly well on medium difficulty specifications but do poorly on those of higher difficulty levels. This is due to more intricate business logic, a greater use of external services, database integration and inclusion of non-functional capabilities such as authentication. We analyzed the errors in LLM-synthesized code and report on the key challenges LLM Agents face in generating code for these specifications. Finally, we show that using a fine-grained approach to code generation improves the correctness of the generated code.
Authors: Ji Xie, Trevor Darrell, Luke Zettlemoyer, XuDong Wang
Abstract: Unified multimodal models (UMMs) unify visual understanding and generation within a single architecture. However, conventional training relies on image-text pairs (or sequences) whose captions are typically sparse and miss fine-grained visual details--even when they use hundreds of words to describe a simple image. We introduce Reconstruction Alignment (RecA), a resource-efficient post-training method that leverages visual understanding encoder embeddings as dense "text prompts," providing rich supervision without captions. Concretely, RecA conditions a UMM on its own visual understanding embeddings and optimizes it to reconstruct the input image with a self-supervised reconstruction loss, thereby realigning understanding and generation. Despite its simplicity, RecA is broadly applicable: across autoregressive, masked-autoregressive, and diffusion-based UMMs, it consistently improves generation and editing fidelity. With only 27 GPU-hours, post-training with RecA substantially improves image generation performance on GenEval (0.73$\rightarrow$0.90) and DPGBench (80.93$\rightarrow$88.15), while also boosting editing benchmarks (ImgEdit 3.38$\rightarrow$3.75, GEdit 6.94$\rightarrow$7.25). Notably, RecA surpasses much larger open-source models and applies broadly across diverse UMM architectures, establishing it as an efficient and general post-training alignment strategy for UMMs
Authors: Simin Chen, Jinjun Peng, Yixin He, Junfeng Yang, Baishakhi Ray
Abstract: Deep learning (DL) compilers are core infrastructure in modern DL systems, offering flexibility and scalability beyond vendor-specific libraries. This work uncovers a fundamental vulnerability in their design: can an official, unmodified compiler alter a model's semantics during compilation and introduce hidden backdoors? We study both adversarial and natural settings. In the adversarial case, we craft benign models where triggers have no effect pre-compilation but become effective backdoors after compilation. Tested on six models, three commercial compilers, and two hardware platforms, our attack yields 100% success on triggered inputs while preserving normal accuracy and remaining undetected by state-of-the-art detectors. The attack generalizes across compilers, hardware, and floating-point settings. In the natural setting, we analyze the top 100 HuggingFace models (including one with 220M+ downloads) and find natural triggers in 31 models. This shows that compilers can introduce risks even without adversarial manipulation. Our results reveal an overlooked threat: unmodified DL compilers can silently alter model semantics. To our knowledge, this is the first work to expose inherent security risks in DL compiler design, opening a new direction for secure and trustworthy ML.
Authors: Momchil S. Tomov, Sang Uk Lee, Hansford Hendrago, Jinwook Huh, Teawon Han, Forbes Howington, Rafael da Silva, Gianmarco Bernasconi, Marc Heim, Samuel Findler, Xiaonan Ji, Alexander Boule, Michael Napoli, Kuo Chen, Jesse Miller, Boaz Floor, Yunqing Hu
Abstract: We present TreeIRL, a novel planner for autonomous driving that combines Monte Carlo tree search (MCTS) and inverse reinforcement learning (IRL) to achieve state-of-the-art performance in simulation and in real-world driving. The core idea is to use MCTS to find a promising set of safe candidate trajectories and a deep IRL scoring function to select the most human-like among them. We evaluate TreeIRL against both classical and state-of-the-art planners in large-scale simulations and on 500+ miles of real-world autonomous driving in the Las Vegas metropolitan area. Test scenarios include dense urban traffic, adaptive cruise control, cut-ins, and traffic lights. TreeIRL achieves the best overall performance, striking a balance between safety, progress, comfort, and human-likeness. To our knowledge, our work is the first demonstration of MCTS-based planning on public roads and underscores the importance of evaluating planners across a diverse set of metrics and in real-world environments. TreeIRL is highly extensible and could be further improved with reinforcement learning and imitation learning, providing a framework for exploring different combinations of classical and learning-based approaches to solve the planning bottleneck in autonomous driving.
Authors: Kijung Yoon
Abstract: Inferring synaptic connectivity from neural population activity is a fundamental challenge in computational neuroscience, complicated by partial observability and mismatches between inference models and true circuit dynamics. In this study, we propose a graph-based neural inference model that simultaneously predicts neural activity and infers latent connectivity by modeling neurons as interacting nodes in a graph. The architecture features two distinct modules: one for learning structural connectivity and another for predicting future spiking activity via a graph neural network (GNN). Our model accommodates unobserved neurons through auxiliary nodes, allowing for inference in partially observed circuits. We evaluate this approach using synthetic data generated from ring attractor network models and real spike recordings from head direction cells in mice. Across a wide range of conditions, including varying recurrent connectivity, external inputs, and incomplete observations, our model reliably resolves spurious correlations and recovers accurate weight profiles. When applied to real data, the inferred connectivity aligns with theoretical predictions of continuous attractor models. These results highlight the potential of GNN-based models to infer latent neural circuitry through self-supervised structure learning, while leveraging the spike prediction task to flexibly link connectivity and dynamics across both simulated and biological neural systems.
Authors: Abdou Karim Kandji, Fr\'ed\'eric Precioso, Cheikh Ba, Samba Ndiaye, Augustin Ndione
Abstract: Intent classification models have made a significant progress in recent years. However, previous studies primarily focus on high-resource language datasets, which results in a gap for low-resource languages and for regions with high rates of illiteracy, where languages are more spoken than read or written. This is the case in Senegal, for example, where Wolof is spoken by around 90\% of the population, while the national illiteracy rate remains at of 42\%. Wolof is actually spoken by more than 10 million people in West African region. To address these limitations, we introduce the Wolof Banking Speech Intent Classification Dataset (WolBanking77), for academic research in intent classification. WolBanking77 currently contains 9,791 text sentences in the banking domain and more than 4 hours of spoken sentences. Experiments on various baselines are conducted in this work, including text and voice state-of-the-art models. The results are very promising on this current dataset. In addition, this paper presents an in-depth examination of the dataset's contents. We report baseline F1-scores and word error rates metrics respectively on NLP and ASR models trained on WolBanking77 dataset and also comparisons between models. Dataset and code available at: https://github.com/abdoukarim/wolbanking77.
