Authors: Zichuan Yang
Abstract: I investigate a stronger form of regularization by deactivating neurons for extended periods, a departure from the temporary changes of methods like Dropout. However, this long-term dynamism introduces a critical challenge: severe training instability when neurons are revived with random weights. To solve this, I propose the Lifecycle (LC) principle, a regularization mechanism centered on a key innovation: state memory. Instead of re-initializing a revived neuron, my method restores its parameters to their last known effective state. This process preserves learned knowledge and avoids destructive optimization shocks. My theoretical analysis reveals that the LC principle smooths the loss landscape, guiding optimization towards flatter minima associated with better generalization. Experiments on image classification benchmarks demonstrate that my method improves generalization and robustness. Crucially, ablation studies confirm that state memory is essential for achieving these gains.
Authors: Mazyar Taghavi, Rahman Farnoosh
Abstract: Protecting endangered wildlife from illegal poaching presents a critical challenge, particularly in vast and partially observable environments where real-time response is essential. This paper introduces a novel Expectation-Maximization (EM) based latent variable modeling approach in the context of Multi-Agent Reinforcement Learning (MARL) for Unmanned Aerial Vehicle (UAV) coordination in wildlife protection. By modeling hidden environmental factors and inter-agent dynamics through latent variables, our method enhances exploration and coordination under uncertainty.We implement and evaluate our EM-MARL framework using a custom simulation involving 10 UAVs tasked with patrolling protected habitats of the endangered Iranian leopard. Extensive experimental results demonstrate superior performance in detection accuracy, adaptability, and policy convergence when compared to standard algorithms such as Proximal Policy Optimization (PPO) and Deep Deterministic Policy Gradient (DDPG). Our findings underscore the potential of combining EM inference with MARL to improve decentralized decisionmaking in complex, high-stakes conservation scenarios. The full implementation, simulation environment, and training scripts are publicly available on GitHub.
Authors: Hunter Gittlin
Abstract: Class imbalance remains a fundamental challenge in machine learning, with traditional solutions often creating as many problems as they solve. We demonstrate that group-aware threshold calibration--setting different decision thresholds for different demographic groups--provides superior robustness compared to synthetic data generation methods. Through extensive experiments, we show that group-specific thresholds achieve 1.5-4% higher balanced accuracy than SMOTE and CT-GAN augmented models while improving worst-group balanced accuracy. Unlike single-threshold approaches that apply one cutoff across all groups, our group-aware method optimizes the Pareto frontier between balanced accuracy and worst-group balanced accuracy, enabling fine-grained control over group-level performance. Critically, we find that applying group thresholds to synthetically augmented data yields minimal additional benefit, suggesting these approaches are fundamentally redundant. Our results span seven model families including linear, tree-based, instance-based, and boosting methods, confirming that group-aware threshold calibration offers a simpler, more interpretable, and more effective solution to class imbalance.
Authors: Cheol Woo Kim, Shresth Verma, Mauricio Tec, Milind Tambe
Abstract: We study an LLM fine-tuning task for designing reward functions for sequential resource allocation problems in public health, guided by human preferences expressed in natural language. This setting presents a challenging testbed for alignment due to complex and ambiguous objectives and limited data availability. We propose DPO-PRO, a robust fine-tuning algorithm based on Direct Preference Optimization (DPO), which accounts for uncertainty in the preference distribution using a lightweight Distributionally Robust Optimization (DRO) formulation. Unlike prior DRO-based DPO methods, DPO-PRO is significantly less conservative. We evaluate DPO-PRO on a real-world maternal mobile health program operated by the non-profit organization ARMMAN, as well as on standard alignment benchmarks. Experimental results demonstrate that our method consistently improves robustness to noisy preference signals compared to existing DPO variants. Moreover, DPO-PRO achieves comparable performance to prior self-reflection-based baseline for reward function design, while requiring significantly lower inference-time cost.
Authors: Zhongzhu Zhou, Yibo Yang, Ziyan Chen, Fengxiang Bie, Haojun Xia, Xiaoxia Wu, Robert Wu, Ben Athiwaratkun, Bernard Ghanem, Shuaiwen Leon Song
Abstract: Policy gradient (PG) methods in reinforcement learning frequently utilize deep neural networks (DNNs) to learn a shared backbone of feature representations used to compute likelihoods in an action selection layer. Numerous studies have been conducted on the convergence and global optima of policy networks, but few have analyzed representational structures of those underlying networks. While training an optimal policy DNN, we observed that under certain constraints, a gentle structure resembling neural collapse, which we refer to as Action Collapse (AC), emerges. This suggests that 1) the state-action activations (i.e. last-layer features) sharing the same optimal actions collapse towards those optimal actions respective mean activations; 2) the variability of activations sharing the same optimal actions converges to zero; 3) the weights of action selection layer and the mean activations collapse to a simplex equiangular tight frame (ETF). Our early work showed those aforementioned constraints to be necessary for these observations. Since the collapsed ETF of optimal policy DNNs maximally separates the pair-wise angles of all actions in the state-action space, we naturally raise a question: can we learn an optimal policy using an ETF structure as a (fixed) target configuration in the action selection layer? Our analytical proof shows that learning activations with a fixed ETF as action selection layer naturally leads to the AC. We thus propose the Action Collapse Policy Gradient (ACPG) method, which accordingly affixes a synthetic ETF as our action selection layer. ACPG induces the policy DNN to produce such an ideal configuration in the action selection layer while remaining optimal. Our experiments across various OpenAI Gym environments demonstrate that our technique can be integrated into any discrete PG methods and lead to favorable reward improvements more quickly and robustly.
Authors: Saarang Panchavati, Corey Arnold, William Speier
Abstract: This work explores the potential of foundation models, specifically a Mamba-based selective state space model, for enhancing EEG analysis in neurological disorder diagnosis. EEG, crucial for diagnosing conditions like epilepsy, presents significant challenges due to its noisy, high-dimensional, and nonlinear nature. Traditional machine learning methods have made advances in automating EEG analysis but often fail to capture its complex spatio-temporal dynamics. Recent advances in deep learning, particularly in sequence modeling, offer new avenues for creating more generalized and expressive models capable of handling such complexities. By training a Mamba-based model on a large dataset containing seizure and non-seizure EEG recordings through a self-supervised reconstruction task followed by a seizure detection task, we demonstrate the model's effectiveness, achieving an AUROC of 0.72 on a held-out test set. This approach marks a significant step toward developing large-scale, clinically applicable foundation models for EEG data analysis.
Authors: Krishna Teja Chitty-Venkata, Sandeep Madireddy, Murali Emani, Venkatram Vishwanath
Abstract: Mixture-of-Experts (MoE) models scale efficiently by activating only a subset of experts per token, offering a computationally sparse alternative to dense architectures. While prior post-training optimizations, such as inter- and intra-expert pruning, reduce memory usage they provide limited gains in inference-time compute efficiency. Moreover, existing MoE architectures typically activate a fixed number of experts uniformly across all layers, resulting in redundant computation and suboptimal performance. In this work, we first demonstrate that MoE pruning strategies improve only the memory footprint but do not significantly improve inference performance on GPU using optimized frameworks such as vLLM. To address this, we introduce LExI, a data-free optimization technique that determines the optimal number of active experts per layer in a pretrained MoE model. LExI leverages only the model weights to estimate the relative importance of each layer and adaptively assigns the number of active experts accordingly per layer. Experiments on state-of-the-art language and vision MoE benchmarks demonstrate that LExI significantly outperforms traditional MoE pruning approaches in terms of inference efficiency with negligible accuracy loss. For example, using LExI, Qwen1.5-MoE achieves the same throughput on Nvidia H100 GPU with 10% better accuracy than traditional expert pruning.
Authors: Arnab Mazumder, Javier E. Santos, Noah Hobbs, Mohamed Mehana, Daniel O'Malley
Abstract: We present the Transparent Earth, a transformer-based architecture for reconstructing subsurface properties from heterogeneous datasets that vary in sparsity, resolution, and modality, where each modality represents a distinct type of observation (e.g., stress angle, mantle temperature, tectonic plate type). The model incorporates positional encodings of observations together with modality encodings, derived from a text embedding model applied to a description of each modality. This design enables the model to scale to an arbitrary number of modalities, making it straightforward to add new ones not considered in the initial design. We currently include eight modalities spanning directional angles, categorical classes, and continuous properties such as temperature and thickness. These capabilities support in-context learning, enabling the model to generate predictions either with no inputs or with an arbitrary number of additional observations from any subset of modalities. On validation data, this reduces errors in predicting stress angle by more than a factor of three. The proposed architecture is scalable and demonstrates improved performance with increased parameters. Together, these advances make the Transparent Earth an initial foundation model for the Earth's subsurface that ultimately aims to predict any subsurface property anywhere on Earth.
Authors: Alejandro Rodriguez Dominguez, Muhammad Shahzad, Xia Hong
Abstract: Existing approaches to predictive uncertainty rely either on multi-hypothesis prediction, which promotes diversity but lacks principled aggregation, or on ensemble learning, which improves accuracy but rarely captures the structured ambiguity. This implicitly means that a unified framework consistent with the loss geometry remains absent. The Structured Basis Function Network addresses this gap by linking multi-hypothesis prediction and ensembling through centroidal aggregation induced by Bregman divergences. The formulation applies across regression and classification by aligning predictions with the geometry of the loss, and supports both a closed-form least-squares estimator and a gradient-based procedure for general objectives. A tunable diversity mechanism provides parametric control of the bias-variance-diversity trade-off, connecting multi-hypothesis generalisation with loss-aware ensemble aggregation. Experiments validate this relation and use the mechanism to study the complexity-capacity-diversity trade-off across datasets of increasing difficulty with deep-learning predictors.
Authors: Howard Dai, Nyambura Njenga, Benjamin Whitsett, Catherine Ma, Darwin Deng, Sara de \'Angel, Alexandre Van Tassel, Siddharth Viswanath, Ryan Pellico, Ian Adelstein, Smita Krishnaswamy
Abstract: We propose a novel framework for pre-training Graph Neural Networks (GNNs) by inductively learning Laplacian eigenvectors. Traditional Message Passing Neural Networks (MPNNs) often struggle to capture global and regional graph structure due to over-smoothing risk as network depth increases. Because the low-frequency eigenvectors of the graph Laplacian matrix encode global information, pre-training GNNs to predict these eigenvectors encourages the network to naturally learn large-scale structural patterns over each graph. Empirically, we show that models pre-trained via our framework outperform baseline models on a variety of graph structure-based tasks. While most existing pre-training methods focus on domain-specific tasks like node or edge feature reconstruction, our self-supervised pre-training framework is structure-based and highly flexible. Eigenvector-learning can be applied to all graph-based datasets, and can be used with synthetic features when task-specific data is sparse.
Authors: Trang Nguyen, Jackson Michaels, Madalina Fiterau, David Jensen
Abstract: This paper highlights the challenge of decomposing conflict detection from conflict resolution in Vision-Language Models (VLMs) and presents potential approaches, including using a supervised metric via linear probes and group-based attention pattern analysis. We conduct a mechanistic investigation of LLaVA-OV-7B, a state-of-the-art VLM that exhibits diverse resolution behaviors when faced with conflicting multimodal inputs. Our results show that a linearly decodable conflict signal emerges in the model's intermediate layers and that attention patterns associated with conflict detection and resolution diverge at different stages of the network. These findings support the hypothesis that detection and resolution are functionally distinct mechanisms. We discuss how such decomposition enables more actionable interpretability and targeted interventions for improving model robustness in challenging multimodal settings.
Authors: Naman Deep Singh, Maximilian M\"uller, Francesco Croce, Matthias Hein
Abstract: Unlearning in large language models (LLMs) involves precisely removing specific information from a pre-trained model. This is crucial to ensure safety of LLMs by deleting private data or harmful knowledge acquired during pre-training. However, existing unlearning methods often fall short when subjected to thorough evaluation. To overcome this, we introduce JensUn, where we leverage the Jensen-Shannon Divergence as the training objective for both forget and retain sets for more stable and effective unlearning dynamics compared to commonly used loss functions. In extensive experiments, JensUn achieves better forget-utility trade-off than competing methods, and even demonstrates strong resilience to benign relearning. Additionally, for a precise unlearning evaluation, we introduce LKF, a curated dataset of lesser-known facts that provides a realistic unlearning scenario. Finally, to comprehensively test unlearning methods, we propose (i) employing an LLM as semantic judge instead of the standard ROUGE score, and (ii) using worst-case unlearning evaluation over various paraphrases and input formats. Our improved evaluation framework reveals that many existing methods are less effective than previously thought.
Authors: Towhidul Islam, Md Sumon Ali
Abstract: Obesity is a critical global health issue driven by dietary, physiological, and environmental factors, and is strongly associated with chronic diseases such as diabetes, cardiovascular disorders, and cancer. Machine learning has emerged as a promising approach for early obesity risk prediction, yet a comparative evaluation of ensemble techniques -- particularly hybrid majority voting and ensemble stacking -- remains limited. This study aims to compare hybrid majority voting and ensemble stacking methods for obesity risk prediction, identifying which approach delivers higher accuracy and efficiency. The analysis seeks to highlight the complementary strengths of these ensemble techniques in guiding better predictive model selection for healthcare applications. Two datasets were utilized to evaluate three ensemble models: Majority Hard Voting, Weighted Hard Voting, and Stacking (with a Multi-Layer Perceptron as meta-classifier). A pool of nine Machine Learning (ML) algorithms, evaluated across a total of 50 hyperparameter configurations, was analyzed to identify the top three models to serve as base learners for the ensemble methods. Preprocessing steps involved dataset balancing, and outlier detection, and model performance was evaluated using Accuracy and F1-Score. On Dataset-1, weighted hard voting and stacking achieved nearly identical performance (Accuracy: 0.920304, F1: 0.920070), outperforming majority hard voting. On Dataset-2, stacking demonstrated superior results (Accuracy: 0.989837, F1: 0.989825) compared to majority hard voting (Accuracy: 0.981707, F1: 0.981675) and weighted hard voting, which showed the lowest performance. The findings confirm that ensemble stacking provides stronger predictive capability, particularly for complex data distributions, while hybrid majority voting remains a robust alternative.
Authors: Sophia Sun, Rose Yu
Abstract: Conformal prediction has been explored as a general and efficient way to provide uncertainty quantification for time series. However, current methods struggle to handle time series data with change points - sudden shifts in the underlying data-generating process. In this paper, we propose a novel Conformal Prediction for Time-series with Change points (CPTC) algorithm, addressing this gap by integrating a model to predict the underlying state with online conformal prediction to model uncertainties in non-stationary time series. We prove CPTC's validity and improved adaptivity in the time series setting under minimum assumptions, and demonstrate CPTC's practical effectiveness on 6 synthetic and real-world datasets, showing improved validity and adaptivity compared to state-of-the-art baselines.
Authors: Siddharth Mansingh, James Amarel, Ragib Arnab, Arvind Mohan, Kamaljeet Singh, Gerd J. Kunde, Nicolas Hengartner, Benjamin Migliori, Emily Casleton, Nathan A. Debarledeben, Ayan Biswas, Diane Oyen, Earl Lawrence
Abstract: Partial Differential Equations (PDEs) are the bedrock for modern computational sciences and engineering, and inherently computationally expensive. While PDE foundation models have shown much promise for simulating such complex spatio-temporal phenomena, existing models remain constrained by the pretraining datasets and struggle with auto-regressive rollout performance, especially in out-of-distribution (OOD) cases. Furthermore, they have significant compute and training data requirements which hamper their use in many critical applications. Inspired by recent advances in ``thinking" strategies used in large language models (LLMs), we introduce the first test-time computing (TTC) strategy for PDEs that utilizes computational resources during inference to achieve more accurate predictions with fewer training samples and smaller models. We accomplish this with two types of reward models that evaluate predictions of a stochastic based model for spatio-temporal consistency. We demonstrate this method on compressible Euler-equation simulations from the PDEGym benchmark and show that TTC captures improved predictions relative to standard non-adaptive auto-regressive inference. This TTC framework marks a foundational step towards more advanced reasoning algorithms or PDE modeling, inluding building reinforcement-learning-based approaches, potentially transforming computational workflows in physics and engineering.
Authors: Carlo Fabrizio, Gianvito Losapio, Marco Mussi, Alberto Maria Metelli, Marcello Restelli
Abstract: The necessary integration of renewable energy sources, combined with the expanding scale of power networks, presents significant challenges in controlling modern power grids. Traditional control systems, which are human and optimization-based, struggle to adapt and to scale in such an evolving context, motivating the exploration of more dynamic and distributed control strategies. This work advances a graph-based distributed reinforcement learning framework for real-time, scalable grid management. The proposed architecture consists of a network of distributed low-level agents acting on individual power lines and coordinated by a high-level manager agent. A Graph Neural Network (GNN) is employed to encode the network's topological information within the single low-level agent's observation. To accelerate convergence and enhance learning stability, the framework integrates imitation learning and potential-based reward shaping. In contrast to conventional decentralized approaches that decompose only the action space while relying on global observations, this method also decomposes the observation space. Each low-level agent acts based on a structured and informative local view of the environment constructed through the GNN. Experiments on the Grid2Op simulation environment show the effectiveness of the approach, which consistently outperforms the standard baseline commonly adopted in the field. Additionally, the proposed model proves to be much more computationally efficient than the simulation-based Expert method.
Authors: Vikas Kashtriya, Pardeep Singh
Abstract: Class imbalance remains a critical challenge in machine learning (ML), particularly in the medical domain, where underrepresented minority classes lead to biased models and reduced predictive performance. This study introduces Quantum-Inspired SMOTE (QI-SMOTE), a novel data augmentation technique that enhances the performance of ML classifiers, including Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), k-Nearest Neighbors (KNN), Gradient Boosting (GB), and Neural Networks, by leveraging quantum principles such as quantum evolution and layered entanglement. Unlike conventional oversampling methods, QI-SMOTE generates synthetic instances that preserve complex data structures, improving model generalization and classification accuracy. We validate QI-SMOTE on the MIMIC-III and MIMIC-IV datasets, using mortality detection as a benchmark task due to their clinical significance and inherent class imbalance. We compare our method against traditional oversampling techniques, including Borderline-SMOTE, ADASYN, SMOTE-ENN, SMOTE-TOMEK, and SVM-SMOTE, using key performance metrics such as Accuracy, F1-score, G-Mean, and AUC-ROC. The results demonstrate that QI-SMOTE significantly improves the effectiveness of ensemble methods (RF, GB, ADA), kernel-based models (SVM), and deep learning approaches by producing more informative and balanced training data. By integrating quantum-inspired transformations into the ML pipeline, QI-SMOTE not only mitigates class imbalance but also enhances the robustness and reliability of predictive models in medical diagnostics and decision-making. This study highlights the potential of quantum-inspired resampling techniques in advancing state-of-the-art ML methodologies.
Authors: Pracheta Amaranath, Vinitra Muralikrishnan, Amit Sharma, David D. Jensen
Abstract: Generating synthetic datasets that accurately reflect real-world observational data is critical for evaluating causal estimators, but remains a challenging task. Existing generative methods offer a solution by producing synthetic datasets anchored in the observed data (source data) while allowing variation in key parameters such as the treatment effect and amount of confounding bias. However, existing methods typically require users to provide point estimates of such parameters (rather than distributions) and fixed estimates (rather than estimates that can be improved with reference to the source data). This denies users the ability to express uncertainty over parameter values and removes the potential for posterior inference, potentially leading to unreliable estimator comparisons. We introduce simulation-based inference for causal evaluation (SBICE), a framework that models generative parameters as uncertain and infers their posterior distribution given a source dataset. Leveraging techniques in simulation-based inference, SBICE identifies parameter configurations that produce synthetic datasets closely aligned with the source data distribution. Empirical results demonstrate that SBICE improves the reliability of estimator evaluations by generating more realistic datasets, which supports a robust and data-consistent approach to causal benchmarking under uncertainty.
Authors: Jaliya L. Wijayaraja, Janaka L. Wijekoon, Malitha Wijesundara
Abstract: Detecting elephants through seismic signals is an emerging research topic aimed at developing solutions for Human-Elephant Conflict (HEC). Despite the promising results, such solutions heavily rely on manual classification of elephant footfalls, which limits their applicability for real-time classification in natural settings. To address this limitation and build on our previous work, this study introduces a classification framework targeting resource-constrained implementations, prioritizing both accuracy and computational efficiency. As part of this framework, a novel event detection technique named Contextually Customized Windowing (CCW), tailored specifically for detecting elephant footfalls, was introduced, and evaluations were conducted by comparing it with the Short-Term Average/Long-Term Average (STA/LTA) method. The yielded results show that the maximum validated detection range was 155.6 m in controlled conditions and 140 m in natural environments. Elephant footfall classification using Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel demonstrated superior performance across multiple settings, achieving an accuracy of 99% in controlled environments, 73% in natural elephant habitats, and 70% in HEC-prone human habitats, the most challenging scenario. Furthermore, feature impact analysis using explainable AI identified the number of Zero Crossings and Dynamic Time Warping (DTW) Alignment Cost as the most influential factors in all experiments, while Predominant Frequency exhibited significant influence in controlled settings.
Authors: Kunal Kumar, Muhammad Ashad Kabir, Luke Donnan, Sayed Ahmed
Abstract: Offloading footwear helps prevent and treat diabetic foot ulcers (DFUs) by lowering plantar pressure (PP), yet prescription decisions remain fragmented: feature selection varies, personalization is limited, and evaluation practices differ. We performed a narrative review of 45 studies (12 guidelines/protocols, 25 knowledge-based systems, 8 machine-learning applications) published to Aug 2025. We thematically analyzed knowledge type, decision logic, evaluation methods, and enabling technologies. Guidelines emphasize PP thresholds (<=200 kPa or >=25--30\% reduction) but rarely yield actionable, feature-level outputs. Knowledge-based systems use rule- and sensor-driven logic, integrating PP monitoring, adherence tracking, and usability testing. ML work introduces predictive, optimization, and generative models with high computational accuracy but limited explainability and clinical validation. Evaluation remains fragmented: protocols prioritize biomechanical tests; knowledge-based systems assess usability/adherence; ML studies focus on technical accuracy with weak linkage to long-term outcomes. From this synthesis we propose a five-part CDSS framework: (1) a minimum viable dataset; (2) a hybrid architecture combining rules, optimization, and explainable ML; (3) structured feature-level outputs; (4) continuous validation and evaluation; and (5) integration with clinical and telehealth workflows. This framework aims to enable scalable, patient-centered CDSSs for DFU care; prioritizing interoperable datasets, explainable models, and outcome-focused evaluation will be key to clinical adoption.
Authors: Shih-Peng Huang, Nontawat Charoenphakdee, Yuta Tsuboi, Yong-Bin Zhuang, Wenwen Li
Abstract: Ensemble method is considered the gold standard for uncertainty quantification (UQ) for machine learning interatomic potentials (MLIPs). However, their high computational cost can limit its practicality. Alternative techniques, such as Monte Carlo dropout and deep kernel learning, have been proposed to improve computational efficiency; however, some of these methods cannot be applied to already trained models and may affect the prediction accuracy. In this paper, we propose a simple and efficient post-hoc framework for UQ that leverages the descriptor of a trained graph neural network potential to estimate residual errors. We refer to this method as post-hoc descriptor-based residual-based learning (PDRL). PDRL models the discrepancy between MLIP predictions and ground truth values, allowing these residuals to act as proxies for prediction uncertainty. We explore multiple variants of PDRL and benchmark them against established UQ methods, evaluating both their effectiveness and limitations.
Authors: Erik M. Lintunen
Abstract: In self-supervised reinforcement learning (RL), one of the key challenges is learning a diverse set of skills to prepare agents for unknown future tasks. Despite impressive advances, scalability and evaluation remain prevalent issues. Regarding scalability, the search for meaningful skills can be obscured by high-dimensional feature spaces, where relevant features may vary across downstream task domains. For evaluating skill diversity, defining what constitutes "diversity" typically requires a hard commitment to a specific notion of what it means for skills to be diverse, potentially leading to inconsistencies in how skill diversity is understood, making results across different approaches hard to compare, and leaving many forms of diversity unexplored. To address these issues, we adopt a measure of sample diversity that translates ideas from ecology to machine learning -- the Vendi Score -- allowing the user to specify and evaluate any desired form of diversity. We demonstrate how this metric facilitates skill evaluation and introduce VendiRL, a unified framework for learning diversely diverse sets of skills. Given distinct similarity functions, VendiRL motivates distinct forms of diversity, which could support skill-diversity pretraining in new and richly interactive environments where optimising for various forms of diversity may be desirable.
Authors: Chen Zeng, Tiehang Xu, Qiao Wang
Abstract: Conventional neural networks frequently face challenges in spectral analysis of signals. To address this challenge, Fourier neural networks (FNNs) and similar approaches integrate components of Fourier series into the structure of neural networks. Nonetheless, a significant hurdle is often overlooked: the superposition of periodic signals does not necessarily result in a periodic signal. For example, when forecasting almost periodic functions composed of signals with incommensurate frequencies, traditional models such as Autoregressive Integrated Moving Average (ARIMA) frequently outperform most neural networks including large language models (LLMs). To tackle this goal, we propose Autoregressive-Weight-Enhanced AR-KAN, a hybrid model that combines the benefits of both methods. Using the Universal Myopic Mapping Theorem, we apply a Kolmogorov-Arnold Network (KAN) for the static nonlinear part and include memory through a pre-trained AR component, which can be explained to retain the most useful information while eliminating redundancy. Experimental data indicates that AR-KAN delivers superior results on $72\%$ of real-world datasets.
Authors: Kaoru Otsuka, Yuki Takezawa, Makoto Yamada
Abstract: Federated Learning (FL) allows distributed model training across multiple clients while preserving data privacy, but it remains vulnerable to Byzantine clients that exhibit malicious behavior. While existing Byzantine-robust FL methods provide strong convergence guarantees (e.g., to a stationary point in expectation) under Byzantine attacks, they typically assume full client participation, which is unrealistic due to communication constraints and client availability. Under partial participation, existing methods fail immediately after the sampled clients contain a Byzantine majority, creating a fundamental challenge for sparse communication. First, we introduce delayed momentum aggregation, a novel principle where the server aggregates the most recently received gradients from non-participating clients alongside fresh momentum from active clients. Our optimizer D-Byz-SGDM (Delayed Byzantine-robust SGD with Momentum) implements this delayed momentum aggregation principle for Byzantine-robust FL with partial participation. Then, we establish convergence guarantees that recover previous full participation results and match the fundamental lower bounds we prove for the partial participation setting. Experiments on deep learning tasks validated our theoretical findings, showing stable and robust training under various Byzantine attacks.
Authors: Minxin Zhang, Yuxuan Liu, Hayden Schaeffer
Abstract: The recently proposed Muon optimizer updates weight matrices via orthogonalized momentum and has demonstrated strong empirical success in large language model training. However, it remains unclear how to determine the learning rates for such orthogonalized updates. AdaGrad, by contrast, is a widely used adaptive method that scales stochastic gradients by accumulated past gradients. We propose a new algorithm, AdaGO, which combines a norm-based AdaGrad-type stepsize with an orthogonalized update direction, bringing together the benefits of both approaches. Unlike other adaptive variants of Muon, AdaGO preserves the orthogonality of the update direction, which can be interpreted as a spectral descent direction, while adapting the stepsizes to the optimization landscape by scaling the direction with accumulated past gradient norms. The implementation of AdaGO requires only minimal modification to Muon, with a single additional scalar variable, the accumulated squared gradient norms, to be computed, making it computationally and memory efficient. Optimal theoretical convergence rates are established for nonconvex functions in both stochastic and deterministic settings under standard smoothness and unbiased bounded-variance noise assumptions. Empirical results on CIFAR-10 classification and function regression demonstrate that AdaGO outperforms Muon and Adam.
Authors: Hritik Arasu, Faisal R Jahangiri
Abstract: Sleep staging models often degrade when deployed on patients with unseen physiology or recording conditions. We propose a streaming, source-free test-time adaptation (TTA) recipe that combines entropy minimization (Tent) with Batch-Norm statistic refresh and two safety rails: an entropy gate to pause adaptation on uncertain windows and an EMA-based reset to reel back drift. On Sleep-EDF Expanded, using single-lead EEG (Fpz-Cz, 100 Hz, 30s epochs; R&K to AASM mapping), we show consistent gains over a frozen baseline at seconds-level latency and minimal memory, reporting per-stage metrics and Cohen's k. The method is model-agnostic, requires no source data or patient calibration, and is practical for on-device or bedside use.
Authors: Satyajit Mojumder, Pallock Halder, Tiana Tonge
Abstract: While multiple sensors are used for real-time monitoring in additive manufacturing, not all provide practical or reliable process insights. For example, high-speed X-ray imaging offers valuable spatial information about subsurface melt pool behavior but is costly and impractical for most industrial settings. In contrast, absorptivity data from low-cost photodiodes correlate with melt pool dynamics but is often too noisy for accurate prediction when used alone. In this paper, we propose a multimodal data fusion approach for predicting melt pool dynamics by combining high-fidelity X-ray data with low-fidelity absorptivity data in the Laser Powder Bed Fusion (LPBF) process. Our multimodal learning framework integrates convolutional neural networks (CNNs) for spatial feature extraction from X-ray data with recurrent neural networks (RNNs) for temporal feature extraction from absorptivity signals, using an early fusion strategy. The multimodal model is further used as a transfer learning model to fine-tune the RNN model that can predict melt pool dynamics only with absorptivity, with greater accuracy compared to the multimodal model. Results show that training with both modalities significantly improves prediction accuracy compared to using either modality alone. Furthermore, once trained, the model can infer melt pool characteristics using only absorptivity data, eliminating the need for expensive X-ray imaging. This multimodal fusion approach enables cost-effective, real-time monitoring and has broad applicability in additive manufacturing.
Authors: Zida Wu, Mathieu Lauriere, Matthieu Geist, Olivier Pietquin, Ankur Mehta
Abstract: Mean Field Games (MFGs) offer a powerful framework for studying large-scale multi-agent systems. Yet, learning Nash equilibria in MFGs remains a challenging problem, particularly when the initial distribution is unknown or when the population is subject to common noise. In this paper, we introduce an efficient deep reinforcement learning (DRL) algorithm designed to achieve population-dependent Nash equilibria without relying on averaging or historical sampling, inspired by Munchausen RL and Online Mirror Descent. The resulting policy is adaptable to various initial distributions and sources of common noise. Through numerical experiments on seven canonical examples, we demonstrate that our algorithm exhibits superior convergence properties compared to state-of-the-art algorithms, particularly a DRL version of Fictitious Play for population-dependent policies. The performance in the presence of common noise underscores the robustness and adaptability of our approach.
Authors: Bilge Taskin, Wenxiong Xie, Teddy Lazebnik
Abstract: Symbolic regression (SR) has emerged as a powerful tool for automated scientific discovery, enabling the derivation of governing equations from experimental data. A growing body of work illustrates the promise of integrating domain knowledge into the SR to improve the discovered equation's generality and usefulness. Physics-informed SR (PiSR) addresses this by incorporating domain knowledge, but current methods often require specialized formulations and manual feature engineering, limiting their adaptability only to domain experts. In this study, we leverage pre-trained Large Language Models (LLMs) to facilitate knowledge integration in PiSR. By harnessing the contextual understanding of LLMs trained on vast scientific literature, we aim to automate the incorporation of domain knowledge, reducing the need for manual intervention and making the process more accessible to a broader range of scientific problems. Namely, the LLM is integrated into the SR's loss function, adding a term of the LLM's evaluation of the SR's produced equation. We extensively evaluate our method using three SR algorithms (DEAP, gplearn, and PySR) and three pre-trained LLMs (Falcon, Mistral, and LLama 2) across three physical dynamics (dropping ball, simple harmonic motion, and electromagnetic wave). The results demonstrate that LLM integration consistently improves the reconstruction of physical dynamics from data, enhancing the robustness of SR models to noise and complexity. We further explore the impact of prompt engineering, finding that more informative prompts significantly improve performance.
Authors: Xinzhe Zheng, Zhen-Qun Yang, Haoran Xie, S. Joe Qin, Arlene Chen, Fangzhen Lin
Abstract: Large Language Models (LLMs) have demonstrated remarkable performance across a wide range of Natural Language Processing (NLP) tasks, but require substantial memory and computational resources. Binary quantization, which compresses model weights from 16-bit Brain Float to 1-bit representations in {-1, 1}, offers significant reductions in storage and inference costs. However, such aggressive quantization often leads to notable performance degradation compared to more conservative 4-bit quantization methods. In this research, we propose a novel optimization objective tailored for binary quantization, along with three algorithms designed to realize it effectively. Our method enhances blocked quantization by dynamically identifying optimal unstructured sub-matrices through adaptive grouping strategies. Experimental results demonstrate that our approach achieves an average bit length of just 1.007 bits, while maintaining high model quality. Specifically, our quantized LLaMA 3.2 3B model attains a perplexity of 8.23, remarkably close to the original 7.81, and surpasses previous SOTA BiLLM with a perplexity of only 123.90. Furthermore, our method is competitive with SOTA 4-bit approaches such as GPTQ in both performance and efficiency. The compression process is highly efficient, requiring only 14 seconds to quantize the full LLaMA 3.2 3B weights on a single CPU core, with the entire process completing in under 100 minutes and exhibiting embarrassingly parallel properties. Code - https://github.com/johnnyzheng0636/WGM_bi_quan
Authors: Sahil Rajesh Dhayalkar
Abstract: We extend the ReLU Transition Graph (RTG) framework into a comprehensive graph-theoretic model for understanding deep ReLU networks. In this model, each node represents a linear activation region, and edges connect regions that differ by a single ReLU activation flip, forming a discrete geometric structure over the network's functional behavior. We prove that RTGs at random initialization exhibit strong expansion, binomial degree distributions, and spectral properties that tightly govern generalization. These structural insights enable new bounds on capacity via region entropy and on generalization via spectral gap and edge-wise KL divergence. Empirically, we construct RTGs for small networks, measure their smoothness and connectivity properties, and validate theoretical predictions. Our results show that region entropy saturates under overparameterization, spectral gap correlates with generalization, and KL divergence across adjacent regions reflects functional smoothness. This work provides a unified framework for analyzing ReLU networks through the lens of discrete functional geometry, offering new tools to understand, diagnose, and improve generalization.
Authors: Xingyue Huang, Rishabh, Gregor Franke, Ziyi Yang, Jiamu Bai, Weijie Bai, Jinhe Bi, Zifeng Ding, Yiqun Duan, Chengyu Fan, Wendong Fan, Xin Gao, Ruohao Guo, Yuan He, Zhuangzhuang He, Xianglong Hu, Neil Johnson, Bowen Li, Fangru Lin, Siyu Lin, Tong Liu, Yunpu Ma, Hao Shen, Hao Sun, Beibei Wang, Fangyijie Wang, Hao Wang, Haoran Wang, Yang Wang, Yifeng Wang, Zhaowei Wang, Ziyang Wang, Yifan Wu, Zikai Xiao, Chengxing Xie, Fan Yang, Junxiao Yang, Qianshuo Ye, Ziyu Ye, Guangtao Zeng, Yuwen Ebony Zhang, Zeyu Zhang, Zihao Zhu, Bernard Ghanem, Philip Torr, Guohao Li
Abstract: Recent advances in Large Language Models (LLMs) have shown that their reasoning capabilities can be significantly improved through Reinforcement Learning with Verifiable Reward (RLVR), particularly in domains like mathematics and programming, where ground-truth correctness can be automatically evaluated. However, extending this success to other reasoning-intensive domains remains challenging due to the scarcity of high-quality, verifiable datasets and the high cost of human supervision. In this work, we introduce the Loong Project: an open-source framework for scalable synthetic data generation and verification across a diverse range of reasoning-intensive domains. The framework consists of two key components: (1) LoongBench, a curated seed dataset containing 8,729 human-vetted examples across 12 domains (e.g., Advanced Mathematics, Chemistry, Logic), each paired with executable code and rich metadata; and (2) LoongEnv, a modular synthetic data generation environment that supports multiple prompting strategies to produce new question-answer-code triples. Together, these components form an agent-environment loop that enables reinforcement learning, where an LLM-based agent is rewarded for generating Chain-of-Thought (CoT) solutions that align with code-executed answers. Empirically, we benchmark LoongBench on a broad suite of both open-source and proprietary LLMs to evaluate domain coverage and reveal performance bottlenecks. In addition, we conduct a comprehensive analysis of synthetic data generated by LoongEnv, examining correctness, difficulty, and diversity. Code and documentation are available at https://github.com/camel-ai/loong.
Authors: Yunfei Teng, Sixin Zhang
Abstract: While Sharpness-Aware Minimization (SAM) improves generalization in deep neural networks by minimizing both loss and sharpness, it suffers from inefficiency in distributed large-batch training. We present Landscape-Smoothed SAM (LSAM), a novel optimizer that preserves SAM's generalization advantages while offering superior efficiency. LSAM integrates SAM's adversarial steps with an asynchronous distributed sampling strategy, generating an asynchronous distributed sampling scheme, producing a smoothed sharpness-aware loss landscape for optimization. This design eliminates synchronization bottlenecks, accelerates large-batch convergence, and delivers higher final accuracy compared to data-parallel SAM.
Authors: Hankang Gu, Yuli Zhang, Chengming Wang, Ruiyuan Jiang, Ziheng Qiao, Pengfei Fan, Dongyao Jia
Abstract: Deep reinforcement learning (DRL) has become a popular approach in traffic signal control (TSC) due to its ability to learn adaptive policies from complex traffic environments. Within DRL-based TSC methods, two primary control paradigms are ``choose phase" and ``switch" strategies. Although the agent in the choose phase paradigm selects the next active phase adaptively, this paradigm may result in unexpected phase sequences for drivers, disrupting their anticipation and potentially compromising safety at intersections. Meanwhile, the switch paradigm allows the agent to decide whether to switch to the next predefined phase or extend the current phase. While this structure maintains a more predictable order, it can lead to unfair and inefficient phase allocations, as certain movements may be extended disproportionately while others are neglected. In this paper, we propose a DRL model, named Deep Hierarchical Cycle Planner (DHCP), to allocate the traffic signal cycle duration hierarchically. A high-level agent first determines the split of the total cycle time between the North-South (NS) and East-West (EW) directions based on the overall traffic state. Then, a low-level agent further divides the allocated duration within each major direction between straight and left-turn movements, enabling more flexible durations for the two movements. We test our model on both real and synthetic road networks, along with multiple sets of real and synthetic traffic flows. Empirical results show our model achieves the best performance over all datasets against baselines.