Authors: Yandan Yang, Baoxiong Jia, Shujie Zhang, Siyuan Huang
Abstract: Indoor scene synthesis has become increasingly important with the rise of Embodied AI, which requires 3D environments that are not only visually realistic but also physically plausible and functionally diverse. While recent approaches have advanced visual fidelity, they often remain constrained to fixed scene categories, lack sufficient object-level detail and physical consistency, and struggle to align with complex user instructions. In this work, we present SceneWeaver, a reflective agentic framework that unifies diverse scene synthesis paradigms through tool-based iterative refinement. At its core, SceneWeaver employs a language model-based planner to select from a suite of extensible scene generation tools, ranging from data-driven generative models to visual- and LLM-based methods, guided by self-evaluation of physical plausibility, visual realism, and semantic alignment with user input. This closed-loop reason-act-reflect design enables the agent to identify semantic inconsistencies, invoke targeted tools, and update the environment over successive iterations. Extensive experiments on both common and open-vocabulary room types demonstrate that SceneWeaver not only outperforms prior methods on physical, visual, and semantic metrics, but also generalizes effectively to complex scenes with diverse instructions, marking a step toward general-purpose 3D environment generation. Project website: https://scene-weaver.github.io/.
Authors: Aymen Bouguerra, Daniel Montoya, Alexandra Gomez-Villa, Fabio Arnez, Chokri Mraidha
Abstract: The powerful zero-shot generalization capabilities of vision-language models (VLMs) like CLIP have enabled new paradigms for safety-related tasks such as out-of-distribution (OOD) detection. However, additional aspects crucial for the computationally efficient and reliable deployment of CLIP are still overlooked. In particular, the impact of quantization on CLIP's performance beyond accuracy remains underexplored. This work presents a large-scale evaluation of quantization on CLIP models, assessing not only in-distribution accuracy but a comprehensive suite of reliability metrics and revealing counterintuitive results driven by pre-training source. We demonstrate that quantization consistently improves calibration for typically underconfident pre-trained models, while often degrading it for overconfident variants. Intriguingly, this degradation in calibration does not preclude gains in other reliability metrics; we find that OOD detection can still improve for these same poorly calibrated models. Furthermore, we identify specific quantization-aware training (QAT) methods that yield simultaneous gains in zero-shot accuracy, calibration, and OOD robustness, challenging the view of a strict efficiency-performance trade-off. These findings offer critical insights for navigating the multi-objective problem of deploying efficient, reliable, and robust VLMs by utilizing quantization beyond its conventional role.
Authors: Yu Gui, Cong Ma, Zongming Ma
Abstract: Learning disentangled representations is a fundamental task in multi-modal learning. In modern applications such as single-cell multi-omics, both shared and modality-specific features are critical for characterizing cell states and supporting downstream analyses. Ideally, modality-specific features should be independent of shared ones while also capturing all complementary information within each modality. This tradeoff is naturally expressed through information-theoretic criteria, but mutual-information-based objectives are difficult to estimate reliably, and their variational surrogates often underperform in practice. In this paper, we introduce IndiSeek, a novel disentangled representation learning approach that addresses this challenge by combining an independence-enforcing objective with a computationally efficient reconstruction loss that bounds conditional mutual information. This formulation explicitly balances independence and completeness, enabling principled extraction of modality-specific features. We demonstrate the effectiveness of IndiSeek on synthetic simulations, a CITE-seq dataset and multiple real-world multi-modal benchmarks.
Authors: Md. Mahfuzur Rahman, Kishor Datta Gupta, Marufa Kamal, Fahad Rahman, Sunzida Siddique, Ahmed Rafi Hasan, Mohd Ariful Haque, Roy George
Abstract: Immediate damage assessment is essential after natural catastrophes; yet, conventional hand evaluation techniques are sluggish and perilous. Although satellite and unmanned aerial vehicle (UAV) photos offer extensive perspectives of impacted regions, current computer vision methodologies generally yield just classification labels or segmentation masks, so constraining their capacity to deliver a thorough situational comprehension. We introduce the Vision Language Caption Enhancer (VLCE), a multimodal system designed to produce comprehensive, contextually-informed explanations of disaster imagery. VLCE employs a dual-architecture approach: a CNN-LSTM model with a ResNet50 backbone pretrained on EuroSat satellite imagery for the xBD dataset, and a Vision Transformer (ViT) model pretrained on UAV pictures for the RescueNet dataset. Both systems utilize external semantic knowledge from ConceptNet and WordNet to expand vocabulary coverage and improve description accuracy. We assess VLCE in comparison to leading vision-language models (LLaVA and QwenVL) utilizing CLIPScore for semantic alignment and InfoMetIC for caption informativeness. Experimental findings indicate that VLCE markedly surpasses baseline models, attaining a maximum of 95.33% on InfoMetIC while preserving competitive semantic alignment. Our dual-architecture system demonstrates significant potential for improving disaster damage assessment by automating the production of actionable, information-dense descriptions from satellite and drone photos.
Authors: Shubhang Bhatnagar, Andy Xu, Kar-Han Tan, Narendra Ahuja
Abstract: Large Language Models (LLMs) with multimodal capabilities have revolutionized vision-language tasks, but their deployment often requires huge memory and computational resources. While post-training quantization (PTQ) has successfully compressed language models to as low as 1-bit precision without significant performance loss, its effectiveness for multimodal LLMs (MLLMs) remains relatively unexplored. In this paper, we present the first study on ultra-low bit (<4-bit) quantization for multimodal LLMs. Our analysis reveals that multimodal tokens and intermediate layer activations produced by them exhibit significantly higher statistical variance and entropy compared to text tokens, making them less tolerant to ultra-low bit quantization. However, the activation distributions of multimodal tokens varies significantly over different layers, with some layers having lower entropy activation distributions. We empirically show that such layers in these models can better tolerate ultra-low bit quantization. Building on these insights, we propose a novel strategy for MLLM quantization, LUQ: Layerwise Ultra-Low Bit Quantization, which selectively applies ultra-low bit quantization to layers that are more resilient to it. Additionally, we also show that using a mix of multimodal tokens (image and text) for PTQ boosts VQA performance in the ultra-low bit regime. We evaluate our method on LLaVA-1.5 and Qwen-2.5-VL across 9 popular VQA benchmarks. The resulting LUQ models use 40% and 31% less memory than their 4-bit counterparts, respectively, while exhibiting a performance degradation of less than 10% on the MME benchmark.
Authors: Yixian Zhang, Shu'ang Yu, Tonghe Zhang, Mo Guang, Haojia Hui, Kaiwen Long, Yu Wang, Chao Yu, Wenbo Ding
Abstract: Training expressive flow-based policies with off-policy reinforcement learning is notoriously unstable due to gradient pathologies in the multi-step action sampling process. We trace this instability to a fundamental connection: the flow rollout is algebraically equivalent to a residual recurrent computation, making it susceptible to the same vanishing and exploding gradients as RNNs. To address this, we reparameterize the velocity network using principles from modern sequential models, introducing two stable architectures: Flow-G, which incorporates a gated velocity, and Flow-T, which utilizes a decoded velocity. We then develop a practical SAC-based algorithm, enabled by a noise-augmented rollout, that facilitates direct end-to-end training of these policies. Our approach supports both from-scratch and offline-to-online learning and achieves state-of-the-art performance on continuous control and robotic manipulation benchmarks, eliminating the need for common workarounds like policy distillation or surrogate objectives.
Authors: Shigui Li, Wei Chen, Delu Zeng
Abstract: Diffusion models (DMs) excel in image generation, but suffer from slow inference and the training-inference discrepancies. Although gradient-based solvers like DPM-Solver accelerate the denoising inference, they lack theoretical foundations in information transmission efficiency. In this work, we introduce an information-theoretic perspective on the inference processes of DMs, revealing that successful denoising fundamentally reduces conditional entropy in reverse transitions. This principle leads to our key insights into the inference processes: (1) data prediction parameterization outperforms its noise counterpart, and (2) optimizing conditional variance offers a reference-free way to minimize both transition and reconstruction errors. Based on these insights, we propose an entropy-aware variance optimized method for the generative process of DMs, called EVODiff, which systematically reduces uncertainty by optimizing conditional entropy during denoising. Extensive experiments on DMs validate our insights and demonstrate that our method significantly and consistently outperforms state-of-the-art (SOTA) gradient-based solvers. For example, compared to the DPM-Solver++, EVODiff reduces the reconstruction error by up to 45.5\% (FID improves from 5.10 to 2.78) at 10 function evaluations (NFE) on CIFAR-10, cuts the NFE cost by 25\% (from 20 to 15 NFE) for high-quality samples on ImageNet-256, and improves text-to-image generation while reducing artifacts. Code is available at https://github.com/ShiguiLi/EVODiff.
Authors: Samuel Pfisterer, Florian Gr\"otschla, Luca A. Lanzend\"orfer, Florian Yan, Roger Wattenhofer
Abstract: Recent progress in speech processing has highlighted that high-quality performance across languages requires substantial training data for each individual language. While existing multilingual datasets cover many languages, they often contain insufficient data for most languages. Thus, trained models perform poorly on the majority of the supported languages. Our work addresses this challenge by introducing a scalable pipeline for constructing speech datasets from parliamentary recordings. The proposed pipeline includes robust components for media retrieval and a two-stage alignment algorithm designed to handle non-verbatim transcripts and long-form audio. Applying this pipeline to recordings from 22 European parliaments, we extract over 61k hours of aligned speech segments, achieving substantial per-language coverage with 19 languages exceeding 1k hours and 22 languages exceeding 500 hours of high-quality speech data. We obtain an average 41.8\% reduction in word error rates over baselines when finetuning an existing ASR model on our dataset, demonstrating the usefulness of our approach.
Authors: Pierre Musacchio, Hyunmin Lee, Jaesik Park
Abstract: Even in controlled settings, understanding instance-wise geometries is a challenging task for a wide range of visual models. Although specialized systems exist, modern arts rely on expensive input formats (category labels, binary segmentation masks) and inference costs (a quadratic amount of forward passes). We mitigate these limitations by proposing InstaFormer, a network capable of holistic order prediction. That is, solely given an input RGB image, InstaFormer returns the full occlusion and depth orderings for all the instances in the scene in a single forward pass. At its core, InstaFormer relies on interactions between object queries and latent mask descriptors that semantically represent the same objects while carrying complementary information. We comprehensively benchmark and ablate our approach to highlight its effectiveness. Our code and models are open-source and available at this URL: https://github.com/SNU-VGILab/InstaOrder.
Authors: Guangliang Liu, Haitao Mao, Bochuan Cao, Zhiyu Xue, Xitong Zhang, Rongrong Wang, Kristen Marie Johnson
Abstract: Large Language Models (LLMs) are able to improve their responses when instructed to do so, a capability known as self-correction. When instructions provide only a general and abstract goal without specific details about potential issues in the response, LLMs must rely on their internal knowledge to improve response quality, a process referred to as intrinsic self-correction. The empirical success of intrinsic self-correction is evident in various applications, but how and why it is effective remains unknown. Focusing on moral self-correction in LLMs, we reveal a key characteristic of intrinsic self-correction: performance convergence through multi-round interactions; and provide a mechanistic analysis of this convergence behavior. Based on our experimental results and analysis, we uncover the underlying mechanism of convergence: consistently injected self-correction instructions activate moral concepts that reduce model uncertainty, leading to converged performance as the activated moral concepts stabilize over successive rounds. This paper demonstrates the strong potential of moral self-correction by showing that it exhibits a desirable property of converged performance.