Authors: Li Yi, Qian Yang
Abstract: Liquid scintillation triple-to-doubly coincident ratio (TDCR) spectroscopy is widely adopted as a standard method for radionuclide quantification because of its inherent advantages such as high precision, self-calibrating capability, and independence from radioactive reference sources. However, multiradionuclide analysis via TDCR faces the challenges of limited automation and reliance on mixture-specific standards, which may not be easily available. Here, we present an Artificial Intelligence (AI) framework that combines numerical spectral simulation and deep learning for standard-free automated analysis. $\beta$ spectra for model training were generated using Geant4 simulations coupled with statistically modeled detector response sampling. A tailored neural network architecture, trained on this dataset covering various nuclei mix ratio and quenching scenarios, enables autonomous resolution of individual radionuclide activities and detecting efficiency through end-to-end learning paradigms. The model delivers consistent high accuracy across tasks: activity proportions (mean absolute error = 0.009), detection efficiencies (mean absolute error = 0.002), and spectral reconstruction (Structural Similarity Index = 0.9998), validating its physical plausibility for quenched $\beta$ spectroscopy. This AI-driven methodology exhibits significant potential for automated safety-compliant multiradionuclide analysis with robust generalization, real-time processing capabilities, and engineering feasibility, particularly in scenarios where reference materials are unavailable or rapid field analysis is required.
Authors: Helge Spieker, J{\o}rn Eirik Betten, Arnaud Gotlieb, Nadjib Lazaar, Nassim Belmecheri
Abstract: Explainable AI (XAI) is essential for validating and trusting models in safety-critical applications like autonomous driving. However, the reliability of XAI is challenged by the Rashomon effect, where multiple, equally accurate models can offer divergent explanations for the same prediction. This paper provides the first empirical quantification of this effect for the task of action prediction in real-world driving scenes. Using Qualitative Explainable Graphs (QXGs) as a symbolic scene representation, we train Rashomon sets of two distinct model classes: interpretable, pair-based gradient boosting models and complex, graph-based Graph Neural Networks (GNNs). Using feature attribution methods, we measure the agreement of explanations both within and between these classes. Our results reveal significant explanation disagreement. Our findings suggest that explanation ambiguity is an inherent property of the problem, not just a modeling artifact.
Authors: Serra Aksoy
Abstract: Attribution methods explain neural network predictions by identifying influential input features, but their evaluation suffers from threshold selection bias that can reverse method rankings and undermine conclusions. Current protocols binarize attribution maps at single thresholds, where threshold choice alone can alter rankings by over 200 percentage points. We address this flaw with a threshold-free framework that computes Area Under the Curve for Intersection over Union (AUC-IoU), capturing attribution quality across the full threshold spectrum. Evaluating seven attribution methods on dermatological imaging, we show single-threshold metrics yield contradictory results, while threshold-free evaluation provides reliable differentiation. XRAI achieves 31% improvement over LIME and 204% over vanilla Integrated Gradients, with size-stratified analysis revealing performance variations up to 269% across lesion scales. These findings establish methodological standards that eliminate evaluation artifacts and enable evidence-based method selection. The threshold-free framework provides both theoretical insight into attribution behavior and practical guidance for robust comparison in medical imaging and beyond.
Authors: Taiga Saito, Yu Otake, Stephen Wu
Abstract: This paper presents a novel application of the Tabular Prior-Data Fitted Network (TabPFN) - a transformer-based foundation model for tabular data - to geotechnical site characterization problems defined in the GEOAI benchmark BM/AirportSoilProperties/2/2025. Two tasks are addressed: (1) predicting the spatial variation of undrained shear strength (su) across borehole depth profiles, and (2) imputing missing mechanical parameters in a dense-site dataset. We apply TabPFN in a zero-training, few-shot, in-context learning setting - without hyper-parameter tuning - and provide it with additional context from the big indirect database (BID). The study demonstrates that TabPFN, as a general-purpose foundation model, achieved superior accuracy and well-calibrated predictive distributions compared to a conventional hierarchical Bayesian model (HBM) baseline, while also offering significant gains in inference efficiency. In Benchmark Problem #1 (spatial su prediction), TabPFN outperformed the HBM in prediction accuracy and delivered an order-of-magnitude faster runtime. In Benchmark Problem #2 (missing mechanical parameter imputation), TabPFN likewise achieved lower RMSE for all target parameters with well-quantified uncertainties, though its cumulative computation cost was higher than HBM's due to its one-variable-at-a-time inference. These results mark the first successful use of a tabular foundation model in geotechnical modeling, suggesting a potential paradigm shift in probabilistic site characterization.
Authors: Kiara Stempel, Mattia Cerrato, Stefan Kramer
Abstract: Decision trees have been studied extensively in the context of fairness, aiming to maximize prediction performance while ensuring non-discrimination against different groups. Techniques in this space usually focus on imposing constraints at training time, constraining the search space so that solutions which display unacceptable values of relevant metrics are not considered, discarded, or discouraged. If we assume one target variable y and one sensitive attribute s, the design space of tree learning algorithms can be spanned as follows: (i) One can have one tree T that is built using an objective function that is a function of y, s, and T. For instance, one can build a tree based on the weighted information gain regarding y (maximizing) and s (minimizing). (ii) The second option is to have one tree model T that uses an objective function in y and T and a constraint on s and T. Here, s is no longer part of the objective, but part of a constraint. This can be achieved greedily by aborting a further split as soon as the condition that optimizes the objective in y fails to satisfy the constraint on s. A simple way to explore other splits is to backtrack during tree construction once a fairness constraint is violated. (iii) The third option is to have two trees T_y and T_s, one for y and one for s, such that the tree structure for y and s does not have to be shared. In this way, information regarding y and regarding s can be used independently, without having to constrain the choices in tree construction by the mutual information between the two variables. Quite surprisingly, of the three options, only the first one and the greedy variant of the second have been studied in the literature so far. In this paper, we introduce the above two additional options from that design space and characterize them experimentally on multiple datasets.
Authors: Zeqiang Zhang, Fabian Wurzberger, Gerrit Schmid, Sebastian Gottwald, Daniel A. Braun
Abstract: Reinforcement learning faces significant challenges when applied to tasks characterized by sparse reward structures. Although imitation learning, within the domain of supervised learning, offers faster convergence, it relies heavily on human-generated demonstrations. Recently, Goal-Conditioned Supervised Learning (GCSL) has emerged as a potential solution by enabling self-imitation learning for autonomous systems. By strategically relabelling goals, agents can derive policy insights from their own experiences. Despite the successes of this framework, it presents two notable limitations: (1) Learning exclusively from self-generated experiences can exacerbate the agents' inherent biases; (2) The relabelling strategy allows agents to focus solely on successful outcomes, precluding them from learning from their mistakes. To address these issues, we propose a novel model that integrates contrastive learning principles into the GCSL framework to learn from both success and failure. Through empirical evaluations, we demonstrate that our algorithm overcomes limitations imposed by agents' initial biases and thereby enables more exploratory behavior. This facilitates the identification and adoption of effective policies, leading to superior performance across a variety of challenging environments.
Authors: Yuxuan Gu, Wuyang Zhou, Giorgos Iacovides, Danilo Mandic
Abstract: Parameter-Efficient Fine-Tuning (PEFT) methods, such as Low-Rank Adaptation (LoRA), have significantly reduced the number of trainable parameters needed in fine-tuning large language models (LLMs). Subsequent developments of LoRA-style adapters have diverged into two main directions: (1) enhancing model expressivity with high-rank adapters, and (2) pushing for further parameter reduction, as exemplified by vector-based methods. However, these approaches present a trade-off, as achieving the expressivity of high-rank weight updates typically comes at the cost of sacrificing the extreme parameter efficiency offered by vector-based techniques. To address this issue, we propose a vector-based random \underline{\textbf{Te}}nsor network for high-\underline{\textbf{R}}ank \underline{\textbf{A}}daptation (TeRA), a novel PEFT method that achieves high-rank weight updates while retaining the parameter efficiency of vector-based PEFT adapters. This is achieved by parameterizing the tensorized weight update matrix as a Tucker-like tensor network (TN), in which large randomly initialized factors are frozen and shared across layers, while only small layer-specific scaling vectors, formed by entries in diagonal factor matrices, are trained. This design effectively decouples the rank of the weight update matrix from the number of trainable parameters. Comprehensive experiments demonstrate that TeRA matches or even outperforms high-rank adapters, while requiring a trainable parameter count similar to vector-based methods. Theoretical analysis and ablation studies further validate the effectiveness of our approach.
Authors: Harald Vilhelm Skat-R{\o}rdam, Sneha Das, Kathrine Sofie Rasmussen, Nicole Nadine L{\o}nfeldt, Line Clemmensen
Abstract: Accurate evaluation of event detection in time series is essential for applications such as stress monitoring with wearable devices, where ground truth is typically annotated as single-point events, even though the underlying phenomena are gradual and temporally diffused. Standard metrics like F1 and point-adjusted F1 (F1$_{pa}$) often misrepresent model performance in such real-world, imbalanced datasets. We introduce a window-based F1 metric (F1$_w$) that incorporates temporal tolerance, enabling a more robust assessment of event detection when exact alignment is unrealistic. Empirical analysis in three physiological datasets, two in-the-wild (ADARP, Wrist Angel) and one experimental (ROAD), indicates that F1$_w$ reveals meaningful model performance patterns invisible to conventional metrics, while its window size can be adapted to domain knowledge to avoid overestimation. We show that the choice of evaluation metric strongly influences the interpretation of model performance: using predictions from TimesFM, only our temporally tolerant metrics reveal statistically significant improvements over random and null baselines in the two in-the-wild use cases. This work addresses key gaps in time series evaluation and provides practical guidance for healthcare applications where requirements for temporal precision vary by context.
Authors: Pujitha Mamillapalli, Yoghitha Ramamoorthi, Abhinav Kumar, Tomoki Murakami, Tomoaki Ogawa, Yasushi Takatori
Abstract: The increasing demand for high data rates and seamless connectivity in wireless systems has sparked significant interest in reconfigurable intelligent surfaces (RIS) and artificial intelligence-based wireless applications. RIS typically comprises passive reflective antenna elements that control the wireless propagation environment by adequately tuning the phase of the reflective elements. The allocation of RIS elements to multipleuser equipment (UEs) is crucial for efficiently utilizing RIS. In this work, we formulate a joint optimization problem that optimizes the RIS phase configuration and resource allocation under an $\alpha$-fair scheduling framework and propose an efficient way of allocating RIS elements. Conventional iterative optimization methods, however, suffer from exponentially increasing computational complexity as the number of RIS elements increases and also complicate the generation of training labels for supervised learning. To overcome these challenges, we propose a five-layer fully connected neural network (FNN) combined with a preprocessing technique to significantly reduce input dimensionality, lower computational complexity, and enhance scalability. The simulation results show that our proposed NN-based solution reduces computational overhead while significantly improving system throughput by 6.8% compared to existing RIS element allocation schemes. Furthermore, the proposed system achieves better performance while reducing computational complexity, making it significantly more scalable than the iterative optimization algorithms.
Authors: Gianmarco De Vita, Nargiz Humbatova, Paolo Tonella
Abstract: Testing Deep Learning (DL)-based systems is an open challenge. Although it is relatively easy to find inputs that cause a DL model to misbehave, the grouping of inputs by features that make the DL model under test fail is largely unexplored. Existing approaches for DL testing introduce perturbations that may focus on specific failure-inducing features, while neglecting others that belong to different regions of the feature space. In this paper, we create an explicit topographical map of the input feature space. Our approach, named TopoMap, is both black-box and model-agnostic as it relies solely on features that characterise the input space. To discriminate the inputs according to the specific features they share, we first apply dimensionality reduction to obtain input embeddings, which are then subjected to clustering. Each DL model might require specific embedding computations and clustering algorithms to achieve a meaningful separation of inputs into discriminative groups. We propose a novel way to evaluate alternative configurations of embedding and clustering techniques. We used a deep neural network (DNN) as an approximation of a human evaluator who could tell whether a pair of clusters can be discriminated based on the features of the included elements. We use such a DNN to automatically select the optimal topographical map of the inputs among all those that are produced by different embedding/clustering configurations. The evaluation results show that the maps generated by TopoMap consist of distinguishable and meaningful regions. In addition, we evaluate the effectiveness of TopoMap using mutation analysis. In particular, we assess whether the clusters in our topographical map allow for an effective selection of mutation-killing inputs. Experimental results show that our approach outperforms random selection by 35% on average on killable mutants; by 61% on non-killable ones.
Authors: Yiming Yao, Fei Liu, Liang Zhao, Xi Lin, Qingfu Zhang
Abstract: Expensive multi-objective optimization is a prevalent and crucial concern in many real-world scenarios, where sample-efficiency is vital due to the limited evaluations to recover the true Pareto front for decision making. Existing works either involve rebuilding Gaussian process surrogates from scratch for each objective in each new problem encountered, or rely on extensive past domain experiments for pre-training deep learning models, making them hard to generalize and impractical to cope with various emerging applications in the real world. To address this issue, we propose a new paradigm named FoMEMO (Foundation Models for Expensive Multi-objective Optimization), which enables the establishment of a foundation model conditioned on any domain trajectory and user preference, and facilitates fast in-context optimization based on the predicted preference-wise aggregation posteriors. Rather than accessing extensive domain experiments in the real world, we demonstrate that pre-training the foundation model with a diverse set of hundreds of millions of synthetic data can lead to superior adaptability to unknown problems, without necessitating any subsequent model training or updates in the optimization process. We evaluate our method across a variety of synthetic benchmarks and real-word applications, and demonstrate its superior generality and competitive performance compared to existing methods.
Authors: Daniel Raggi, Gem Stapleton, Mateja Jamnik, Aaron Stockdill, Grecia Garcia Garcia, Peter C-H. Cheng
Abstract: Representation choice is of fundamental importance to our ability to communicate and reason effectively. A major unsolved problem, addressed in this paper, is how to devise \textit{representational-system (RS) agnostic} techniques that drive representation transformation and choice. We present a novel calculus, called \textit{structure transfer}, that enables representation transformation across diverse RSs. Specifically, given a \textit{source} representation drawn from a source RS, the rules of structure transfer allow us to generate a \textit{target} representation for a target RS. The generality of structure transfer comes in part from its ability to ensure that the source representation and the generated target representation satisfy \textit{any} specified relation (such as semantic equivalence). This is done by exploiting \textit{schemas}, which encode knowledge about RSs. Specifically, schemas can express \textit{preservation of information} across relations between any pair of RSs, and this knowledge is used by structure transfer to derive a structure for the target representation which ensures that the desired relation holds. We formalise this using Representational Systems Theory~\cite{raggi2022rst}, building on the key concept of a \textit{construction space}. The abstract nature of construction spaces grants them the generality to model RSs of diverse kinds, including formal languages, geometric figures and diagrams, as well as informal notations. Consequently, structure transfer is a system-agnostic calculus that can be used to identify alternative representations in a wide range of practical settings.
Authors: Minjung Park, Gyuyeon Na, Soyoun Kim, Sunyoung Moon, HyeonJeong Cha, Sangmi Chai
Abstract: Abnormal cryptocurrency transactions - such as mixing services, fraudulent transfers, and pump-and-dump operations -- pose escalating risks to financial integrity but remain notoriously difficult to detect due to class imbalance, temporal volatility, and complex network dependencies. Existing approaches are predominantly model-centric and post hoc, flagging anomalies only after they occur and thus offering limited preventive value. This paper introduces HyPV-LEAD (Hyperbolic Peak-Valley Lead-time Enabled Anomaly Detection), a data-driven early-warning framework that explicitly incorporates lead time into anomaly detection. Unlike prior methods, HyPV-LEAD integrates three innovations: (1) window-horizon modeling to guarantee actionable lead-time alerts, (2) Peak-Valley (PV) sampling to mitigate class imbalance while preserving temporal continuity, and (3) hyperbolic embedding to capture the hierarchical and scale-free properties of blockchain transaction networks. Empirical evaluation on large-scale Bitcoin transaction data demonstrates that HyPV-LEAD consistently outperforms state-of-the-art baselines, achieving a PR-AUC of 0.9624 with significant gains in precision and recall. Ablation studies further confirm that each component - PV sampling, hyperbolic embedding, and structural-temporal modeling - provides complementary benefits, with the full framework delivering the highest performance. By shifting anomaly detection from reactive classification to proactive early-warning, HyPV-LEAD establishes a robust foundation for real-time risk management, anti-money laundering (AML) compliance, and financial security in dynamic blockchain environments.
Authors: David Cortes, Carlos Juiz, Belen Bermejo
Abstract: Training large-scale deep learning models has become a key challenge for the scientific community and industry. While the massive use of GPUs can significantly speed up training times, this approach has a negative impact on efficiency. In this article, we present a detailed analysis of the times reported by MLPerf Training v4.1 on four workloads: BERT, Llama2 LoRA, RetinaNet, and Stable Diffusion, showing that there are configurations that optimise the relationship between performance, GPU usage, and efficiency. The results point to a break-even point that allows training times to be reduced while maximising efficiency.
Authors: Fatemeh Azad, Zoran Bosni\'c, Matja\v{z} Kukar
Abstract: Missing data represents a fundamental challenge in machine learning applications, often reducing model performance and reliability. This problem is particularly acute in fields like bioinformatics and clinical machine learning, where datasets are frequently incomplete due to the nature of both data generation and data collection. While numerous imputation methods exist, from simple statistical techniques to advanced deep learning models, no single method consistently performs well across diverse datasets and missingness mechanisms. This paper proposes a novel Meta-Imputation approach that learns to combine the outputs of multiple base imputers to predict missing values more accurately. By training the proposed method called Meta-Imputation Balanced (MIB) on synthetically masked data with known ground truth, the system learns to predict the most suitable imputed value based on the behavior of each method. Our work highlights the potential of ensemble learning in imputation and paves the way for more robust, modular, and interpretable preprocessing pipelines in real-world machine learning systems.
Authors: Leizhen Wang, Peibo Duan, Hao Wang, Yue Wang, Jian Xu, Nan Zheng, Zhenliang Ma
Abstract: In traffic engineering, the fixed-time traffic signal control remains widely used for its low cost, stability, and interpretability. However, its design depends on hand-crafted formulas (e.g., Webster) and manual re-timing by engineers to adapt to demand changes, which is labor-intensive and often yields suboptimal results under heterogeneous or congested conditions. This paper introduces the EvolveSignal, a large language models (LLMs) powered coding agent to automatically discover new traffic signal control algorithms. We formulate the problem as program synthesis, where candidate algorithms are represented as Python functions with fixed input-output structures, and iteratively optimized through external evaluations (e.g., a traffic simulator) and evolutionary search. Experiments on a signalized intersection demonstrate that the discovered algorithms outperform Webster's baseline, reducing average delay by 20.1% and average stops by 47.1%. Beyond performance, ablation and incremental analyses reveal that EvolveSignal modifications-such as adjusting cycle length bounds, incorporating right-turn demand, and rescaling green allocations-can offer practically meaningful insights for traffic engineers. This work opens a new research direction by leveraging AI for algorithm design in traffic signal control, bridging program synthesis with transportation engineering.
Authors: Fleur Hendriks, Ond\v{r}ej Roko\v{s}, Martin Do\v{s}k\'a\v{r}, Marc G. D. Geers, Vlado Menkovski
Abstract: Bifurcation phenomena in nonlinear dynamical systems often lead to multiple coexisting stable solutions, particularly in the presence of symmetry breaking. Deterministic machine learning models struggle to capture this multiplicity, averaging over solutions and failing to represent lower-symmetry outcomes. In this work, we propose a generative framework based on flow matching to model the full probability distribution over bifurcation outcomes. Our method enables direct sampling of multiple valid solutions while preserving system symmetries through equivariant modeling. We introduce a symmetric matching strategy that aligns predicted and target outputs under group actions, allowing accurate learning in equivariant settings. We validate our approach on a range of systems, from toy models to complex physical problems such as buckling beams and the Allen-Cahn equation. Our results demonstrate that flow matching significantly outperforms non-probabilistic and variational methods in capturing multimodal distributions and symmetry-breaking bifurcations, offering a principled and scalable solution for modeling multistability in high-dimensional systems.