Authors: Irash Perera (Department of Computer Science and Engineering, University of Moratuwa, Colombo, Sri Lanka), Uthayasanker Thayasivam (Department of Computer Science and Engineering, University of Moratuwa, Colombo, Sri Lanka)
Abstract: Intelligent analysis of medical imaging plays a crucial role in assisting clinical diagnosis, especially for identifying subtle pathological features. This paper introduces a novel multi-branch ConvNeXt architecture designed specifically for the nuanced challenges of medical image analysis. While applied here to the specific problem of COVID-19 diagnosis, the methodology offers a generalizable framework for classifying a wide range of pathologies from CT scans. The proposed model incorporates a rigorous end-to-end pipeline, from meticulous data preprocessing and augmentation to a disciplined two-phase training strategy that leverages transfer learning effectively. The architecture uniquely integrates features extracted from three parallel branches: Global Average Pooling, Global Max Pooling, and a new Attention-weighted Pooling mechanism. The model was trained and validated on a combined dataset of 2,609 CT slices derived from two distinct datasets. Experimental results demonstrate a superior performance on the validation set, achieving a final ROC-AUC of 0.9937, a validation accuracy of 0.9757, and an F1-score of 0.9825 for COVID-19 cases, outperforming all previously reported models on this dataset. These findings indicate that a modern, multi-branch architecture, coupled with careful data handling, can achieve performance comparable to or exceeding contemporary state-of-the-art models, thereby proving the efficacy of advanced deep learning techniques for robust medical diagnostics.
Authors: Mir Tafseer Nayeem, Sawsan Alqahtani, Md Tahmid Rahman Laskar, Tasnim Mohiuddin, M Saiful Bari
Abstract: Tokenization is a crucial but under-evaluated step in large language models (LLMs). The standard metric, fertility (the average number of tokens per word), captures compression efficiency but obscures how vocabularies are allocated across languages and domains. We analyze six widely used tokenizers across seven languages and two domains, finding stable fertility for English, high fertility for Chinese, and little domain sensitivity. To address fertility's blind spots, we propose the Single Token Retention Rate (STRR), which measures the proportion of words preserved as single tokens. STRR reveals systematic prioritization of English, strong support for Chinese, and fragmentation in Hindi, offering an interpretable view of cross-lingual fairness. Our results show that STRR complements fertility and provides practical guidance for designing more equitable multilingual tokenizers.
Authors: Yuanche Liu, Yingxuan Xu, Yang Zhang
Abstract: We introduce a machine-learning framework based on symbolic regression to extract the full symbol alphabet of multi-loop Feynman integrals. By targeting the analytic structure rather than reduction, the method is broadly applicable and interpretable across different families of integrals. It successfully reconstructs complete symbol alphabets in nontrivial examples, demonstrating both robustness and generality. Beyond accelerating computations case by case, it uncovers the analytic structure universally. This framework opens new avenues for multi-loop amplitude analysis and provides a versatile tool for exploring scattering amplitudes.
Authors: Jiafang Song, Sandipan Pramanik, Abhirup Datta
Abstract: In many scientific applications, uncertainty of estimates from an earlier (upstream) analysis needs to be propagated in subsequent (downstream) Bayesian analysis, without feedback. Cutting feedback methods, also termed cut-Bayes, achieve this by constructing a cut-posterior distribution that prevents backward information flow. Cutting feedback like nested MCMC is computationally challenging while variational inference (VI) cut-Bayes methods need two variational approximations and require access to the upstream data and model. In this manuscript we propose, NeVI-Cut, a provably accurate and modular neural network-based variational inference method for cutting feedback. We directly utilize samples from the upstream analysis without requiring access to the upstream data or model. This simultaneously preserves modularity of analysis and reduces approximation errors by avoiding a variational approximation for the upstream model. We then use normalizing flows to specify the conditional variational family for the downstream parameters and estimate the conditional cut-posterior as a variational solution of Monte Carlo average loss over all the upstream samples. We provide theoretical guarantees on the NeVI-Cut estimate to approximate any cut-posterior. Our results are in a fixed-data regime and provide convergence rates of the actual variational solution, quantifying how richness of the neural architecture and the complexity of the target cut-posterior dictate the approximation quality. In the process, we establish new results on uniform Kullback-Leibler approximation rates of conditional normalizing flows. Simulation studies and two real-world analyses illustrate how NeVI-Cut achieves significant computational gains over traditional cutting feedback methods and is considerably more accurate than parametric variational cut approaches.
Authors: Shudong Sun, Hao Helen Zhang
Abstract: Transfer learning has become a cornerstone of modern machine learning, as it can empower models by leveraging knowledge from related domains to improve learning effectiveness. However, transferring from poorly aligned data can harm rather than help performance, making it crucial to determine whether the transfer will be beneficial before implementation. This work aims to address this challenge by proposing an innovative metric to measure dataset similarity and provide quantitative guidance on transferability. In the literature, existing methods largely focus on feature distributions while overlooking label information and predictive relationships, potentially missing critical transferability insights. In contrast, our proposed metric, the Cross-Learning Score (CLS), measures dataset similarity through bidirectional generalization performance between domains. We provide a theoretical justification for CLS by establishing its connection to the cosine similarity between the decision boundaries for the target and source datasets. Computationally, CLS is efficient and fast to compute as it bypasses the problem of expensive distribution estimation for high-dimensional problems. We further introduce a general framework that categorizes source datasets into positive, ambiguous, or negative transfer zones based on their CLS relative to the baseline error, enabling informed decisions. Additionally, we extend this approach to encoder-head architectures in deep learning to better reflect modern transfer pipelines. Extensive experiments on diverse synthetic and real-world tasks demonstrate that CLS can reliably predict whether transfer will improve or degrade performance, offering a principled tool for guiding data selection in transfer learning.
Authors: Alain Riou, Joan Serr\`a, Yuki Mitsufuji
Abstract: Sampling, the technique of reusing pieces of existing audio tracks to create new music content, is a very common practice in modern music production. In this paper, we tackle the challenging task of automatic sample identification, that is, detecting such sampled content and retrieving the material from which it originates. To do so, we adopt a self-supervised learning approach that leverages a multi-track dataset to create positive pairs of artificial mixes, and design a novel contrastive learning objective. We show that such method significantly outperforms previous state-of-the-art baselines, that is robust to various genres, and that scales well when increasing the number of noise songs in the reference database. In addition, we extensively analyze the contribution of the different components of our training pipeline and highlight, in particular, the need for high-quality separated stems for this task.
Authors: Pin-Lun Chen, Chi-Hsi Kung, Che-Han Chang, Wei-Chen Chiu, Yi-Ting Chen
Abstract: Evaluating the safety of autonomous vehicles (AVs) requires diverse, safety-critical scenarios, with collisions being especially important yet rare and unsafe to collect in the real world. Therefore, the community has been focusing on generating safety-critical scenarios in simulation. However, controlling attributes such as collision type and time-to-accident (TTA) remains challenging. We introduce a new task called controllable collision scenario generation, where the goal is to produce trajectories that realize a user-specified collision type and TTA, to investigate the feasibility of automatically generating desired collision scenarios. To support this task, we present COLLIDE, a large-scale collision scenario dataset constructed by transforming real-world driving logs into diverse collisions, balanced across five representative collision types and different TTA intervals. We propose a framework that predicts Collision Pattern, a compact and interpretable representation that captures the spatial configuration of the ego and the adversarial vehicles at impact, before rolling out full adversarial trajectories. Experiments show that our approach outperforms strong baselines in both collision rate and controllability. Furthermore, generated scenarios consistently induce higher planner failure rates, revealing limitations of existing planners. We demonstrate that these scenarios fine-tune planners for robustness improvements, contributing to safer AV deployment in different collision scenarios. Project page is available at https://submit-user.github.io/anon2025
Authors: Michele Loi, Marcello Di Bello, Nicol\`o Cangiotti
Abstract: The growing philosophical literature on algorithmic fairness has examined statistical criteria such as equalized odds and calibration, causal and counterfactual approaches, and the role of structural and compounding injustices. Yet an important dimension has been overlooked: whether the evidential value of an algorithmic output itself depends on structural injustice. We contrast a predictive policing algorithm, which relies on historical crime data, with a camera-based system that records ongoing offenses, where both are designed to guide police deployment. In evaluating the moral acceptability of acting on a piece of evidence, we must ask not only whether the evidence is probative in the actual world, but also whether it would remain probative in nearby worlds without the relevant injustices. The predictive policing algorithm fails this test, but the camera-based system passes it. When evidence fails the test, it is morally problematic to use it punitively, more so than evidence that passes the test.
Authors: Deeptimaan Banerjee, Prateek Gothwal, Ashis Kumer Biswas
Abstract: In many domains, including online education, healthcare, security, and human-computer interaction, facial emotion recognition (FER) is essential. Real-world FER is still difficult despite its significance because of some factors such as variable head positions, occlusions, illumination shifts, and demographic diversity. Engagement detection, which is essential for applications like virtual learning and customer services, is frequently challenging due to FER limitations by many current models. In this article, we propose ExpressNet-MoE, a novel hybrid deep learning model that blends both Convolution Neural Networks (CNNs) and Mixture of Experts (MoE) framework, to overcome the difficulties. Our model dynamically chooses the most pertinent expert networks, thus it aids in the generalization and providing flexibility to model across a wide variety of datasets. Our model improves on the accuracy of emotion recognition by utilizing multi-scale feature extraction to collect both global and local facial features. ExpressNet-MoE includes numerous CNN-based feature extractors, a MoE module for adaptive feature selection, and finally a residual network backbone for deep feature learning. To demonstrate efficacy of our proposed model we evaluated on several datasets, and compared with current state-of-the-art methods. Our model achieves accuracies of 74.77% on AffectNet (v7), 72.55% on AffectNet (v8), 84.29% on RAF-DB, and 64.66% on FER-2013. The results show how adaptive our model is and how it may be used to develop end-to-end emotion recognition systems in practical settings. Reproducible codes and results are made publicly accessible at https://github.com/DeeptimaanB/ExpressNet-MoE.
Authors: Xiaojian Ding, Lin Zhao, Xian Li, Xiaoying Zhu
Abstract: Incomplete multi-view data, where certain views are entirely missing for some samples, poses significant challenges for traditional multi-view clustering methods. Existing deep incomplete multi-view clustering approaches often rely on static fusion strategies or two-stage pipelines, leading to suboptimal fusion results and error propagation issues. To address these limitations, this paper proposes a novel incomplete multi-view clustering framework based on Hierarchical Semantic Alignment and Cooperative Completion (HSACC). HSACC achieves robust cross-view fusion through a dual-level semantic space design. In the low-level semantic space, consistency alignment is ensured by maximizing mutual information across views. In the high-level semantic space, adaptive view weights are dynamically assigned based on the distributional affinity between individual views and an initial fused representation, followed by weighted fusion to generate a unified global representation. Additionally, HSACC implicitly recovers missing views by projecting aligned latent representations into high-dimensional semantic spaces and jointly optimizes reconstruction and clustering objectives, enabling cooperative learning of completion and clustering. Experimental results demonstrate that HSACC significantly outperforms state-of-the-art methods on five benchmark datasets. Ablation studies validate the effectiveness of the hierarchical alignment and dynamic weighting mechanisms, while parameter analysis confirms the model's robustness to hyperparameter variations.
Authors: Maor Reuben, Ido Mendel, Or Feldman, Moshe Kravchik, Mordehai Guri, Rami Puzis
Abstract: Supply chain attacks significantly threaten software security with malicious code injections within legitimate projects. Such attacks are very rare but may have a devastating impact. Detecting spurious code injections using automated tools is further complicated as it often requires deciphering the intention of both the inserted code and its context. In this study, we propose an unsupervised approach for highlighting spurious code injections by quantifying cohesion disruptions in the source code. Using a name-prediction-based cohesion (NPC) metric, we analyze how function cohesion changes when malicious code is introduced compared to natural cohesion fluctuations. An analysis of 54,707 functions over 369 open-source C++ repositories reveals that code injection reduces cohesion and shifts naming patterns toward shorter, less descriptive names compared to genuine function updates. Considering the sporadic nature of real supply-chain attacks, we evaluate the proposed method with extreme test-set imbalance and show that monitoring high-cohesion functions with NPC can effectively detect functions with injected code, achieving a Precision@100 of 36.41% at a 1:1,000 ratio and 12.47% at 1:10,000. These results suggest that automated cohesion measurements, in general, and name-prediction-based cohesion, in particular, may help identify supply chain attacks, improving source code integrity.