Authors: Ilana Sebag, Jean-Yves Franceschi, Alain Rakotomamonjy, Alexandre Allauzen, Jamal Atif
Abstract: Generative Adversarial Networks (GANs) and diffusion models have emerged as leading approaches for high-quality image synthesis. While both can be trained under differential privacy (DP) to protect sensitive data, their sensitivity to membership inference attacks (MIAs), a key threat to data confidentiality, remains poorly understood. In this work, we present the first unified theoretical and empirical analysis of the privacy risks faced by differentially private generative models. We begin by showing, through a stability-based analysis, that GANs exhibit fundamentally lower sensitivity to data perturbations than diffusion models, suggesting a structural advantage in resisting MIAs. We then validate this insight with a comprehensive empirical study using a standardized MIA pipeline to evaluate privacy leakage across datasets and privacy budgets. Our results consistently reveal a marked privacy robustness gap in favor of GANs, even in strong DP regimes, highlighting that model type alone can critically shape privacy leakage.
Authors: Natalia Flechas Manrique, Alberto Mart\'inez, Elena L\'opez-Mart\'inez, Luc Andrea, Rom\'an Orus, Aitor Manteca, Aitziber L. Cortajarena, Lloren\c{c} Espinosa-Portal\'es
Abstract: Epitopes are short antigenic peptide sequences which are recognized by antibodies or immune cell receptors. These are central to the development of immunotherapies, vaccines, and diagnostics. However, the rational design of synthetic epitope libraries is challenging due to the large combinatorial sequence space, $20^n$ combinations for linear epitopes of n amino acids, making screening and testing unfeasible, even with high throughput experimental techniques. In this study, we present a large language model, epiGPTope, pre-trained on protein data and specifically fine-tuned on linear epitopes, which for the first time can directly generate novel epitope-like sequences, which are found to possess statistical properties analogous to the ones of known epitopes. This generative approach can be used to prepare libraries of epitope candidate sequences. We further train statistical classifiers to predict whether an epitope sequence is of bacterial or viral origin, thus narrowing the candidate library and increasing the likelihood of identifying specific epitopes. We propose that such combination of generative and predictive models can be of assistance in epitope discovery. The approach uses only primary amino acid sequences of linear epitopes, bypassing the need for a geometric framework or hand-crafted features of the sequences. By developing a method to create biologically feasible sequences, we anticipate faster and more cost-effective generation and screening of synthetic epitopes, with relevant applications in the development of new biotechnologies.
Authors: Oguzhan Baser, Kaan Kale, Po-han Li, Sandeep Chinchali
Abstract: Resource allocation is crucial for the performance optimization of cloud-assisted multi-agent intelligence. Traditional methods often overlook agents' diverse computational capabilities and complex operating environments, leading to inefficient and unfair resource distribution. To address this, we open-sourced Fair-Synergy, an algorithmic framework that utilizes the concave relationship between the agents' accuracy and the system resources to ensure fair resource allocation across fleet intelligence. We extend traditional allocation approaches to encompass a multidimensional machine learning utility landscape defined by model parameters, training data volume, and task complexity. We evaluate Fair-Synergy with advanced vision and language models such as BERT, VGG16, MobileNet, and ResNets on datasets including MNIST, CIFAR-10, CIFAR-100, BDD, and GLUE. We demonstrate that Fair-Synergy outperforms standard benchmarks by up to 25% in multi-agent inference and 11% in multi-agent learning settings. Also, we explore how the level of fairness affects the least advantaged, most advantaged, and average agents, providing insights for equitable fleet intelligence.
Authors: Bhanu Sharma, Eugene Pinsky
Abstract: In this study we analyzed average sleep durations across 61 countries to examine the impact of Daylight Saving Time (DST) practices. Key metrics influencing sleep were identified, and statistical correlation analysis was applied to explore relationships among these factors. Countries were grouped based on DST observance, and visualizations compared sleep patterns between DST and non-DST regions. Results show that, on average, countries observing DST tend to report longer sleep durations than those that do not. A more detailed pattern emerged when accounting for latitude: at lower latitudes, DST-observing countries reported shorter sleep durations compared to non-DST countries, while at higher latitudes, DST-observing countries reported longer average sleep durations. These findings suggest that the influence of DST on sleep may be moderated by geographical location.
Authors: Paul-Gauthier No\'e, Andreas Nautsch, Driss Matrouf, Pierre-Michel Bousquet, Jean-Fran\c{c}ois Bonastre
Abstract: While calibration of probabilistic predictions has been widely studied, this paper rather addresses calibration of likelihood functions. This has been discussed, especially in biometrics, in cases with only two exhaustive and mutually exclusive hypotheses (classes) where likelihood functions can be written as log-likelihood-ratios (LLRs). After defining calibration for LLRs and its connection with the concept of weight-of-evidence, we present the idempotence property and its associated constraint on the distribution of the LLRs. Although these results have been known for decades, they have been limited to the binary case. Here, we extend them to cases with more than two hypotheses by using the Aitchison geometry of the simplex, which allows us to recover, in a vector form, the additive form of the Bayes' rule; extending therefore the LLR and the weight-of-evidence to any number of hypotheses. Especially, we extend the definition of calibration, the idempotence, and the constraint on the distribution of likelihood functions to this multiple hypotheses and multiclass counterpart of the LLR: the isometric-log-ratio transformed likelihood function. This work is mainly conceptual, but we still provide one application to machine learning by presenting a non-linear discriminant analysis where the discriminant components form a calibrated likelihood function over the classes, improving therefore the interpretability and the reliability of the method.
Authors: Elizabeth Coda, Ery Arias-Castro, Gal Mishne
Abstract: Dimensionality reduction methods such as t-SNE and UMAP are popular methods for visualizing data with a potential (latent) clustered structure. They are known to group data points at the same time as they embed them, resulting in visualizations with well-separated clusters that preserve local information well. However, t-SNE and UMAP also tend to distort the global geometry of the underlying data. We propose a more transparent, modular approach consisting of first clustering the data, then embedding each cluster, and finally aligning the clusters to obtain a global embedding. We demonstrate this approach on several synthetic and real-world datasets and show that it is competitive with existing methods, while being much more transparent.
Authors: Taisiya Khakharova, Lucas Sakizloglou, Leen Lambers
Abstract: Sepsis is a serious, life-threatening condition. When treating sepsis, it is challenging to determine the correct amount of intravenous fluids and vasopressors for a given patient. While automated reinforcement learning (RL)-based methods have been used to support these decisions with promising results, previous studies have relied on relational data. Given the complexity of modern healthcare data, representing data as a graph may provide a more natural and effective approach. This study models patient data from the well-known MIMIC-III dataset as a heterogeneous graph that evolves over time. Subsequently, we explore two Graph Neural Network architectures - GraphSAGE and GATv2 - for learning patient state representations, adopting the approach of decoupling representation learning from policy learning. The encoders are trained to produce latent state representations, jointly with decoders that predict the next patient state. These representations are then used for policy learning with the dBCQ algorithm. The results of our experimental evaluation confirm the potential of a graph-based approach, while highlighting the complexity of representation learning in this domain.
Authors: Chenlu Ye, Zhou Yu, Ziji Zhang, Hao Chen, Narayanan Sadagopan, Jing Huang, Tong Zhang, Anurag Beniwal
Abstract: Reinforcement learning with verifiable rewards (RLVR) has emerged to be a predominant paradigm for mathematical reasoning tasks, offering stable improvements in reasoning ability. However, Outcome Reward Models (ORMs) in RLVR are too coarse-grained to distinguish flawed reasoning within correct answers or valid reasoning within incorrect answers. This lack of granularity introduces noisy and misleading gradients significantly and hinders further progress in reasoning process quality. While Process Reward Models (PRMs) offer fine-grained guidance for intermediate steps, they frequently suffer from inaccuracies and are susceptible to reward hacking. To resolve this dilemma, we introduce PRocess cOnsistency Filter (PROF), an effective data process curation method that harmonizes noisy, fine-grained process rewards with accurate, coarse-grained outcome rewards. Rather than naively blending PRM and ORM in the objective function (arXiv:archive/2506.18896), PROF leverages their complementary strengths through consistency-driven sample selection. Our approach retains correct responses with higher averaged process values and incorrect responses with lower averaged process values, while maintaining positive/negative training sample balance. Extensive experiments demonstrate that our method not only consistently improves the final accuracy over $4\%$ compared to the blending approaches, but also strengthens the quality of intermediate reasoning steps. Codes and training recipes are available at https://github.com/Chenluye99/PROF.
Authors: Spyros Rigas, Dhruv Verma, Georgios Alexandridis, Yixuan Wang
Abstract: Kolmogorov-Arnold Networks (KANs) are a recently introduced neural architecture that replace fixed nonlinearities with trainable activation functions, offering enhanced flexibility and interpretability. While KANs have been applied successfully across scientific and machine learning tasks, their initialization strategies remain largely unexplored. In this work, we study initialization schemes for spline-based KANs, proposing two theory-driven approaches inspired by LeCun and Glorot, as well as an empirical power-law family with tunable exponents. Our evaluation combines large-scale grid searches on function fitting and forward PDE benchmarks, an analysis of training dynamics through the lens of the Neural Tangent Kernel, and evaluations on a subset of the Feynman dataset. Our findings indicate that the Glorot-inspired initialization significantly outperforms the baseline in parameter-rich models, while power-law initialization achieves the strongest performance overall, both across tasks and for architectures of varying size. All code and data accompanying this manuscript are publicly available at https://github.com/srigas/KAN_Initialization_Schemes.
Authors: Phuc Pham, Viet Thanh Duy Nguyen, Truong-Son Hy
Abstract: Accurate identification of interactions between protein residues and ligand functional groups is essential to understand molecular recognition and guide rational drug design. Existing deep learning approaches for protein-ligand interpretability often rely on 3D structural input or use distance-based contact labels, limiting both their applicability and biological relevance. We introduce LINKER, the first sequence-based model to predict residue-functional group interactions in terms of biologically defined interaction types, using only protein sequences and the ligand SMILES as input. LINKER is trained with structure-supervised attention, where interaction labels are derived from 3D protein-ligand complexes via functional group-based motif extraction. By abstracting ligand structures into functional groups, the model focuses on chemically meaningful substructures while predicting interaction types rather than mere spatial proximity. Crucially, LINKER requires only sequence-level input at inference time, enabling large-scale application in settings where structural data is unavailable. Experiments on the LP-PDBBind benchmark demonstrate that structure-informed supervision over functional group abstractions yields interaction predictions closely aligned with ground-truth biochemical annotations.
Authors: Niteesh Midlagajni, Constantin A. Rothkopf
Abstract: Simulating particle dynamics with high fidelity is crucial for solving real-world interaction and control tasks involving liquids in design, graphics, and robotics. Recently, data-driven approaches, particularly those based on graph neural networks (GNNs), have shown progress in tackling such problems. However, these approaches are often limited to learning fluid behavior in static free-fall environments or simple manipulation settings involving primitive objects, often overlooking complex interactions with dynamically moving kinematic rigid bodies. Here, we propose a GNN-based framework designed from the ground up to learn the dynamics of liquids under rigid body interactions and active manipulations, where particles are represented as graph nodes and particle-object collisions are handled using surface representations with the bounding volume hierarchy (BVH) algorithm. This approach enables the network to model complex interactions between liquid particles and intricate surface geometries. Our model accurately captures fluid behavior in dynamic settings and can also function as a simulator in static free-fall environments. Despite being trained on a single-object manipulation task of pouring, our model generalizes effectively to environments with unseen objects and novel manipulation tasks such as stirring and scooping. Finally, we show that the learned dynamics can be leveraged to solve control and manipulation tasks using gradient-based optimization methods.
Authors: Yubo Gao, Renbo Tu, Gennady Pekhimenko, Nandita Vijaykumar
Abstract: Differentially-Private SGD (DP-SGD) is a powerful technique to protect user privacy when using sensitive data to train neural networks. During training, converting model weights and activations into low-precision formats, i.e., quantization, can drastically reduce training times, energy consumption, and cost, and is thus a widely used technique. In this work, we demonstrate that quantization causes significantly higher accuracy degradation in DP-SGD compared to regular SGD. We observe that this is caused by noise injection in DP-SGD, which amplifies quantization variance, leading to disproportionately large accuracy degradation. To address this challenge, we present QPQuant, a dynamic quantization framework that adaptively selects a changing subset of layers to quantize at each epoch. Our method combines two key ideas that effectively reduce quantization variance: (i) probabilistic sampling of the layers that rotates which layers are quantized every epoch, and (ii) loss-aware layer prioritization, which uses a differentially private loss sensitivity estimator to identify layers that can be quantized with minimal impact on model quality. This estimator consumes a negligible fraction of the overall privacy budget, preserving DP guarantees. Empirical evaluations on ResNet18, ResNet50, and DenseNet121 across a range of datasets demonstrate that DPQuant consistently outperforms static quantization baselines, achieving near Pareto-optimal accuracy-compute trade-offs and up to 2.21x theoretical throughput improvements on low-precision hardware, with less than 2% drop in validation accuracy.
Authors: Erkan Bayram, Mohamed-Ali Belabbas, Tamer Ba\c{s}ar
Abstract: In our earlier work, we introduced the principle of Tuning without Forgetting (TwF) for sequential training of neural ODEs, where training samples are added iteratively and parameters are updated within the subspace of control functions that preserves the end-point mapping at previously learned samples on the manifold of output labels in the first-order approximation sense. In this letter, we prove that this parameter subspace forms a Banach submanifold of finite codimension under nonsingular controls, and we characterize its tangent space. This reveals that TwF corresponds to a continuation/deformation of the control function along the tangent space of this Banach submanifold, providing a theoretical foundation for its mapping-preserving (not forgetting) during the sequential training exactly, beyond first-order approximation.
Authors: Duy A. Nguyen, Abhi Kamboj, Minh N. Do
Abstract: Addressing missing modalities and limited labeled data is crucial for advancing robust multimodal learning. We propose Robult, a scalable framework designed to mitigate these challenges by preserving modality-specific information and leveraging redundancy through a novel information-theoretic approach. Robult optimizes two core objectives: (1) a soft Positive-Unlabeled (PU) contrastive loss that maximizes task-relevant feature alignment while effectively utilizing limited labeled data in semi-supervised settings, and (2) a latent reconstruction loss that ensures unique modality-specific information is retained. These strategies, embedded within a modular design, enhance performance across various downstream tasks and ensure resilience to incomplete modalities during inference. Experimental results across diverse datasets validate that Robult achieves superior performance over existing approaches in both semi-supervised learning and missing modality contexts. Furthermore, its lightweight design promotes scalability and seamless integration with existing architectures, making it suitable for real-world multimodal applications.
Authors: Jigang Fan, Zhenghong Zhou, Ruofan Jin, Le Cong, Mengdi Wang, Zaixi Zhang
Abstract: Proteins play crucial roles in almost all biological processes. The advancement of deep learning has greatly accelerated the development of protein foundation models, leading to significant successes in protein understanding and design. However, the lack of systematic red-teaming for these models has raised serious concerns about their potential misuse, such as generating proteins with biological safety risks. This paper introduces SafeProtein, the first red-teaming framework designed for protein foundation models to the best of our knowledge. SafeProtein combines multimodal prompt engineering and heuristic beam search to systematically design red-teaming methods and conduct tests on protein foundation models. We also curated SafeProtein-Bench, which includes a manually constructed red-teaming benchmark dataset and a comprehensive evaluation protocol. SafeProtein achieved continuous jailbreaks on state-of-the-art protein foundation models (up to 70% attack success rate for ESM3), revealing potential biological safety risks in current protein foundation models and providing insights for the development of robust security protection technologies for frontier models. The codes will be made publicly available at https://github.com/jigang-fan/SafeProtein.
Authors: Han Shen
Abstract: For RL algorithms, appropriate entropy control is crucial to their effectiveness. To control the policy entropy, a commonly used method is entropy regularization, which is adopted in various popular RL algorithms including PPO, SAC and A3C. Although entropy regularization proves effective in robotic and games RL conventionally, studies found that it gives weak to no gains in LLM-RL training. In this work, we study the issues of entropy bonus in LLM-RL setting. Specifically, we first argue that the conventional entropy regularization suffers from the LLM's extremely large response space and the sparsity of the optimal outputs. As a remedy, we propose AEnt, an entropy control method that utilizes a new clamped entropy bonus with an automatically adjusted coefficient. The clamped entropy is evaluated with the re-normalized policy defined on certain smaller token space, which encourages exploration within a more compact response set. In addition, the algorithm automatically adjusts entropy coefficient according to the clamped entropy value, effectively controlling the entropy-induced bias while leveraging the entropy's benefits. AEnt is tested in math-reasoning tasks under different base models and datasets, and it is observed that AEnt outperforms the baselines consistently across multiple benchmarks.
Authors: Xin-Yi Tong, Sherrie Wang
Abstract: Accurately obtaining crop type and its spatial distribution at a global scale is critical for food security, agricultural policy-making, and sustainable development. Remote sensing offers an efficient solution for large-scale crop classification, but the limited availability of reliable ground samples in many regions constrains applicability across geographic areas. To address performance declines under geospatial shifts, this study identifies remote sensing features that are invariant to geographic variation and proposes strategies to enhance cross-regional generalization. We construct CropGlobe, a global crop type dataset with 300,000 pixel-level samples from eight countries across five continents, covering six major food and industrial crops (corn, soybeans, rice, wheat, sugarcane, cotton). With broad geographic coverage, CropGlobe enables a systematic evaluation under cross-country, cross-continent, and cross-hemisphere transfer. We compare the transferability of temporal multi-spectral features (Sentinel-2-based 1D/2D median features and harmonic coefficients) and hyperspectral features (from EMIT). To improve generalization under spectral and phenological shifts, we design CropNet, a lightweight and robust CNN tailored for pixel-level crop classification, coupled with temporal data augmentation (time shift, time scale, and magnitude warping) that simulates realistic cross-regional phenology. Experiments show that 2D median temporal features from Sentinel-2 consistently exhibit the strongest invariance across all transfer scenarios, and augmentation further improves robustness, particularly when training data diversity is limited. Overall, the work identifies more invariant feature representations that enhance geographic transferability and suggests a promising path toward scalable, low-cost crop type applications across globally diverse regions.
Authors: Gwen Legate, Irina Rish, Eugene Belilovsky
Abstract: Federated learning enables collaborative model training across numerous edge devices without requiring participants to share data; however, memory and communication constraints on these edge devices may preclude their participation in training. We consider a setting in which a subset of edge devices are below a critical memory or communication threshold required to conduct model updates. Under typical federated optimization algorithms, these devices are excluded from training which renders their data inaccessible and increases system induced bias. We are inspired by MeZO, a zeroth-order method used for memory-efficient fine-tuning. The increased variance inherent to zeroth-order gradient approximations has relegated previous zeroth-order optimizers exclusively to the domain of fine tuning; a limitation we seek to correct. We devise a federated, memory-efficient zeroth-order optimizer, ZOWarmUp that permits zeroth-order training from a random initialization. ZOWarmUp leverages differing client capabilities and careful variance reduction techniques to facilitate participation of under-represented, low-resource clients in model training. Like other federated zeroth-order methods, ZOWarmUp eliminates the need for edge devices to transmit their full gradients to the server and instead relies on only a small set of random seeds, rendering the up-link communication cost negligible. We present experiments using various datasets and model architectures to show that ZOWarmUp is a robust algorithm that can can be applied under a wide variety of circumstances. For systems with a high proportion of edge devices that would otherwise be excluded from training, this algorithm provides access to a greater volume and diversity of data, thus improving training outcomes.