Authors: Jiani Huang, Amish Sethi, Matthew Kuo, Mayank Keoliya, Neelay Velingker, JungHo Jung, Ser-Nam Lim, Ziyang Li, Mayur Naik
Abstract: Multi-modal large language models (MLLMs) are making rapid progress toward general-purpose embodied agents. However, existing MLLMs do not reliably capture fine-grained links between low-level visual features and high-level textual semantics, leading to weak grounding and inaccurate perception. To overcome this challenge, we propose ESCA, a framework that contextualizes embodied agents by grounding their perception in spatial-temporal scene graphs. At its core is SGCLIP, a novel, open-domain, promptable foundation model for generating scene graphs that is based on CLIP. SGCLIP is trained on 87K+ open-domain videos using a neurosymbolic pipeline that aligns automatically generated captions with scene graphs produced by the model itself, eliminating the need for human-labeled annotations. We demonstrate that SGCLIP excels in both prompt-based inference and task-specific fine-tuning, achieving state-of-the-art results on scene graph generation and action localization benchmarks. ESCA with SGCLIP improves perception for embodied agents based on both open-source and commercial MLLMs, achieving state of-the-art performance across two embodied environments. Notably, ESCA significantly reduces agent perception errors and enables open-source models to surpass proprietary baselines. We release the source code for SGCLIP model training at https://github.com/video-fm/LASER and for the embodied agent at https://github.com/video-fm/ESCA.
URLs: https://github.com/video-fm/LASER, https://github.com/video-fm/ESCA.
Authors: Mansi Phute, Matthew Hull, Haoran Wang, Alec Helbling, ShengYun Peng, Willian Lunardi, Martin Andreoni, Wenke Lee, Duen Horng Chau
Abstract: Deep learning models deployed in safety critical applications like autonomous driving use simulations to test their robustness against adversarial attacks in realistic conditions. However, these simulations are non-differentiable, forcing researchers to create attacks that do not integrate simulation environmental factors, reducing attack success. To address this limitation, we introduce UNDREAM, the first software framework that bridges the gap between photorealistic simulators and differentiable renderers to enable end-to-end optimization of adversarial perturbations on any 3D objects. UNDREAM enables manipulation of the environment by offering complete control over weather, lighting, backgrounds, camera angles, trajectories, and realistic human and object movements, thereby allowing the creation of diverse scenes. We showcase a wide array of distinct physically plausible adversarial objects that UNDREAM enables researchers to swiftly explore in different configurable environments. This combination of photorealistic simulation and differentiable optimization opens new avenues for advancing research of physical adversarial attacks.
Authors: Tiancheng Hu, Joachim Baumann, Lorenzo Lupo, Nigel Collier, Dirk Hovy, Paul R\"ottger
Abstract: Large language model (LLM) simulations of human behavior have the potential to revolutionize the social and behavioral sciences, if and only if they faithfully reflect real human behaviors. Current evaluations are fragmented, based on bespoke tasks and metrics, creating a patchwork of incomparable results. To address this, we introduce SimBench, the first large-scale, standardized benchmark for a robust, reproducible science of LLM simulation. By unifying 20 diverse datasets covering tasks from moral decision-making to economic choice across a large global participant pool, SimBench provides the necessary foundation to ask fundamental questions about when, how, and why LLM simulations succeed or fail. We show that, while even the best LLMs today have limited simulation ability (score: 40.80/100), performance scales log-linearly with model size. Simulation performance is not improved by increased inference-time compute. We demonstrate an alignment-simulation trade-off: instruction-tuning improves performance on low-entropy (consensus) questions but degrades it on high-entropy (diverse) ones. Models particularly struggle when simulating specific demographic groups. Finally, we demonstrate that simulation ability correlates most strongly with deep, knowledge-intensive reasoning (MMLU-Pro, r=0.939). By making progress measurable, we aim to accelerate the development of more faithful LLM simulators.
Authors: Anjie Liu, Jianhong Wang, Samuel Kaski, Jun Wang, Mengyue Yang
Abstract: Steering cooperative multi-agent reinforcement learning (MARL) towards desired outcomes is challenging, particularly when the global guidance from a human on the whole multi-agent system is impractical in a large-scale MARL. On the other hand, designing external mechanisms (e.g., intrinsic rewards and human feedback) to coordinate agents mostly relies on empirical studies, lacking a easy-to-use research tool. In this work, we employ multi-agent influence diagrams (MAIDs) as a graphical framework to address the above issues. First, we introduce the concept of MARL interaction paradigms (orthogonal to MARL learning paradigms), using MAIDs to analyze and visualize both unguided self-organization and global guidance mechanisms in MARL. Then, we design a new MARL interaction paradigm, referred to as the targeted intervention paradigm that is applied to only a single targeted agent, so the problem of global guidance can be mitigated. In implementation, we introduce a causal inference technique, referred to as Pre-Strategy Intervention (PSI), to realize the targeted intervention paradigm. Since MAIDs can be regarded as a special class of causal diagrams, a composite desired outcome that integrates the primary task goal and an additional desired outcome can be achieved by maximizing the corresponding causal effect through the PSI. Moreover, the bundled relevance graph analysis of MAIDs provides a tool to identify whether an MARL learning paradigm is workable under the design of an MARL interaction paradigm. In experiments, we demonstrate the effectiveness of our proposed targeted intervention, and verify the result of relevance graph analysis.
Authors: Atticus McWhorter, Daryl DeFord
Abstract: Novel Markov Chain Monte Carlo (MCMC) methods have enabled the generation of large ensembles of redistricting plans through graph partitioning. However, existing algorithms such as Reversible Recombination (RevReCom) and Metropolized Forest Recombination (MFR) are constrained to sampling from distributions related to spanning trees. We introduce the marked edge walk (MEW), a novel MCMC algorithm for sampling from the space of graph partitions under a tunable distribution. The walk operates on the space of spanning trees with marked edges, allowing for calculable transition probabilities for use in the Metropolis-Hastings algorithm. Empirical results on real-world dual graphs show convergence under target distributions unrelated to spanning trees. For this reason, MEW represents an advancement in flexible ensemble generation.