Authors: Xingxuan Zhang, Gang Ren, Han Yu, Hao Yuan, Hui Wang, Jiansheng Li, Jiayun Wu, Lang Mo, Li Mao, Mingchao Hao, Ningbo Dai, Renzhe Xu, Shuyang Li, Tianyang Zhang, Yue He, Yuanrui Wang, Yunjia Zhang, Zijing Xu, Dongzhe Li, Fang Gao, Hao Zou, Jiandong Liu, Jiashuo Liu, Jiawei Xu, Kaijie Cheng, Kehan Li, Linjun Zhou, Qing Li, Shaohua Fan, Xiaoyu Lin, Xinyan Han, Xuanyue Li, Yan Lu, Yuan Xue, Yuanyuan Jiang, Zimu Wang, Zhenlei Wang, Peng Cui
Abstract: We argue that progress toward general intelligence requires complementary foundation models grounded in language, the physical world, and structured data. This report presents LimiX, the first installment of our large structured-data models (LDMs). LimiX treats structured data as a joint distribution over variables and missingness, thus capable of addressing a wide range of tabular tasks through query-based conditional prediction via a single model. LimiX is pretrained using masked joint-distribution modeling with an episodic, context-conditional objective, where the model predicts for query subsets conditioned on dataset-specific contexts, supporting rapid, training-free adaptation at inference. We evaluate LimiX across 10 large structured-data benchmarks with broad regimes of sample size, feature dimensionality, class number, categorical-to-numerical feature ratio, missingness, and sample-to-feature ratios. With a single model and a unified interface, LimiX consistently surpasses strong baselines including gradient-boosting trees, deep tabular networks, recent tabular foundation models, and automated ensembles, as shown in Figure 1 and Figure 2. The superiority holds across a wide range of tasks, such as classification, regression, missing value imputation, and data generation, often by substantial margins, while avoiding task-specific architectures or bespoke training per task. All LimiX models are publicly accessible under Apache 2.0.
Authors: Haoran Huan, Mihir Prabhudesai, Mengning Wu, Shantanu Jaiswal, Deepak Pathak
Abstract: Large language models (LLMs) have demonstrated impressive capabilities across a variety of tasks, but their increasing autonomy in real-world applications raises concerns about their trustworthiness. While hallucinations-unintentional falsehoods-have been widely studied, the phenomenon of lying, where an LLM knowingly generates falsehoods to achieve an ulterior objective, remains underexplored. In this work, we systematically investigate the lying behavior of LLMs, differentiating it from hallucinations and testing it in practical scenarios. Through mechanistic interpretability techniques, we uncover the neural mechanisms underlying deception, employing logit lens analysis, causal interventions, and contrastive activation steering to identify and control deceptive behavior. We study real-world lying scenarios and introduce behavioral steering vectors that enable fine-grained manipulation of lying tendencies. Further, we explore the trade-offs between lying and end-task performance, establishing a Pareto frontier where dishonesty can enhance goal optimization. Our findings contribute to the broader discourse on AI ethics, shedding light on the risks and potential safeguards for deploying LLMs in high-stakes environments. Code and more illustrations are available at https://llm-liar.github.io/
Authors: Mohammad Mehedi Hasan, Pedro G. Lind, Hernando Ombao, Anis Yazidi, Rabindra Khadka
Abstract: Dementia (DEM) is a growing global health challenge, underscoring the need for early and accurate diagnosis. Electroencephalography (EEG) provides a non-invasive window into brain activity, but conventional methods struggle to capture its transient complexity. We present the \textbf{EEG Microstate Analysis Framework (EEG-MSAF)}, an end-to-end pipeline that leverages EEG microstates discrete, quasi-stable topographies to identify DEM-related biomarkers and distinguish DEM, mild cognitive impairment (MCI), and normal cognition (NC). EEG-MSAF comprises three stages: (1) automated microstate feature extraction, (2) classification with machine learning (ML), and (3) feature ranking using Shapley Additive Explanations (SHAP) to highlight key biomarkers. We evaluate on two EEG datasets: the public Chung-Ang University EEG (CAUEEG) dataset and a clinical cohort from Thessaloniki Hospital. Our framework demonstrates strong performance and generalizability. On CAUEEG, EEG-MSAF-SVM achieves \textbf{89\% $\pm$ 0.01 accuracy}, surpassing the deep learning baseline CEEDNET by \textbf{19.3\%}. On the Thessaloniki dataset, it reaches \textbf{95\% $\pm$ 0.01 accuracy}, comparable to EEGConvNeXt. SHAP analysis identifies mean correlation and occurrence as the most informative metrics: disruption of microstate C (salience/attention network) dominates DEM prediction, while microstate F, a novel default-mode pattern, emerges as a key early biomarker for both MCI and DEM. By combining accuracy, generalizability, and interpretability, EEG-MSAF advances EEG-based dementia diagnosis and sheds light on brain dynamics across the cognitive spectrum.
Authors: Diego Di Carlo (RIKEN AIP), Koyama Shoichi (UTokyo), Nugraha Aditya Arie (RIKEN AIP), Fontaine Mathieu (LTCI, S2A), Bando Yoshiaki (AIST), Yoshii Kazuyoshi (RIKEN AIP)
Abstract: This paper investigates continuous representations of steering vectors over frequency and position of microphone and source for augmented listening (e.g., spatial filtering and binaural rendering) with precise control of the sound field perceived by the user. Steering vectors have typically been used for representing the spatial characteristics of the sound field as a function of the listening position. The basic algebraic representation of steering vectors assuming an idealized environment cannot deal with the scattering effect of the sound field. One may thus collect a discrete set of real steering vectors measured in dedicated facilities and super-resolve (i.e., upsample) them. Recently, physics-aware deep learning methods have been effectively used for this purpose. Such deterministic super-resolution, however, suffers from the overfitting problem due to the non-uniform uncertainty over the measurement space. To solve this problem, we integrate an expressive representation based on the neural field (NF) into the principled probabilistic framework based on the Gaussian process (GP). Specifically, we propose a physics-aware composite kernel that model the directional incoming waves and the subsequent scattering effect. Our comprehensive comparative experiment showed the effectiveness of the proposed method under data insufficiency conditions. In downstream tasks such as speech enhancement and binaural rendering using the simulated data of the SPEAR challenge, the oracle performances were attained with less than ten times fewer measurements.
Authors: Thanh Tung Khuat, Johnny Peng, Robert Bassett, Ellen Otte, Bogdan Gabrys
Abstract: This study explores the deployment of three machine learning (ML) approaches for real-time prediction of glucose, lactate, and ammonium concentrations in cell culture processes, using Raman spectroscopy as input features. The research addresses challenges associated with limited data availability and process variability, providing a comparative analysis of pretrained models, just-in-time learning (JITL), and online learning algorithms. Two industrial case studies are presented to evaluate the impact of varying bioprocess conditions on model performance. The findings highlight the specific conditions under which pretrained models demonstrate superior predictive accuracy and identify scenarios where JITL or online learning approaches are more effective for adaptive process monitoring. This study also highlights the critical importance of updating the deployed models/agents with the latest offline analytical measurements during bioreactor operations to maintain the model performance against the changes in cell growth behaviours and operating conditions throughout the bioreactor run. Additionally, the study confirms the usefulness of a simple mixture-of-experts framework in achieving enhanced accuracy and robustness for real-time predictions of metabolite concentrations based on Raman spectral data. These insights contribute to the development of robust strategies for the efficient deployment of ML models in dynamic and changing biomanufacturing environments.
Authors: Xinbo Zhang, Montserrat Guillen, Lishuai Li, Xin Li, Youhua Frank Chen
Abstract: Driving behavior big data leverages multi-sensor telematics to understand how people drive and powers applications such as risk evaluation, insurance pricing, and targeted intervention. Usage-based insurance (UBI) built on these data has become mainstream. Telematics-captured near-miss events (NMEs) provide a timely alternative to claim-based risk, but weekly NMEs are sparse, highly zero-inflated, and behaviorally heterogeneous even after exposure normalization. Analyzing multi-sensor telematics and ADAS warnings, we show that the traditional statistical models underfit the dataset. We address these challenges by proposing a set of zero-inflated Poisson (ZIP) frameworks that learn latent behavior groups and fit offset-based count models via EM to yield calibrated, interpretable weekly risk predictions. Using a naturalistic dataset from a fleet of 354 commercial drivers over a year, during which the drivers completed 287,511 trips and logged 8,142,896 km in total, our results show consistent improvements over baselines and prior telematics models, with lower AIC/BIC values in-sample and better calibration out-of-sample. We also conducted sensitivity analyses on the EM-based grouping for the number of clusters, finding that the gains were robust and interpretable. Practically, this supports context-aware ratemaking on a weekly basis and fairer premiums by recognizing heterogeneous driving styles.
Authors: Pucheng Tang, Hongqiao Wang, Wenzhou Lin, Qian Chen, Heng Yong
Abstract: Parametric partial differential equations (PDEs) are fundamental mathematical tools for modeling complex physical systems, yet their numerical evaluation across parameter spaces remains computationally intensive when using conventional high-fidelity solvers. To address this challenge, we propose a novel physical law-corrected prior Gaussian process (LC-prior GP) surrogate modeling framework that effectively integrates data-driven learning with underlying physical constraints to flexibly handle multi-coupled variables defined on complex geometries. The proposed approach leverages proper orthogonal decomposition (POD) to parameterize high-dimensional PDE solutions via their dominant modes and associated coefficients, thereby enabling efficient Gaussian process (GP) surrogate modeling within a reduced-dimensional coefficient space. A key contribution lies in the incorporation of physical laws together with a limited number of parameter samples to correct the GP posterior mean, thus avoiding reliance on computationally expensive numerical solvers. Furthermore, interpolation functions are constructed to describe the mapping from the full parameter space to the physics-based correction term. This mapping is subsequently backpropagated to constrain the original GP surrogate, yielding a more physically consistent conditional prior. To handle irregular geometries, the radial basis function-finite difference (RBF-FD) method is incorporated during training set computation, with its inherent differentiation matrices providing both computational efficiency and numerical accuracy for physical constraint optimization. The effectiveness of the proposed method is demonstrated through numerical experiments involving a reaction-diffusion model, miscible flooding models, and Navier-Stokes equations with multi-physics coupling defined on irregular domains.
Authors: Guangzheng Zhong, Yang Liu, Jiming Liu
Abstract: In decision-dependent games, multiple players optimize their decisions under a data distribution that shifts with their joint actions, creating complex dynamics in applications like market pricing. A practical consequence of these dynamics is the \textit{performatively stable equilibrium}, where each player's strategy is a best response under the induced distribution. Prior work relies on $\beta$-smoothness, assuming Lipschitz continuity of loss function gradients with respect to the data distribution, which is impractical as the data distribution maps, i.e., the relationship between joint decision and the resulting distribution shifts, are typically unknown, rendering $\beta$ unobtainable. To overcome this limitation, we propose a gradient-based sensitivity measure that directly quantifies the impact of decision-induced distribution shifts. Leveraging this measure, we derive convergence guarantees for performatively stable equilibria under a practically feasible assumption of strong monotonicity. Accordingly, we develop a sensitivity-informed repeated retraining algorithm that adjusts players' loss functions based on the sensitivity measure, guaranteeing convergence to performatively stable equilibria for arbitrary data distribution maps. Experiments on prediction error minimization game, Cournot competition, and revenue maximization game show that our approach outperforms state-of-the-art baselines, achieving lower losses and faster convergence.
Authors: John Zobolas, Anne-Marie George, Alberto L\'opez, Sebastian Fischer, Marc Becker, Tero Aittokallio
Abstract: Prediction of patient survival using high-dimensional multi-omics data requires systematic feature selection methods that ensure predictive performance, sparsity, and reliability for prognostic biomarker discovery. We developed a hybrid ensemble feature selection (hEFS) approach that combines data subsampling with multiple prognostic models, integrating both embedded and wrapper-based strategies for survival prediction. Omics features are ranked using a voting-theory-inspired aggregation mechanism across models and subsamples, while the optimal number of features is selected via a Pareto front, balancing predictive accuracy and model sparsity without any user-defined thresholds. When applied to multi-omics datasets from three pancreatic cancer cohorts, hEFS identifies significantly fewer and more stable biomarkers compared to the conventional, late-fusion CoxLasso models, while maintaining comparable discrimination performance. Implemented within the open-source mlr3fselect R package, hEFS offers a robust, interpretable, and clinically valuable tool for prognostic modelling and biomarker discovery in high-dimensional survival settings.
Authors: Nathan Doum\`eche (LPSM, EDF R&D OSIRIS), Francis Bach (ENS-PSL), G\'erard Biau (LPSM, IUF), Claire Boyer (LMO)
Abstract: Kernel methods are powerful tools in statistical learning, but their cubic complexity in the sample size n limits their use on large-scale datasets. In this work, we introduce a scalable framework for kernel regression with O(n log n) complexity, fully leveraging GPU acceleration. The approach is based on a Fourier representation of kernels combined with non-uniform fast Fourier transforms (NUFFT), enabling exact, fast, and memory-efficient computations. We instantiate our framework in three settings: Sobolev kernel regression, physics-informed regression, and additive models. When known, the proposed estimators are shown to achieve minimax convergence rates, consistent with classical kernel theory. Empirical results demonstrate that our methods can process up to tens of billions of samples within minutes, providing both statistical accuracy and computational scalability. These contributions establish a flexible approach, paving the way for the routine application of kernel methods in large-scale learning tasks.
Authors: Alissa A. Valentine, Lauren A. Lepow, Lili Chan, Alexander W. Charney, Isotta Landi
Abstract: In the United States, schizophrenia (SCZ) carries a race and sex disparity that may be explained by clinician bias - a belief held by a clinician about a patient that prevents impartial clinical decision making. The emergency department (ED) is marked by higher rates of stress that lead to clinicians relying more on implicit biases during decision making. In this work, we considered a large cohort of psychiatric patients in the ED from the Mount Sinai Health System (MSHS) in New York City to investigate the effects of clinician bias on SCZ diagnosis while controlling for known risk factors and patient sociodemographic information. Clinician bias was quantified as the ratio of negative to total sentences within a patient's first ED note. We utilized a logistic regression to predict SCZ diagnosis given patient race, sex, age, history of trauma or substance use disorder, and the ratio of negative sentences. Our findings showed that an increased ratio of negative sentences is associated with higher odds of obtaining a SCZ diagnosis [OR (95% CI)=1.408 (1.361-1.456)]. Identifying as male [OR (95% CI)=1.112 (1.055-1.173)] or Black [OR (95% CI)=1.081(1.031-1.133)] increased one's odds of being diagnosed with SCZ. However, from an intersectional lens, Black female patients with high SES have the highest odds of obtaining a SCZ diagnosis [OR (95% CI)=1.629 (1.535-1.729)]. Results such as these suggest that SES does not act as a protective buffer against SCZ diagnosis in all patients, demanding more attention to the quantification of health disparities. Lastly, we demonstrated that clinician bias is operational with real world data and related to increased odds of obtaining a stigmatizing diagnosis such as SCZ.
Authors: Xiangpeng Li, Junwei Ma, Bo Li, Ali Mostafavi
Abstract: The multifaceted nature of disaster impact shows that densely populated areas contribute more to aggregate burden, while sparsely populated but heavily affected regions suffer disproportionately at the individual level. This study introduces a framework for quantifying the societal impacts of power outages by translating customer weighted outage exposure into deprivation measures, integrating welfare metrics with three recovery indicators, average outage days per customer, restoration duration, and relative restoration rate, computed from sequential EAGLE I observations and linked to Zip Code Tabulation Area demographics. Applied to four United States hurricanes, Beryl 2024 Texas, Helene 2024 Florida, Milton 2024 Florida, and Ida 2021 Louisiana, this standardized pipeline provides the first cross event, fine scale evaluation of outage impacts and their drivers. Results demonstrate regressive patterns with greater burdens in lower income areas, mechanistic analysis shows deprivation increases with longer restoration durations and decreases with faster restoration rates, explainable modeling identifies restoration duration as the dominant driver, and clustering reveals distinct recovery typologies not captured by conventional reliability metrics. This framework delivers a transferable method for assessing outage impacts and equity, comparative cross event evidence linking restoration dynamics to social outcomes, and actionable spatial analyses that support equity informed restoration planning and resilience investment.
Authors: Andrew Ferguson, Marisa LaFleur, Lars Ruthotto, Jesse Thaler, Yuan-Sen Ting, Pratyush Tiwary, Soledad Villar, E. Paulo Alves, Jeremy Avigad, Simon Billinge, Camille Bilodeau, Keith Brown, Emmanuel Candes, Arghya Chattopadhyay, Bingqing Cheng, Jonathan Clausen, Connor Coley, Andrew Connolly, Fred Daum, Sijia Dong, Chrisy Xiyu Du, Cora Dvorkin, Cristiano Fanelli, Eric B. Ford, Luis Manuel Frutos, Nicol\'as Garc\'ia Trillos, Cecilia Garraffo, Robert Ghrist, Rafael Gomez-Bombarelli, Gianluca Guadagni, Sreelekha Guggilam, Sergei Gukov, Juan B. Guti\'errez, Salman Habib, Johannes Hachmann, Boris Hanin, Philip Harris, Murray Holland, Elizabeth Holm, Hsin-Yuan Huang, Shih-Chieh Hsu, Nick Jackson, Olexandr Isayev, Heng Ji, Aggelos Katsaggelos, Jeremy Kepner, Yannis Kevrekidis, Michelle Kuchera, J. Nathan Kutz, Branislava Lalic, Ann Lee, Matt LeBlanc, Josiah Lim, Rebecca Lindsey, Yongmin Liu, Peter Y. Lu, Sudhir Malik, Vuk Mandic, Vidya Manian, Emeka P. Mazi, Pankaj Mehta, Peter Melchior, Brice M\'enard, Jennifer Ngadiuba, Stella Offner, Elsa Olivetti, Shyue Ping Ong, Christopher Rackauckas, Philippe Rigollet, Chad Risko, Philip Romero, Grant Rotskoff, Brett Savoie, Uros Seljak, David Shih, Gary Shiu, Dima Shlyakhtenko, Eva Silverstein, Taylor Sparks, Thomas Strohmer, Christopher Stubbs, Stephen Thomas, Suriyanarayanan Vaikuntanathan, Rene Vidal, Francisco Villaescusa-Navarro, Gregory Voth, Benjamin Wandelt, Rachel Ward, Melanie Weber, Risa Wechsler, Stephen Whitelam, Olaf Wiest, Mike Williams, Zhuoran Yang, Yaroslava G. Yingling, Bin Yu, Shuwen Yue, Ann Zabludoff, Huimin Zhao, Tong Zhang
Abstract: This community paper developed out of the NSF Workshop on the Future of Artificial Intelligence (AI) and the Mathematical and Physics Sciences (MPS), which was held in March 2025 with the goal of understanding how the MPS domains (Astronomy, Chemistry, Materials Research, Mathematical Sciences, and Physics) can best capitalize on, and contribute to, the future of AI. We present here a summary and snapshot of the MPS community's perspective, as of Spring/Summer 2025, in a rapidly developing field. The link between AI and MPS is becoming increasingly inextricable; now is a crucial moment to strengthen the link between AI and Science by pursuing a strategy that proactively and thoughtfully leverages the potential of AI for scientific discovery and optimizes opportunities to impact the development of AI by applying concepts from fundamental science. To achieve this, we propose activities and strategic priorities that: (1) enable AI+MPS research in both directions; (2) build up an interdisciplinary community of AI+MPS researchers; and (3) foster education and workforce development in AI for MPS researchers and students. We conclude with a summary of suggested priorities for funding agencies, educational institutions, and individual researchers to help position the MPS community to be a leader in, and take full advantage of, the transformative potential of AI+MPS.