Authors: Mohammad Abdul Rehman, Syed Imad Ali Shah, Abbas Anwar, Noor Islam
Abstract: Large Language Models (LLMs) can reason over natural-language inputs, but their role in intrusion detection without fine-tuning remains uncertain. This study evaluates a prompt-only approach on UNSW-NB15 by converting each network flow to a compact textual record and augmenting it with lightweight, domain-inspired boolean flags (asymmetry, burst rate, TTL irregularities, timer anomalies, rare service/state, short bursts). To reduce output drift and support measurement, the model is constrained to produce structured, grammar-valid responses, and a single decision threshold is calibrated on a small development split. We compare zero-shot, instruction-guided, and few-shot prompting to strong tabular and neural baselines under identical splits, reporting accuracy, precision, recall, F1, and macro scores. Empirically, unguided prompting is unreliable, while instructions plus flags substantially improve detection quality; adding calibrated scoring further stabilizes results. On a balanced subset of two hundred flows, a 7B instruction-tuned model with flags reaches macro-F1 near 0.78; a lighter 3B model with few-shot cues and calibration attains F1 near 0.68 on one thousand examples. As the evaluation set grows to two thousand flows, decision quality decreases, revealing sensitivity to coverage and prompting. Tabular baselines remain more stable and faster, yet the prompt-only pipeline requires no gradient training, produces readable artifacts, and adapts easily through instructions and flags. Contributions include a flow-to-text protocol with interpretable cues, a calibration method for thresholding, a systematic baseline comparison, and a reproducibility bundle with prompts, grammar, metrics, and figures.
Authors: Mohammad Abdul Rehman, Syed Imad Ali Shah, Abbas Anwar, Noor Islam
Abstract: The remarkable capabilities of Large Language Models (LLMs) in natural language understanding and generation have sparked interest in their potential for cybersecurity applications, including password guessing. In this study, we conduct an empirical investigation into the efficacy of pre-trained LLMs for password cracking using synthetic user profiles. Specifically, we evaluate the performance of state-of-the-art open-source LLMs such as TinyLLaMA, Falcon-RW-1B, and Flan-T5 by prompting them to generate plausible passwords based on structured user attributes (e.g., name, birthdate, hobbies). Our results, measured using Hit@1, Hit@5, and Hit@10 metrics under both plaintext and SHA-256 hash comparisons, reveal consistently poor performance, with all models achieving less than 1.5% accuracy at Hit@10. In contrast, traditional rule-based and combinator-based cracking methods demonstrate significantly higher success rates. Through detailed analysis and visualization, we identify key limitations in the generative reasoning of LLMs when applied to the domain-specific task of password guessing. Our findings suggest that, despite their linguistic prowess, current LLMs lack the domain adaptation and memorization capabilities required for effective password inference, especially in the absence of supervised fine-tuning on leaked password datasets. This study provides critical insights into the limitations of LLMs in adversarial contexts and lays the groundwork for future efforts in secure, privacy-preserving, and robust password modeling.
Authors: Prateek Gothwal, Deeptimaan Banerjee, Ashis Kumer Biswas
Abstract: Engagement detection in online learning environments is vital for improving student outcomes and personalizing instruction. We present ViBED-Net (Video-Based Engagement Detection Network), a novel deep learning framework designed to assess student engagement from video data using a dual-stream architecture. ViBED-Net captures both facial expressions and full-scene context by processing facial crops and entire video frames through EfficientNetV2 for spatial feature extraction. These features are then analyzed over time using two temporal modeling strategies: Long Short-Term Memory (LSTM) networks and Transformer encoders. Our model is evaluated on the DAiSEE dataset, a large-scale benchmark for affective state recognition in e-learning. To enhance performance on underrepresented engagement classes, we apply targeted data augmentation techniques. Among the tested variants, ViBED-Net with LSTM achieves 73.43\% accuracy, outperforming existing state-of-the-art approaches. ViBED-Net demonstrates that combining face-aware and scene-aware spatiotemporal cues significantly improves engagement detection accuracy. Its modular design allows flexibility for application across education, user experience research, and content personalization. This work advances video-based affective computing by offering a scalable, high-performing solution for real-world engagement analysis. The source code for this project is available on https://github.com/prateek-gothwal/ViBED-Net .
Authors: Kevin Huang, Rosario Scalise, Cleah Winston, Ayush Agrawal, Yunchu Zhang, Rohan Baijal, Markus Grotz, Byron Boots, Benjamin Burchfiel, Masha Itkina, Paarth Shah, Abhishek Gupta
Abstract: Imitation learning has proven effective for training robots to perform complex tasks from expert human demonstrations. However, it remains limited by its reliance on high-quality, task-specific data, restricting adaptability to the diverse range of real-world object configurations and scenarios. In contrast, non-expert data -- such as play data, suboptimal demonstrations, partial task completions, or rollouts from suboptimal policies -- can offer broader coverage and lower collection costs. However, conventional imitation learning approaches fail to utilize this data effectively. To address these challenges, we posit that with right design decisions, offline reinforcement learning can be used as a tool to harness non-expert data to enhance the performance of imitation learning policies. We show that while standard offline RL approaches can be ineffective at actually leveraging non-expert data under the sparse data coverage settings typically encountered in the real world, simple algorithmic modifications can allow for the utilization of this data, without significant additional assumptions. Our approach shows that broadening the support of the policy distribution can allow imitation algorithms augmented by offline RL to solve tasks robustly, showing considerably enhanced recovery and generalization behavior. In manipulation tasks, these innovations significantly increase the range of initial conditions where learned policies are successful when non-expert data is incorporated. Moreover, we show that these methods are able to leverage all collected data, including partial or suboptimal demonstrations, to bolster task-directed policy performance. This underscores the importance of algorithmic techniques for using non-expert data for robust policy learning in robotics. Website: https://uwrobotlearning.github.io/RISE-offline/
Authors: Antonio Norelli, Michael Bronstein
Abstract: A meaningful text can be hidden inside another, completely different yet still coherent and plausible, text of the same length. For example, a tweet containing a harsh political critique could be embedded in a tweet that celebrates the same political leader, or an ordinary product review could conceal a secret manuscript. This uncanny state of affairs is now possible thanks to Large Language Models, and in this paper we present a simple and efficient protocol to achieve it. We show that even modest 8-billion-parameter open-source LLMs are sufficient to obtain high-quality results, and a message as long as this abstract can be encoded and decoded locally on a laptop in seconds. The existence of such a protocol demonstrates a radical decoupling of text from authorial intent, further eroding trust in written communication, already shaken by the rise of LLM chatbots. We illustrate this with a concrete scenario: a company could covertly deploy an unfiltered LLM by encoding its answers within the compliant responses of a safe model. This possibility raises urgent questions for AI safety and challenges our understanding of what it means for a Large Language Model to know something.