Authors: Gabriel A. B. do Nascimento, Vincent Dong, Guilherme J. Cavalcante, Alex Nguyen, Tha\'is G. do R\^ego, Yuri Malheiros, Telmo M. Silva Filho, Carla R. Zeballos Torrez, James C. Gee, Anne Marie McCarthy, Andrew D. A. Maidment, Bruno Barufaldi
Abstract: Accurate breast MRI lesion detection is critical for early cancer diagnosis, especially in high-risk populations. We present a classification pipeline that adapts a pretrained foundation model, the Medical Slice Transformer (MST), for breast lesion classification using dynamic contrast-enhanced MRI (DCE-MRI). Leveraging DINOv2-based self-supervised pretraining, MST generates robust per-slice feature embeddings, which are then used to train a Kolmogorov--Arnold Network (KAN) classifier. The KAN provides a flexible and interpretable alternative to conventional convolutional networks by enabling localized nonlinear transformations via adaptive B-spline activations. This enhances the model's ability to differentiate benign from malignant lesions in imbalanced and heterogeneous clinical datasets. Experimental results demonstrate that the MST+KAN pipeline outperforms the baseline MST classifier, achieving AUC = 0.80 \pm 0.02 while preserving interpretability through attention-based heatmaps. Our findings highlight the effectiveness of combining foundation model embeddings with advanced classification strategies for building robust and generalizable breast MRI analysis tools.
Authors: Matthew Russo, Tim Kraska
Abstract: With advances in large language models (LLMs), researchers are creating new systems that can perform AI-driven analytics over large unstructured datasets. Recent work has explored executing such analytics queries using semantic operators -- a declarative set of AI-powered data transformations with natural language specifications. However, even when optimized, these operators can be expensive to execute on millions of records and their iterator execution semantics make them ill-suited for interactive data analytics tasks. In another line of work, Deep Research systems have demonstrated an ability to answer natural language question(s) over large datasets. These systems use one or more LLM agent(s) to plan their execution, process the dataset(s), and iteratively refine their answer. However, these systems do not explicitly optimize their query plans which can lead to poor plan execution. In order for AI-driven analytics to excel, we need a runtime which combines the optimized execution of semantic operators with the flexibility and more dynamic execution of Deep Research systems. As a first step towards this vision, we build a prototype which enables Deep Research agents to write and execute optimized semantic operator programs. We evaluate our prototype and demonstrate that it can outperform a handcrafted semantic operator program and open Deep Research systems on two basic queries. Compared to a standard open Deep Research agent, our prototype achieves up to 1.95x better F1-score. Furthermore, even if we give the agent access to semantic operators as tools, our prototype still achieves cost and runtime savings of up to 76.8% and 72.7% thanks to its optimized execution.
Authors: Nico Bohlinger, Jan Peters
Abstract: We present a single, general locomotion policy trained on a diverse collection of 50 legged robots. By combining an improved embodiment-aware architecture (URMAv2) with a performance-based curriculum for extreme Embodiment Randomization, our policy learns to control millions of morphological variations. Our policy achieves zero-shot transfer to unseen real-world humanoid and quadruped robots.
Authors: Johannes J. Brust, Michael A. Saunders
Abstract: For a datastream, the change over a short interval is often of low rank. For high throughput information arranged in matrix format, recomputing an optimal SVD approximation after each step is typically prohibitive. Instead, incremental and truncated updating strategies are used, which may not scale for large truncation ranks. Therefore, we propose a set of efficient new algorithms that update a bidiagonal factorization, and which are similarly accurate as the SVD methods. In particular, we develop a compact Householder-type algorithm that decouples a sparse part from a low-rank update and has about half the memory requirements of standard bidiagonalization methods. A second algorithm based on Givens rotations has only about 10 flops per rotation and scales quadratically with the problem size, compared to a typical cubic scaling. The algorithm is therefore effective for processing high-throughput updates, as we demonstrate in tracking large subspaces of recommendation systems and networks, and when compared to well known software such as LAPACK or the incremental SVD.
Authors: Sadra Saremi, Amirhossein Ahmadkhan Kordbacheh
Abstract: Colon cancer also known as Colorectal cancer, is one of the most malignant types of cancer worldwide. Early-stage detection of colon cancer is highly crucial to prevent its deterioration. This research presents a hybrid multi-scale deep learning architecture that synergizes capsule networks, graph attention mechanisms, transformer modules, and residual learning to advance colon cancer classification on the Lung and Colon Cancer Histopathological Image Dataset (LC25000) dataset. The proposed model in this paper utilizes the HG-TNet model that introduces a hybrid architecture that joins strength points in transformers and convolutional neural networks to capture multi-scale features in histopathological images. Mainly, a transformer branch extracts global contextual bonds by partitioning the image into patches by convolution-based patch embedding and then processing these patches through a transformer encoder. Analogously, a dedicated CNN branch captures fine-grained, local details through successive Incorporation these diverse features, combined with a self-supervised rotation prediction objective, produce a robust diagnostic representation that surpasses standard architectures in performance. Results show better performance not only in accuracy or loss function but also in these algorithms by utilizing capsule networks to preserve spatial orders and realize how each element individually combines and forms whole structures.
Authors: Syomantak Chaudhuri, Thomas A. Courtade
Abstract: Previous works in the differential privacy literature that allow users to choose their privacy levels typically operate under the heterogeneous differential privacy (HDP) framework with the simplifying assumption that user data and privacy levels are not correlated. Firstly, we demonstrate that the standard HDP framework falls short when user data and privacy demands are allowed to be correlated. Secondly, to address this shortcoming, we propose an alternate framework, Add-remove Heterogeneous Differential Privacy (AHDP), that jointly accounts for user data and privacy preference. We show that AHDP is robust to possible correlations between data and privacy. Thirdly, we formalize the guarantees of the proposed AHDP framework through an operational hypothesis testing perspective. The hypothesis testing setup may be of independent interest in analyzing other privacy frameworks as well. Fourthly, we show that there exists non-trivial AHDP mechanisms that notably do not require prior knowledge of the data-privacy correlations. We propose some such mechanisms and apply them to core statistical tasks such as mean estimation, frequency estimation, and linear regression. The proposed mechanisms are simple to implement with minimal assumptions and modeling requirements, making them attractive for real-world use. Finally, we empirically evaluate proposed AHDP mechanisms, highlighting their trade-offs using LLM-generated synthetic datasets, which we release for future research.
Authors: Zhicheng Tang, Jinwen Tang, Yi Shang
Abstract: In the realm of aerial imaging, the ability to detect small objects is pivotal for a myriad of applications, encompassing environmental surveillance, urban design, and crisis management. Leveraging RetinaNet, this work unveils DDR-Net: a data-driven, deep-learning model devised to enhance the detection of diminutive objects. DDR-Net introduces novel, data-driven techniques to autonomously ascertain optimal feature maps and anchor estimations, cultivating a tailored and proficient training process while maintaining precision. Additionally, this paper presents an innovative sampling technique to bolster model efficacy under limited data training constraints. The model's enhanced detection capabilities support critical applications including wildlife and habitat monitoring, traffic flow optimization, and public safety improvements through accurate identification of small objects like vehicles and pedestrians. DDR-Net significantly reduces the cost and time required for data collection and training, offering efficient performance even with limited data. Empirical assessments over assorted aerial avian imagery datasets demonstrate that DDR-Net markedly surpasses RetinaNet and alternative contemporary models. These innovations advance current aerial image analysis technologies and promise wide-ranging impacts across multiple sectors including agriculture, security, and archaeology.
Authors: Lesi Chen, Junru Li, Jingzhao Zhang
Abstract: This paper studies the complexity of finding an $\epsilon$-stationary point for stochastic bilevel optimization when the upper-level problem is nonconvex and the lower-level problem is strongly convex. Recent work proposed the first-order method, F${}^2$SA, achieving the $\tilde{\mathcal{O}}(\epsilon^{-6})$ upper complexity bound for first-order smooth problems. This is slower than the optimal $\Omega(\epsilon^{-4})$ complexity lower bound in its single-level counterpart. In this work, we show that faster rates are achievable for higher-order smooth problems. We first reformulate F$^2$SA as approximating the hyper-gradient with a forward difference. Based on this observation, we propose a class of methods F${}^2$SA-$p$ that uses $p$th-order finite difference for hyper-gradient approximation and improves the upper bound to $\tilde{\mathcal{O}}(p \epsilon^{4-p/2})$ for $p$th-order smooth problems. Finally, we demonstrate that the $\Omega(\epsilon^{-4})$ lower bound also holds for stochastic bilevel problems when the high-order smoothness holds for the lower-level variable, indicating that the upper bound of F${}^2$SA-$p$ is nearly optimal in the highly smooth region $p = \Omega( \log \epsilon^{-1} / \log \log \epsilon^{-1})$.
Authors: Renzhi Wu, Junjie Yang, Li Chen, Hong Li, Li Yu, Hong Yan
Abstract: Cross-domain recommendation systems face the challenge of integrating fine-grained user and item relationships across various product domains. To address this, we introduce RankGraph, a scalable graph learning framework designed to serve as a core component in recommendation foundation models (FMs). By constructing and leveraging graphs composed of heterogeneous nodes and edges across multiple products, RankGraph enables the integration of complex relationships between users, posts, ads, and other entities. Our framework employs a GPU-accelerated Graph Neural Network and contrastive learning, allowing for dynamic extraction of subgraphs such as item-item and user-user graphs to support similarity-based retrieval and real-time clustering. Furthermore, RankGraph integrates graph-based pretrained representations as contextual tokens into FM sequence models, enriching them with structured relational knowledge. RankGraph has demonstrated improvements in click (+0.92%) and conversion rates (+2.82%) in online A/B tests, showcasing its effectiveness in cross-domain recommendation scenarios.
Authors: Kaustuv Mukherji, Jaikrishna Manojkumar Patil, Dyuman Aditya, Paulo Shakarian, Devendra Parkar, Lahari Pokala, Clark Dorman, Gerardo I. Simari
Abstract: We introduce Lattice Annotated Temporal (LAT) Logic, an extension of Generalized Annotated Logic Programs (GAPs) that incorporates temporal reasoning and supports open-world semantics through the use of a lower lattice structure. This logic combines an efficient deduction process with temporal logic programming to support non-Markovian relationships and open-world reasoning capabilities. The open-world aspect, a by-product of the use of the lower-lattice annotation structure, allows for efficient grounding through a Skolemization process, even in domains with infinite or highly diverse constants. We provide a suite of theoretical results that bound the computational complexity of the grounding process, in addition to showing that many of the results on GAPs (using an upper lattice) still hold with the lower lattice and temporal extensions (though different proof techniques are required). Our open-source implementation, PyReason, features modular design, machine-level optimizations, and direct integration with reinforcement learning environments. Empirical evaluations across multi-agent simulations and knowledge graph tasks demonstrate up to three orders of magnitude speedup and up to five orders of magnitude memory reduction while maintaining or improving task performance. Additionally, we evaluate LAT Logic's value in reinforcement learning environments as a non-Markovian simulator, achieving up to three orders of magnitude faster simulation with improved agent performance, including a 26% increase in win rate due to capturing richer temporal dependencies. These results highlight LAT Logic's potential as a unified, extensible framework for open-world temporal reasoning in dynamic and uncertain environments. Our implementation is available at: pyreason.syracuse.edu.
Authors: Yifan Chen, Eric Vanden-Eijnden
Abstract: Flow-based generative models can face significant challenges when modeling scientific data with multiscale Fourier spectra, often producing large errors in fine-scale features. We address this problem within the framework of stochastic interpolants, via principled design of noise distributions and interpolation schedules. The key insight is that the noise should not be smoother than the target data distribution -- measured by Fourier spectrum decay rates -- to ensure bounded drift fields near the initial time. For Gaussian and near-Gaussian distributions whose fine-scale structure is known, we show that spectrum-matched noise improves numerical efficiency compared to standard white-noise approaches. For complex non-Gaussian distributions, we develop scale-adaptive interpolation schedules that address the numerical ill-conditioning arising from rougher-than-data noise. Numerical experiments on synthetic Gaussian random fields and solutions to the stochastic Allen-Cahn and Navier-Stokes equations validate our approach and demonstrate its ability to generate high-fidelity samples at lower computational cost than traditional approaches.
Authors: Fong-Chun Tsai, Kuan-Tang Huang, Bi-Cheng Yan, Tien-Hong Lo, Berlin Chen
Abstract: Automated Speaking Assessment (ASA) plays a crucial role in evaluating second-language (L2) learners proficiency. However, ASA models often suffer from class imbalance, leading to biased predictions. To address this, we introduce a novel objective for training ASA models, dubbed the Balancing Logit Variation (BLV) loss, which perturbs model predictions to improve feature representation for minority classes without modifying the dataset. Evaluations on the ICNALE benchmark dataset show that integrating the BLV loss into a celebrated text-based (BERT) model significantly enhances classification accuracy and fairness, making automated speech evaluation more robust for diverse learners.
Authors: Yangtao Deng, Lei Zhang, Qinlong Wang, Xiaoyun Zhi, Xinlei Zhang, Zhuo Jiang, Haohan Xu, Lei Wang, Zuquan Song, Gaohong Liu, Yang Bai, Shuguang Wang, Wencong Xiao, Jianxi Ye, Minlan Yu, Hong Xu
Abstract: Reliability is essential for ensuring efficiency in LLM training. However, many real-world reliability issues remain difficult to resolve, resulting in wasted resources and degraded model performance. Unfortunately, today's collective communication libraries operate as black boxes, hiding critical information needed for effective root cause analysis. We propose Mycroft, a lightweight distributed tracing and root cause analysis system designed to address previously hidden reliability issues in collective communication. Mycroft's key idea is to trace collective communication states and leverage internal control and data dependencies to resolve reliability problems in LLM training. Mycroft has been deployed at ByteDance for over six months to debug collective communication related issues at runtime. It detected anomalies within 15 seconds in 90% of cases and identified the root cause within 20 seconds in 60% of cases. We also conducted extensive fault injection experiments to demonstrate Mycroft's capability and efficiency.
Authors: Moontaha Nishat Chowdhury, Andr\'e Bauer, Minxuan Zhou
Abstract: In today's data-driven world, recommendation systems personalize user experiences across industries but rely on sensitive data, raising privacy concerns. Fully homomorphic encryption (FHE) can secure these systems, but a significant challenge in applying FHE to recommendation systems is efficiently handling the inherently large and sparse user-item rating matrices. FHE operations are computationally intensive, and naively processing various sparse matrices in recommendation systems would be prohibitively expensive. Additionally, the communication overhead between parties remains a critical concern in encrypted domains. We propose a novel approach combining Compressed Sparse Row (CSR) representation with FHE-based matrix factorization that efficiently handles matrix sparsity in the encrypted domain while minimizing communication costs. Our experimental results demonstrate high recommendation accuracy with encrypted data while achieving the lowest communication costs, effectively preserving user privacy.
Authors: Huaicheng Zhang, Ruoxin Wang, Chenlian Zhou, Jiguang Shi, Yue Ge, Zhoutong Li, Sheng Chang, Hao Wang, Jin He, Qijun Huang
Abstract: As one of the most effective methods for cardiovascular disease (CVD) diagnosis, multi-lead Electrocardiogram (ECG) signals present a characteristic multi-sensor information fusion challenge that has been continuously researched in deep learning domains. Despite the numerous algorithms proposed with different DL architectures, maintaining a balance among performance, computational complexity, and multi-source ECG feature fusion remains challenging. Recently, state space models (SSMs), particularly Mamba, have demonstrated remarkable effectiveness across various fields. Their inherent design for high-efficiency computation and linear complexity makes them particularly suitable for low-dimensional data like ECGs. This work proposes S2M2ECG, an SSM architecture featuring three-level fusion mechanisms: (1) Spatio-temporal bi-directional SSMs with segment tokenization for low-level signal fusion, (2) Intra-lead temporal information fusion with bi-directional scanning to enhance recognition accuracy in both forward and backward directions, (3) Cross-lead feature interaction modules for spatial information fusion. To fully leverage the ECG-specific multi-lead mechanisms inherent in ECG signals, a multi-branch design and lead fusion modules are incorporated, enabling individual analysis of each lead while ensuring seamless integration with others. Experimental results reveal that S2M2ECG achieves superior performance in the rhythmic, morphological, and clinical scenarios. Moreover, its lightweight architecture ensures it has nearly the fewest parameters among existing models, making it highly suitable for efficient inference and convenient deployment. Collectively, S2M2ECG offers a promising alternative that strikes an excellent balance among performance, computational complexity, and ECG-specific characteristics, paving the way for high-performance, lightweight computations in CVD diagnosis.
Authors: Derek Jones, Yue Yang, Felice C. Lightstone, Niema Moshiri, Jonathan E. Allen, Tajana S. Rosing
Abstract: Self-supervised pretraining from static structures of drug-like compounds and proteins enable powerful learned feature representations. Learned features demonstrate state of the art performance on a range of predictive tasks including molecular properties, structure generation, and protein-ligand interactions. The majority of approaches are limited by their use of static structures and it remains an open question, how best to use atomistic molecular dynamics (MD) simulations to develop more generalized models to improve prediction accuracy for novel molecular structures. We present SURrogate mmGBSA (SurGBSA) as a new modeling approach for MD-based representation learning, which learns a surrogate function of the Molecular Mechanics Generalized Born Surface Area (MMGBSA). We show for the first time the benefits of physics-informed pre-training to train a surrogate MMGBSA model on a collection of over 1.4 million 3D trajectories collected from MD simulations of the CASF-2016 benchmark. SurGBSA demonstrates a dramatic 6,497x speedup versus a traditional physics-based single-point MMGBSA calculation while nearly matching single-point MMGBSA accuracy on the challenging pose ranking problem for identification of the correct top pose (-0.4% difference). Our work advances the development of molecular foundation models by showing model improvements when training on MD simulations. Models, code and training data are made publicly available.
Authors: Cl\'ement Herv\'e, Paul Garnier, Jonathan Viquerat, Elie Hachem
Abstract: Intracranial aneurysms pose a significant clinical risk yet are difficult to detect, delineate and model due to limited annotated 3D data. We propose a cross-domain feature-transfer approach that leverages the latent geometric embeddings learned by TRELLIS, a generative model trained on large-scale non-medical 3D datasets, to augment neural networks for aneurysm analysis. By replacing conventional point normals or mesh descriptors with TRELLIS surface features, we systematically enhance three downstream tasks: (i) classifying aneurysms versus healthy vessels in the Intra3D dataset, (ii) segmenting aneurysm and vessel regions on 3D meshes, and (iii) predicting time-evolving blood-flow fields using a graph neural network on the AnXplore dataset. Our experiments show that the inclusion of these features yields strong gains in accuracy, F1-score and segmentation quality over state-of-the-art baselines, and reduces simulation error by 15\%. These results illustrate the broader potential of transferring 3D representations from general-purpose generative models to specialized medical tasks.
Authors: Yue Li, Xin Yi, Dongsheng Shi, Yongyi Cui, Gerard de Melo, Xiaoling Wang, Linlin Wang
Abstract: The intellectual property (IP) protection of Large Language Models (LLMs) is increasingly critical. Injecting specialized fingerprints into LLMs through instruction tuning is a common IP protection technique. However, this may significantly degrade model performance, requires substantial computational resources, and exhibits poor persistence under model modifications. We argue that knowledge editing offers a lightweight alternative that is more suitable for fingerprint injection. Accordingly, we apply knowledge editing to fingerprint injection for the first time and demonstrate its strong capability. Despite using scrambled text as fingerprints to prevent them from being overwritten during fine-tuning, degradation still occurs under large-scale fine-tuning. To address this, we propose Fingerprint Subspace-aware Fine-Tuning (FSFT), which reduces fingerprint degradation by constraining the update of the fingerprint subspace. The performance of FSFT exceeds fine-tuning by 10% even in the worst-case scenario. Additionally, we observe that the fingerprint-injected models struggle to distinguish between fingerprints and similar texts due to the high similarity of their features. This finding underscores the urgent need for more robust and fine-grained fingerprinting injection methods for LLMs.
Authors: Sashuai Zhou, Weinan Gan, Qijiong Liu, Ke Lei, Jieming Zhu, Hai Huang, Yan Xia, Ruiming Tang, Zhenhua Dong, Zhou Zhao
Abstract: Recent advances in LLM-based recommendation have shown promise, yet their cross-domain generalization is hindered by a fundamental mismatch between language-centric pretraining and the recommendation task. Existing methods, relying on language-level knowledge, fail to capture dynamic, item-level user interests across domains. To bridge this gap, we propose RecBase, a domain-agnostic foundational model pretrained with a recommendation-oriented objective. RecBase leverages a large-scale, heterogeneous, cross-domain corpus with unified textual representations and feature mappings to enhance cross-domain generalization. To further align item semantics across domains, we introduce a unified item tokenizer that encodes items into hierarchical concept identifiers, enabling structured representation and efficient vocabulary sharing. The model is trained using an autoregressive objective to capture complex item-level sequential patterns. On eight real-world datasets, our 1.5B-parameter model matches or surpasses the performance of LLM baselines up to 7B parameters in zero-shot and cross-domain recommendation tasks.
Authors: Mattia Litrico, Francesco Guarnera, Mario Valerio Giuffrida, Daniele Rav\`i, Sebastiano Battiato
Abstract: Generating realistic MRIs to accurately predict future changes in the structure of brain is an invaluable tool for clinicians in assessing clinical outcomes and analysing the disease progression at the patient level. However, current existing methods present some limitations: (i) some approaches fail to explicitly capture the relationship between structural changes and time intervals, especially when trained on age-imbalanced datasets; (ii) others rely only on scan interpolation, which lack clinical utility, as they generate intermediate images between timepoints rather than future pathological progression; and (iii) most approaches rely on 2D slice-based architectures, thereby disregarding full 3D anatomical context, which is essential for accurate longitudinal predictions. We propose a 3D Temporally-Aware Diffusion Model (TADM-3D), which accurately predicts brain progression on MRI volumes. To better model the relationship between time interval and brain changes, TADM-3D uses a pre-trained Brain-Age Estimator (BAE) that guides the diffusion model in the generation of MRIs that accurately reflect the expected age difference between baseline and generated follow-up scans. Additionally, to further improve the temporal awareness of TADM-3D, we propose the Back-In-Time Regularisation (BITR), by training TADM-3D to predict bidirectionally from the baseline to follow-up (forward), as well as from the follow-up to baseline (backward). Although predicting past scans has limited clinical applications, this regularisation helps the model generate temporally more accurate scans. We train and evaluate TADM-3D on the OASIS-3 dataset, and we validate the generalisation performance on an external test set from the NACC dataset. The code will be available upon acceptance.
Authors: Mattia Litrico, Feng Chen, Michael Pound, Sotirios A Tsaftaris, Sebastiano Battiato, Mario Valerio Giuffrida
Abstract: Crowd density estimation is a well-known computer vision task aimed at estimating the density distribution of people in an image. The main challenge in this domain is the reliance on fine-grained location-level annotations, (i.e. points placed on top of each individual) to train deep networks. Collecting such detailed annotations is both tedious, time-consuming, and poses a significant barrier to scalability for real-world applications. To alleviate this burden, we present Count2Density: a novel pipeline designed to predict meaningful density maps containing quantitative spatial information using only count-level annotations (i.e., total number of people) during training. To achieve this, Count2Density generates pseudo-density maps leveraging past predictions stored in a Historical Map Bank, thereby reducing confirmation bias. This bank is initialised using an unsupervised saliency estimator to provide an initial spatial prior and is iteratively updated with an EMA of predicted density maps. These pseudo-density maps are obtained by sampling locations from estimated crowd areas using a hypergeometric distribution, with the number of samplings determined by the count-level annotations. To further enhance the spatial awareness of the model, we add a self-supervised contrastive spatial regulariser to encourage similar feature representations within crowded regions while maximising dissimilarity with background regions. Experimental results demonstrate that our approach significantly outperforms cross-domain adaptation methods and achieves better results than recent state-of-the-art approaches in semi-supervised settings across several datasets. Additional analyses validate the effectiveness of each individual component of our pipeline, confirming the ability of Count2Density to effectively retrieve spatial information from count-level annotations and enabling accurate subregion counting.
Authors: Mingfeng Lin
Abstract: Coronary artery disease is a leading cause of mortality, underscoring the critical importance of precise diagnosis through X-ray angiography. Manual coronary artery segmentation from these images is time-consuming and inefficient, prompting the development of automated models. However, existing methods, whether rule-based or deep learning models, struggle with issues like poor performance and limited generalizability. Moreover, current knowledge distillation methods applied in this field have not fully exploited the hierarchical knowledge of the model, leading to certain information waste and insufficient enhancement of the model's performance capabilities for segmentation tasks. To address these issues, this paper introduces Deep Self-knowledge Distillation, a novel approach for coronary artery segmentation that leverages hierarchical outputs for supervision. By combining Deep Distribution Loss and Pixel-wise Self-knowledge Distillation Loss, our method enhances the student model's segmentation performance through a hierarchical learning strategy, effectively transferring knowledge from the teacher model. Our method combines a loosely constrained probabilistic distribution vector with tightly constrained pixel-wise supervision, providing dual regularization for the segmentation model while also enhancing its generalization and robustness. Extensive experiments on XCAD and DCA1 datasets demonstrate that our approach outperforms the dice coefficient, accuracy, sensitivity and IoU compared to other models in comparative evaluations.
Authors: Yaniv Goren, Yuval Cohen, Alexander Apartsin, Yehudit Aperstein
Abstract: In the realm of human-computer interaction, fostering a natural dialogue between humans and machines is paramount. A key, often overlooked, component of this dialogue is the use of interjections such as "mmm" and "hmm". Despite their frequent use to express agreement, hesitation, or requests for information, these interjections are typically dismissed as "non-words" by Automatic Speech Recognition (ASR) engines. Addressing this gap, we introduce a novel task dedicated to interjection classification, a pioneer in the field to our knowledge. This task is challenging due to the short duration of interjection signals and significant inter- and intra-speaker variability. In this work, we present and publish a dataset of interjection signals collected specifically for interjection classification. We employ this dataset to train and evaluate a baseline deep learning model. To enhance performance, we augment the training dataset using techniques such as tempo and pitch transformation, which significantly improve classification accuracy, making models more robust. The interjection dataset, a Python library for the augmentation pipeline, baseline model, and evaluation scripts, are available to the research community.
Authors: Xiaoxiao Xu, Hao Wu, Wenhui Yu, Lantao Hu, Peng Jiang, Kun Gai
Abstract: We propose a general model-agnostic Contrastive learning framework with Counterfactual Samples Synthesizing (CCSS) for modeling the monotonicity between the neural network output and numerical features which is critical for interpretability and effectiveness of recommender systems. CCSS models the monotonicity via a two-stage process: synthesizing counterfactual samples and contrasting the counterfactual samples. The two techniques are naturally integrated into a model-agnostic framework, forming an end-to-end training process. Abundant empirical tests are conducted on a publicly available dataset and a real industrial dataset, and the results well demonstrate the effectiveness of our proposed CCSS. Besides, CCSS has been deployed in our real large-scale industrial recommender, successfully serving over hundreds of millions users.
Authors: Leonidas Bakopoulos, Georgios Chalkiadakis
Abstract: Adaptive exploration methods propose ways to learn complex policies via alternating between exploration and exploitation. An important question for such methods is to determine the appropriate moment to switch between exploration and exploitation and vice versa. This is critical in domains that require the learning of long and complex sequences of actions. In this work, we present a generic adaptive exploration framework that employs uncertainty to address this important issue in a principled manner. Our framework includes previous adaptive exploration approaches as special cases. Moreover, we can incorporate in our framework any uncertainty-measuring mechanism of choice, for instance mechanisms used in intrinsic motivation or epistemic uncertainty-based exploration methods. We experimentally demonstrate that our framework gives rise to adaptive exploration strategies that outperform standard ones across several MuJoCo environments.
Authors: Sophia Bianchi Moyen, Rickmer Krohn, Sophie Lueth, Kay Pompetzki, Jan Peters, Vignesh Prasad, Georgia Chalvatzaki
Abstract: Intuitive Teleoperation interfaces are essential for mobile manipulation robots to ensure high quality data collection while reducing operator workload. A strong sense of embodiment combined with minimal physical and cognitive demands not only enhances the user experience during large-scale data collection, but also helps maintain data quality over extended periods. This becomes especially crucial for challenging long-horizon mobile manipulation tasks that require whole-body coordination. We compare two distinct robot control paradigms: a coupled embodiment integrating arm manipulation and base navigation functions, and a decoupled embodiment treating these systems as separate control entities. Additionally, we evaluate two visual feedback mechanisms: immersive virtual reality and conventional screen-based visualization of the robot's field of view. These configurations were systematically assessed across a complex, multi-stage task sequence requiring integrated planning and execution. Our results show that the use of VR as a feedback modality increases task completion time, cognitive workload, and perceived effort of the teleoperator. Coupling manipulation and navigation leads to a comparable workload on the user as decoupling the embodiments, while preliminary experiments suggest that data acquired by coupled teleoperation leads to better imitation learning performance. Our holistic view on intuitive teleoperation interfaces provides valuable insight into collecting high-quality, high-dimensional mobile manipulation data at scale with the human operator in mind. Project website:https://sophiamoyen.github.io/role-embodiment-wbc-moma-teleop/
URLs: https://sophiamoyen.github.io/role-embodiment-wbc-moma-teleop/
Authors: Siqi Xiang, Sheng Wang, Xiaokui Xiao, Cong Yue, Zhanhao Zhao, Beng Chin Ooi
Abstract: With the prevalence of in-database AI-powered analytics, there is an increasing demand for database systems to efficiently manage the ever-expanding number and size of deep learning models. However, existing database systems typically store entire models as monolithic files or apply compression techniques that overlook the structural characteristics of deep learning models, resulting in suboptimal model storage overhead. This paper presents NeurStore, a novel in-database model management system that enables efficient storage and utilization of deep learning models. First, NeurStore employs a tensor-based model storage engine to enable fine-grained model storage within databases. In particular, we enhance the hierarchical navigable small world (HNSW) graph to index tensors, and only store additional deltas for tensors within a predefined similarity threshold to ensure tensor-level deduplication. Second, we propose a delta quantization algorithm that effectively compresses delta tensors, thus achieving a superior compression ratio with controllable model accuracy loss. Finally, we devise a compression-aware model loading mechanism, which improves model utilization performance by enabling direct computation on compressed tensors. Experimental evaluations demonstrate that NeurStore achieves superior compression ratios and competitive model loading throughput compared to state-of-the-art approaches.
Authors: Babak Azkaei, Kishor Chandra Joshi, George Exarchakos
Abstract: The ever-increasing reliance of critical services on network infrastructure coupled with the increased operational complexity of beyond-5G/6G networks necessitate the need for proactive and automated network fault management. The provision for open interfaces among different radio access network\,(RAN) elements and the integration of AI/ML into network architecture enabled by the Open RAN\,(O-RAN) specifications bring new possibilities for active network health monitoring and anomaly detection. In this paper we leverage these advantages and develop an anomaly detection framework that proactively detect the possible throughput drops for a UE and minimize the post-handover failures. We propose two actionable anomaly detection algorithms tailored for real-world deployment. The first algorithm identifies user equipment (UE) at risk of severe throughput degradation by analyzing key performance indicators (KPIs) such as resource block utilization and signal quality metrics, enabling proactive handover initiation. The second algorithm evaluates neighbor cell radio coverage quality, filtering out cells with anomalous signal strength or interference levels. This reduces candidate targets for handover by 41.27\% on average. Together, these methods mitigate post-handover failures and throughput drops while operating much faster than the near-real-time latency constraints. This paves the way for self-healing 6G networks.
Authors: Dyah A. M. G. Wisnu, Ryandhimas E. Zezario, Stefano Rini, Hsin-Min Wang, Yu Tsao
Abstract: We present a system for automatic multi-axis perceptual quality prediction of generative audio, developed for Track 2 of the AudioMOS Challenge 2025. The task is to predict four Audio Aesthetic Scores--Production Quality, Production Complexity, Content Enjoyment, and Content Usefulness--for audio generated by text-to-speech (TTS), text-to-audio (TTA), and text-to-music (TTM) systems. A main challenge is the domain shift between natural training data and synthetic evaluation data. To address this, we combine BEATs, a pretrained transformer-based audio representation model, with a multi-branch long short-term memory (LSTM) predictor and use a triplet loss with buffer-based sampling to structure the embedding space by perceptual similarity. Our results show that this improves embedding discriminability and generalization, enabling domain-robust audio quality assessment without synthetic training data.
Authors: Napsu Karmitsa, Antti Airola, Tapio Pahikkala, Tinja Pitk\"am\"aki
Abstract: The increasing availability of personal data has enabled significant advances in fields such as machine learning, healthcare, and cybersecurity. However, this data abundance also raises serious privacy concerns, especially in light of powerful re-identification attacks and growing legal and ethical demands for responsible data use. Differential privacy (DP) has emerged as a principled, mathematically grounded framework for mitigating these risks. This review provides a comprehensive survey of DP, covering its theoretical foundations, practical mechanisms, and real-world applications. It explores key algorithmic tools and domain-specific challenges - particularly in privacy-preserving machine learning and synthetic data generation. The report also highlights usability issues and the need for improved communication and transparency in DP systems. Overall, the goal is to support informed adoption of DP by researchers and practitioners navigating the evolving landscape of data privacy.
Authors: Arnau Quera-Bofarull, Nicholas Bishop, Joel Dyer, Daniel Jarne Ornia, Anisoara Calinescu, Doyne Farmer, Michael Wooldridge
Abstract: Agent-based models (ABMs) simulate complex systems by capturing the bottom-up interactions of individual agents comprising the system. Many complex systems of interest, such as epidemics or financial markets, involve thousands or even millions of agents. Consequently, ABMs often become computationally demanding and rely on the calibration of numerous free parameters, which has significantly hindered their widespread adoption. In this paper, we demonstrate that automatic differentiation (AD) techniques can effectively alleviate these computational burdens. By applying AD to ABMs, the gradients of the simulator become readily available, greatly facilitating essential tasks such as calibration and sensitivity analysis. Specifically, we show how AD enables variational inference (VI) techniques for efficient parameter calibration. Our experiments demonstrate substantial performance improvements and computational savings using VI on three prominent ABMs: Axtell's model of firms; Sugarscape; and the SIR epidemiological model. Our approach thus significantly enhances the practicality and scalability of ABMs for studying complex systems.
Authors: Seokhun Park, Insung Kong, Yongdai Kim
Abstract: Bayesian Additive Regression Trees (BART) is a powerful statistical model that leverages the strengths of Bayesian inference and regression trees. It has received significant attention for capturing complex non-linear relationships and interactions among predictors. However, the accuracy of BART often comes at the cost of interpretability. To address this limitation, we propose ANOVA Bayesian Additive Regression Trees (ANOVA-BART), a novel extension of BART based on the functional ANOVA decomposition, which is used to decompose the variability of a function into different interactions, each representing the contribution of a different set of covariates or factors. Our proposed ANOVA-BART enhances interpretability, preserves and extends the theoretical guarantees of BART, and achieves superior predictive performance. Specifically, we establish that the posterior concentration rate of ANOVA-BART is nearly minimax optimal, and further provides the same convergence rates for each interaction that are not available for BART. Moreover, comprehensive experiments confirm that ANOVA-BART surpasses BART in both accuracy and uncertainty quantification, while also demonstrating its effectiveness in component selection. These results suggest that ANOVA-BART offers a compelling alternative to BART by balancing predictive accuracy, interpretability, and theoretical consistency.
Authors: Joel Jaskari, Chandreyee Roy, Fumiko Ogushi, Mikko Saukkoriipi, Jaakko Sahlsten, Kimmo Kaski
Abstract: Graph neural networks (GNNs) have emerged as a state-of-the-art data-driven tool for modeling connectivity data of graph-structured complex networks and integrating information of their nodes and edges in space and time. However, as of yet, the analysis of social networks using the time series of people's mobile connectivity data has not been extensively investigated. In the present study, we investigate four snapshot - based temporal GNNs in predicting the phone call and SMS activity between users of a mobile communication network. In addition, we develop a simple non - GNN baseline model using recently proposed EdgeBank method. Our analysis shows that the ROLAND temporal GNN outperforms the baseline model in most cases, whereas the other three GNNs perform on average worse than the baseline. The results show that GNN based approaches hold promise in the analysis of temporal social networks through mobile connectivity data. However, due to the relatively small performance margin between ROLAND and the baseline model, further research is required on specialized GNN architectures for temporal social network analysis.
Authors: Yewen Li, Jingtong Gao, Nan Jiang, Shuai Mao, Ruyi An, Fei Pan, Xiangyu Zhao, Bo An, Qingpeng Cai, Peng Jiang
Abstract: Auto-bidding is central to computational advertising, achieving notable commercial success by optimizing advertisers' bids within economic constraints. Recently, large generative models show potential to revolutionize auto-bidding by generating bids that could flexibly adapt to complex, competitive environments. Among them, diffusers stand out for their ability to address sparse-reward challenges by focusing on trajectory-level accumulated rewards, as well as their explainable capability, i.e., planning a future trajectory of states and executing bids accordingly. However, diffusers struggle with generation uncertainty, particularly regarding dynamic legitimacy between adjacent states, which can lead to poor bids and further cause significant loss of ad impression opportunities when competing with other advertisers in a highly competitive auction environment. To address it, we propose a Causal auto-Bidding method based on a Diffusion completer-aligner framework, termed CBD. Firstly, we augment the diffusion training process with an extra random variable t, where the model observes t-length historical sequences with the goal of completing the remaining sequence, thereby enhancing the generated sequences' dynamic legitimacy. Then, we employ a trajectory-level return model to refine the generated trajectories, aligning more closely with advertisers' objectives. Experimental results across diverse settings demonstrate that our approach not only achieves superior performance on large-scale auto-bidding benchmarks, such as a 29.9% improvement in conversion value in the challenging sparse-reward auction setting, but also delivers significant improvements on the Kuaishou online advertising platform, including a 2.0% increase in target cost.
Authors: Tien-Hong Lo, Szu-Yu Chen, Yao-Ting Sung, Berlin Chen
Abstract: A recent line of research on automated speaking assessment (ASA) has benefited from self-supervised learning (SSL) representations, which capture rich acoustic and linguistic patterns in non-native speech without underlying assumptions of feature curation. However, speech-based SSL models capture acoustic-related traits but overlook linguistic content, while text-based SSL models rely on ASR output and fail to encode prosodic nuances. Moreover, most prior arts treat proficiency levels as nominal classes, ignoring their ordinal structure and non-uniform intervals between proficiency labels. To address these limitations, we propose an effective ASA approach combining SSL with handcrafted indicator features via a novel modeling paradigm. We further introduce a multi-margin ordinal loss that jointly models both the score ordinality and non-uniform intervals of proficiency labels. Extensive experiments on the TEEMI corpus show that our method consistently outperforms strong baselines and generalizes well to unseen prompts.