Authors: Krishnakumar Balasubramanian, Sayan Banerjee, Philippe Rigollet
Abstract: We study stationary solutions of McKean-Vlasov equations on the circle. Our main contributions stem from observing an exact equivalence between solutions of the stationary McKean-Vlasov equation and an infinite-dimensional quadratic system of equations over Fourier coefficients, which allows explicit characterization of the stationary states in a sequence space rather than a function space. This framework provides a transparent description of local bifurcations, characterizing their periodicity, and resonance structures, while accommodating singular potentials. We derive analytic expressions that characterize the emergence, form and shape (supercritical, critical, subcritical or transcritical) of bifurcations involving possibly multiple Fourier modes and connect them with discontinuous phase transitions. We also characterize, under suitable assumptions, the detailed structure of the stationary bifurcating solutions that are accurate upto an arbitrary number of Fourier modes. At the global level, we establish regularity and concavity properties of the free energy landscape, proving existence, compactness, and coexistence of globally minimizing stationary measures, further identifying discontinuous phase transitions with points of non-differentiability of the minimum free energy map. As an application, we specialize the theory to the Noisy Mean-Field Transformer model, where we show how changing the inverse temperature parameter $\beta$ affects the geometry of the infinitely many bifurcations from the uniform measure. We also explain how increasing $\beta$ can lead to a rich class of approximate multi-mode stationary solutions which can be seen as `metastable states'. Further, a sharp transition from continuous to discontinuous (first-order) phase behavior is observed as $\beta$ increases.
Authors: Nimrod Berman, Omkar Joglekar, Eitan Kosman, Dotan Di Castro, Omri Azencot
Abstract: Recent advances in generative modeling have positioned diffusion models as state-of-the-art tools for sampling from complex data distributions. While these models have shown remarkable success across single-modality domains such as images and audio, extending their capabilities to Modality Translation (MT), translating information across different sensory modalities, remains an open challenge. Existing approaches often rely on restrictive assumptions, including shared dimensionality, Gaussian source priors, and modality-specific architectures, which limit their generality and theoretical grounding. In this work, we propose the Latent Denoising Diffusion Bridge Model (LDDBM), a general-purpose framework for modality translation based on a latent-variable extension of Denoising Diffusion Bridge Models. By operating in a shared latent space, our method learns a bridge between arbitrary modalities without requiring aligned dimensions. We introduce a contrastive alignment loss to enforce semantic consistency between paired samples and design a domain-agnostic encoder-decoder architecture tailored for noise prediction in latent space. Additionally, we propose a predictive loss to guide training toward accurate cross-domain translation and explore several training strategies to improve stability. Our approach supports arbitrary modality pairs and performs strongly on diverse MT tasks, including multi-view to 3D shape generation, image super-resolution, and multi-view scene synthesis. Comprehensive experiments and ablations validate the effectiveness of our framework, establishing a new strong baseline in general modality translation. For more information, see our project page: https://sites.google.com/view/lddbm/home.
Authors: Naomi Desobry, Elnura Zhalieva, Souhaib Ben Taieb
Abstract: Probabilistic models must be well calibrated to support reliable decision-making. While calibration in single-output regression is well studied, defining and achieving multivariate calibration in multi-output regression remains considerably more challenging. The existing literature on multivariate calibration primarily focuses on diagnostic tools based on pre-rank functions, which are projections that reduce multivariate prediction-observation pairs to univariate summaries to detect specific types of miscalibration. In this work, we go beyond diagnostics and introduce a general regularization framework to enforce multivariate calibration during training for arbitrary pre-rank functions. This framework encompasses existing approaches such as highest density region calibration and copula calibration. Our method enforces calibration by penalizing deviations of the projected probability integral transforms (PITs) from the uniform distribution, and can be added as a regularization term to the loss function of any probabilistic predictor. Specifically, we propose a regularization loss that jointly enforces both marginal and multivariate pre-rank calibration. We also introduce a new PCA-based pre-rank that captures calibration along directions of maximal variance in the predictive distribution, while also enabling dimensionality reduction. Across 18 real-world multi-output regression datasets, we show that unregularized models are consistently miscalibrated, and that our methods significantly improve calibration across all pre-rank functions without sacrificing predictive accuracy.
Authors: Christiaan M. Geldenhuys, G\"unther Tonitz, Thomas R. Niesler
Abstract: While recent sound event detection (SED) systems can identify baleen whale calls in marine audio, challenges related to false positive and minority-class detection persist. We propose the boundary proposal network (BPN), which extends an existing lightweight SED system. The BPN is inspired by work in image object detection and aims to reduce the number of false positive detections. It achieves this by using intermediate latent representations computed within the backbone classification model to gate the final output. When added to an existing SED system, the BPN achieves a 16.8 % absolute increase in precision, as well as 21.3 % and 9.4 % improvements in the F1-score for minority-class d-calls and bp-calls, respectively. We further consider two approaches to the selection of post-processing hyperparameters: a forward-search and a backward-search. By separately optimising event-level and frame-level hyperparameters, these two approaches lead to considerable performance improvements over parameters selected using empirical methods. The complete WhaleVAD-BPN system achieves a cross-validated development F1-score of 0.475, which is a 9.8 % absolute improvement over the baseline.
Authors: Qiang Liu, Wuganjing Song, Zhenzhou Lin, Feifan Chen, Qiaolong Cai, Chen Li, Yongduo Sui
Abstract: The reasoning capabilities of Large Language Models (LLMs) are typically developed through the single-turn reinforcement learning, whereas real-world applications often involve multi-turn interactions with human feedback, leading to a potential mismatch between training and deployment conditions. In this work, we study whether multi-turn training with human feedback is necessary for reasoning tasks. We compare conventional single-turn training with three multi-turn strategies and reach contrary conclusions to previous research. We find that models trained in a single-turn setting generalize effectively to both single- and multi-turn evaluations, while models trained with multi-turn strategies exhibit a significant degradation in single-turn reasoning performance. These results suggest that for tasks with complete information, robust single-turn training remains more effective and reliable, as multi-turn training with basic feedback provides limited benefits and can even degrade reasoning capabilities.