Authors: Wu Lin, Scott C. Lowe, Felix Dangel, Runa Eschenhagen, Zikun Xu, Roger B. Grosse
Abstract: As an adaptive method, Shampoo employs a structured second-moment estimation, and its effectiveness has attracted growing attention. Prior work has primarily analyzed its estimation scheme through the Frobenius norm. Motivated by the natural connection between the second moment and a covariance matrix, we propose studying Shampoo's estimation as covariance estimation through the lens of Kullback-Leibler (KL) minimization. This alternative perspective reveals a previously hidden limitation, motivating improvements to Shampoo's design. Building on this insight, we develop a practical estimation scheme, termed KL-Shampoo, that eliminates Shampoo's reliance on Adam for stabilization, thereby removing the additional memory overhead introduced by Adam. Preliminary results show that KL-Shampoo improves Shampoo's performance, enabling it to stabilize without Adam and even outperform its Adam-stabilized variant, SOAP, in neural network pretraining.
Authors: Amirhossein Shahbazinia, Darong Huang, Luis Costero, David Atienza
Abstract: Cloud platforms are increasingly relied upon to host diverse, resource-intensive workloads due to their scalability, flexibility, and cost-efficiency. In multi-tenant cloud environments, virtual machines are consolidated on shared physical servers to improve resource utilization. While virtualization guarantees resource partitioning for CPU, memory, and storage, it cannot ensure performance isolation. Competition for shared resources such as last-level cache, memory bandwidth, and network interfaces often leads to severe performance degradation. Existing management techniques, including VM scheduling and resource provisioning, require accurate performance prediction to mitigate interference. However, this remains challenging in public clouds due to the black-box nature of VMs and the highly dynamic nature of workloads. To address these limitations, we propose CloudFormer, a dual-branch Transformer-based model designed to predict VM performance degradation in black-box environments. CloudFormer jointly models temporal dynamics and system-level interactions, leveraging 206 system metrics at one-second resolution across both static and dynamic scenarios. This design enables the model to capture transient interference effects and adapt to varying workload conditions without scenario-specific tuning. Complementing the methodology, we provide a fine-grained dataset that significantly expands the temporal resolution and metric diversity compared to existing benchmarks. Experimental results demonstrate that CloudFormer consistently outperforms state-of-the-art baselines across multiple evaluation metrics, achieving robust generalization across diverse and previously unseen workloads. Notably, CloudFormer attains a mean absolute error (MAE) of just 7.8%, representing a substantial improvement in predictive accuracy and outperforming existing methods at least by 28%.
Authors: Mohammed Amer, Mohamed A. Suliman, Tu Bui, Nuria Garcia, Serban Georgescu
Abstract: Healthcare applications are inherently multimodal, benefiting greatly from the integration of diverse data sources. However, the modalities available in clinical settings can vary across different locations and patients. A key area that stands to gain from multimodal integration is breast cancer molecular subtyping, an important clinical task that can facilitate personalized treatment and improve patient prognosis. In this work, we propose a scalable and loosely-coupled multimodal framework that seamlessly integrates data from various modalities, including copy number variation (CNV), clinical records, and histopathology images, to enhance breast cancer subtyping. While our primary focus is on breast cancer, our framework is designed to easily accommodate additional modalities, offering the flexibility to scale up or down with minimal overhead without requiring re-training of existing modalities, making it applicable to other types of cancers as well. We introduce a dual-based representation for whole slide images (WSIs), combining traditional image-based and graph-based WSI representations. This novel dual approach results in significant performance improvements. Moreover, we present a new multimodal fusion strategy, demonstrating its ability to enhance performance across a range of multimodal conditions. Our comprehensive results show that integrating our dual-based WSI representation with CNV and clinical health records, along with our pipeline and fusion strategy, outperforms state-of-the-art methods in breast cancer subtyping.
Authors: Benjamin Heymann, Otmane Sakhi
Abstract: We consider the problem of directly optimizing a non-linear function of an outcome, where this outcome itself is the sum of many small contributions. The non-linearity of the function means that the problem is not equivalent to the maximization of the expectation of the individual contribution. By leveraging the concentration properties of the sum of individual outcomes, we derive a scalable descent algorithm that directly optimizes for our stated objective. This allows for instance to maximize the probability of successful A/B test, for which it can be wiser to target a success criterion, such as exceeding a given uplift, rather than chasing the highest expected payoff.
Authors: Imad Aouali, Otmane Sakhi
Abstract: Off-policy evaluation (OPE) and off-policy learning (OPL) are foundational for decision-making in offline contextual bandits. Recent advances in OPL primarily optimize OPE estimators with improved statistical properties, assuming that better estimators inherently yield superior policies. Although theoretically justified, we argue this estimator-centric approach neglects a critical practical obstacle: challenging optimization landscapes. In this paper, we provide theoretical insights and extensive empirical evidence showing that current OPL methods encounter severe optimization issues, particularly as action spaces become large. We demonstrate that simpler weighted log-likelihood objectives enjoy substantially better optimization properties and still recover competitive, often superior, learned policies. Our findings emphasize the necessity of explicitly addressing optimization considerations in the development of OPL algorithms for large action spaces.
Authors: Hong Ye Tan, Subhadip Mukherjee, Junqi Tang
Abstract: Inverse problems lie at the heart of modern imaging science, with broad applications in areas such as medical imaging, remote sensing, and microscopy. Recent years have witnessed a paradigm shift in solving imaging inverse problems, where data-driven regularizers are used increasingly, leading to remarkably high-fidelity reconstruction. A particularly notable approach for data-driven regularization is to use learned image denoisers as implicit priors in iterative image reconstruction algorithms. This survey presents a comprehensive overview of this powerful and emerging class of algorithms, commonly referred to as plug-and-play (PnP) methods. We begin by providing a brief background on image denoising and inverse problems, followed by a short review of traditional regularization strategies. We then explore how proximal splitting algorithms, such as the alternating direction method of multipliers (ADMM) and proximal gradient descent (PGD), can naturally accommodate learned denoisers in place of proximal operators, and under what conditions such replacements preserve convergence. The role of Tweedie's formula in connecting optimal Gaussian denoisers and score estimation is discussed, which lays the foundation for regularization-by-denoising (RED) and more recent diffusion-based posterior sampling methods. We discuss theoretical advances regarding the convergence of PnP algorithms, both within the RED and proximal settings, emphasizing the structural assumptions that the denoiser must satisfy for convergence, such as non-expansiveness, Lipschitz continuity, and local homogeneity. We also address practical considerations in algorithm design, including choices of denoiser architecture and acceleration strategies.
Authors: Thinh Viet Le, Md Obaidur Rahman, Vassilis Kekatos
Abstract: Interconnection studies require solving numerous instances of the AC load or power flow (AC PF) problem to simulate diverse scenarios as power systems navigate the ongoing energy transition. To expedite such studies, this work leverages recent advances in quantum computing to find or predict AC PF solutions using a variational quantum circuit (VQC). VQCs are trainable models that run on modern-day noisy intermediate-scale quantum (NISQ) hardware to accomplish elaborate optimization and machine learning (ML) tasks. Our first contribution is to pose a single instance of the AC PF as a nonlinear least-squares fit over the VQC trainable parameters (weights) and solve it using a hybrid classical/quantum computing approach. The second contribution is to feed PF specifications as features into a data-embedded VQC and train the resultant quantum ML (QML) model to predict general PF solutions. The third contribution is to develop a novel protocol to efficiently measure AC-PF quantum observables by exploiting the graph structure of a power network. Preliminary numerical tests indicate that the proposed VQC models attain enhanced prediction performance over a deep neural network despite using much fewer weights. The proposed quantum AC-PF framework sets the foundations for addressing more elaborate grid tasks via quantum computing.
Authors: Honglu Zhou, Xiangyu Peng, Shrikant Kendre, Michael S. Ryoo, Silvio Savarese, Caiming Xiong, Juan Carlos Niebles
Abstract: Next-generation AI companions must go beyond general video understanding to resolve spatial and temporal references in dynamic, real-world environments. Existing Video Large Language Models (Video LLMs), while capable of coarse-level comprehension, struggle with fine-grained, spatiotemporal reasoning, especially when user queries rely on time-based event references for temporal anchoring, or gestural cues for spatial anchoring to clarify object references and positions. To bridge this critical gap, we introduce Strefer, a synthetic instruction data generation framework designed to equip Video LLMs with spatiotemporal referring and reasoning capabilities. Strefer produces diverse instruction-tuning data using a data engine that pseudo-annotates temporally dense, fine-grained video metadata, capturing rich spatial and temporal information in a structured manner, including subjects, objects, their locations as masklets, and their action descriptions and timelines. Our approach enhances the ability of Video LLMs to interpret spatial and temporal references, fostering more versatile, space-time-aware reasoning essential for real-world AI companions. Without using proprietary models, costly human annotation, or the need to annotate large volumes of new videos, experimental evaluations show that models trained with data produced by Strefer outperform baselines on tasks requiring spatial and temporal disambiguation. Additionally, these models exhibit enhanced space-time-aware reasoning, establishing a new foundation for perceptually grounded, instruction-tuned Video LLMs.
Authors: Yanlin Zhang, Sungyong Chung, Nachuan Li, Dana Monzer, Hani S. Mahmassani, Samer H. Hamdar, Alireza Talebpour
Abstract: The Waymo Open Motion Dataset (WOMD) has become a popular resource for data-driven modeling of autonomous vehicles (AVs) behavior. However, its validity for behavioral analysis remains uncertain due to proprietary post-processing, the absence of error quantification, and the segmentation of trajectories into 20-second clips. This study examines whether WOMD accurately captures the dynamics and interactions observed in real-world AV operations. Leveraging an independently collected naturalistic dataset from Level 4 AV operations in Phoenix, Arizona (PHX), we perform comparative analyses across three representative urban driving scenarios: discharging at signalized intersections, car-following, and lane-changing behaviors. For the discharging analysis, headways are manually extracted from aerial video to ensure negligible measurement error. For the car-following and lane-changing cases, we apply the Simulation-Extrapolation (SIMEX) method to account for empirically estimated error in the PHX data and use Dynamic Time Warping (DTW) distances to quantify behavioral differences. Results across all scenarios consistently show that behavior in PHX falls outside the behavioral envelope of WOMD. Notably, WOMD underrepresents short headways and abrupt decelerations. These findings suggest that behavioral models calibrated solely on WOMD may systematically underestimate the variability, risk, and complexity of naturalistic driving. Caution is therefore warranted when using WOMD for behavior modeling without proper validation against independently collected data.
Authors: Yi Tian, Kaiqing Zhang, Russ Tedrake, Suvrit Sra
Abstract: We study the task of learning state representations from potentially high-dimensional observations, with the goal of controlling an unknown partially observable system. We pursue a cost-driven approach, where a dynamic model in some latent state space is learned by predicting the costs without predicting the observations or actions. In particular, we focus on an intuitive cost-driven state representation learning method for solving Linear Quadratic Gaussian (LQG) control, one of the most fundamental partially observable control problems. As our main results, we establish finite-sample guarantees of finding a near-optimal state representation function and a near-optimal controller using the directly learned latent model, for finite-horizon time-varying LQG control problems. To the best of our knowledge, despite various empirical successes, finite-sample guarantees of such a cost-driven approach remain elusive. Our result underscores the value of predicting multi-step costs, an idea that is key to our theory, and notably also an idea that is known to be empirically valuable for learning state representations. A second part of this work, that is to appear as Part II, addresses the infinite-horizon linear time-invariant setting; it also extends the results to an approach that implicitly learns the latent dynamics, inspired by the recent empirical breakthrough of MuZero in model-based reinforcement learning.
Authors: Yewei Xu, Shi Chen, Qin Li
Abstract: Does the use of auto-differentiation yield reasonable updates for deep neural networks (DNNs)? Specifically, when DNNs are designed to adhere to neural ODE architectures, can we trust the gradients provided by auto-differentiation? Through mathematical analysis and numerical evidence, we demonstrate that when neural networks employ high-order methods, such as Linear Multistep Methods (LMM) or Explicit Runge-Kutta Methods (ERK), to approximate the underlying ODE flows, brute-force auto-differentiation often introduces artificial oscillations in the gradients that prevent convergence. In the case of Leapfrog and 2-stage ERK, we propose simple post-processing techniques that effectively eliminates these oscillations, correct the gradient computation and thus returns the accurate updates.
Authors: Eslam Abdelaleem, Ilya Nemenman, K. Michael Martini
Abstract: Variational dimensionality reduction methods are widely used for their accuracy, generative capabilities, and robustness. We introduce a unifying framework that generalizes both such as traditional and state-of-the-art methods. The framework is based on an interpretation of the multivariate information bottleneck, trading off the information preserved in an encoder graph (defining what to compress) against that in a decoder graph (defining a generative model for data). Using this approach, we rederive existing methods, including the deep variational information bottleneck, variational autoencoders, and deep multiview information bottleneck. We naturally extend the deep variational CCA (DVCCA) family to beta-DVCCA and introduce a new method, the deep variational symmetric information bottleneck (DVSIB). DSIB, the deterministic limit of DVSIB, connects to modern contrastive learning approaches such as Barlow Twins, among others. We evaluate these methods on Noisy MNIST and Noisy CIFAR-100, showing that algorithms better matched to the structure of the problem like DVSIB and beta-DVCCA produce better latent spaces as measured by classification accuracy, dimensionality of the latent variables, sample efficiency, and consistently outperform other approaches under comparable conditions. Additionally, we benchmark against state-of-the-art models, achieving superior or competitive accuracy. Our results demonstrate that this framework can seamlessly incorporate diverse multi-view representation learning algorithms, providing a foundation for designing novel, problem-specific loss functions.
Authors: Zhiyuan Wang, Xiaoyang Qu, Jing Xiao, Bokui Chen, Jianzong Wang
Abstract: Catastrophic forgetting poses a substantial challenge for managing intelligent agents controlled by a large model, causing performance degradation when these agents face new tasks. In our work, we propose a novel solution - the Progressive Prompt Decision Transformer (P2DT). This method enhances a transformer-based model by dynamically appending decision tokens during new task training, thus fostering task-specific policies. Our approach mitigates forgetting in continual and offline reinforcement learning scenarios. Moreover, P2DT leverages trajectories collected via traditional reinforcement learning from all tasks and generates new task-specific tokens during training, thereby retaining knowledge from previous studies. Preliminary results demonstrate that our model effectively alleviates catastrophic forgetting and scales well with increasing task environments.
Authors: Zhiyuan Wang, Xiaoyang Qu, Jing Xiao, Bokui Chen, Jianzong Wang
Abstract: This paper introduces INCPrompt, an innovative continual learning solution that effectively addresses catastrophic forgetting. INCPrompt's key innovation lies in its use of adaptive key-learner and task-aware prompts that capture task-relevant information. This unique combination encapsulates general knowledge across tasks and encodes task-specific knowledge. Our comprehensive evaluation across multiple continual learning benchmarks demonstrates INCPrompt's superiority over existing algorithms, showing its effectiveness in mitigating catastrophic forgetting while maintaining high performance. These results highlight the significant impact of task-aware incremental prompting on continual learning performance.
Authors: Tianyue Zhou, Jung-Hoon Cho, Cathy Wu
Abstract: Recommendation systems now pervade the digital world, ranging from advertising to entertainment. However, it remains challenging to implement effective recommendation systems in the physical world, such as in mobility or health. This work focuses on a key challenge: in the physical world, it is often easy for the user to opt out of taking any recommendation if they are not to her liking, and to fall back to her baseline behavior. It is thus crucial in cyber-physical recommendation systems to operate with an interaction model that is aware of such user behavior, lest the user abandon the recommendations altogether. This paper thus introduces the Nah Bandit, a tongue-in-cheek reference to describe a Bandit problem where users can say `nah' to the recommendation and opt for their preferred option instead. As such, this problem lies in between a typical bandit setup and supervised learning. We model the user non-compliance by parameterizing an anchoring effect of recommendations on users. We then propose the Expert with Clustering (EWC) algorithm, a hierarchical approach that incorporates feedback from both recommended and non-recommended options to accelerate user preference learning. In a recommendation scenario with $N$ users, $T$ rounds per user, and $K$ clusters, EWC achieves a regret bound of $O(N\sqrt{T\log K} + NT)$, achieving superior theoretical performance in the short term compared to LinUCB algorithm. Experimental results also highlight that EWC outperforms both supervised learning and traditional contextual bandit approaches. This advancement reveals that effective use of non-compliance feedback can accelerate preference learning and improve recommendation accuracy. This work lays the foundation for future research in Nah Bandit, providing a robust framework for more effective recommendation systems.
Authors: Yuhang Yao, Yuan Li, Xinyi Fan, Junhao Li, Kay Liu, Weizhao Jin, Yu Yang, Srivatsan Ravi, Philip S. Yu, Carlee Joe-Wong
Abstract: Federated graph learning is an emerging field with significant practical challenges. While algorithms have been proposed to improve the accuracy of training graph neural networks, such as node classification on federated graphs, the system performance is often overlooked, despite it is crucial for real-world deployment. To bridge this gap, we introduce FedGraph, a research library designed for practical distributed training and comprehensive benchmarking of FGL algorithms. FedGraph supports a range of state-of-the-art graph learning methods and includes a monitoring class that evaluates system performance, with a particular focus on communication and computation costs during training. Unlike existing federated learning platforms, FedGraph natively integrates homomorphic encryption to enhance privacy preservation and supports scalable deployment across multiple physical machines with system-level performance evaluation to guide the system design of future algorithms. To enhance efficiency and privacy, we propose a low-rank communication scheme for algorithms like FedGCN that require pre-training communication, accelerating both the pre-training and training phases. Extensive experiments benchmark FGL algorithms on three major graph learning tasks and demonstrate FedGraph as the first efficient FGL framework to support encrypted low-rank communication and scale to graphs with 100 million nodes.
Authors: Yunkai Dang, Mengxi Gao, Yibo Yan, Xin Zou, Yanggan Gu, Jungang Li, Jingyu Wang, Peijie Jiang, Aiwei Liu, Jia Liu, Xuming Hu
Abstract: Multimodal large language models (MLLMs) have recently achieved state-of-the-art performance on tasks ranging from visual question answering to video understanding. However, existing studies have concentrated mainly on visual-textual misalignment, leaving largely unexplored the MLLMs' ability to preserve an originally correct answer when confronted with misleading information. We reveal a response uncertainty phenomenon: across nine standard datasets, twelve state-of-the-art open-source MLLMs overturn a previously correct answer in 65% of cases after receiving a single deceptive cue. To systematically quantify this vulnerability, we propose a two-stage evaluation pipeline: (1) elicit each model's original response on unperturbed inputs; (2) inject explicit (false-answer hints) and implicit (contextual contradictions) misleading instructions, and compute the misleading rate - the fraction of correct-to-incorrect flips. Leveraging the most susceptible examples, we curate the Multimodal Uncertainty Benchmark (MUB), a collection of image-question pairs stratified into low, medium, and high difficulty based on how many of twelve state-of-the-art MLLMs they mislead. Extensive evaluation on twelve open-source and five closed-source models reveals a high uncertainty: average misleading rates exceed 86%, with explicit cues over 67.19% and implicit cues over 80.67%. To reduce the misleading rate, we then fine-tune all open-source MLLMs on a compact 2000-sample mixed-instruction dataset, reducing misleading rates to 6.97% (explicit) and 32.77% (implicit), boosting consistency by nearly 29.37% on highly deceptive inputs, and slightly improving accuracy on standard benchmarks. Our code is available at https://github.com/Yunkaidang/uncertainty
Authors: Tengjie Zheng, Haipeng Chen, Lin Cheng, Shengping Gong, Xu Huang
Abstract: Learning dynamical models from data is not only fundamental but also holds great promise for advancing principle discovery, time-series prediction, and controller design. Among various approaches, Gaussian Process State-Space Models (GPSSMs) have recently gained significant attention due to their combination of flexibility and interpretability. However, for online learning, the field lacks an efficient method suitable for scenarios where prior information regarding data distribution and model function is limited. To address this issue, this paper proposes a recursive GPSSM method with adaptive capabilities for both operating domains and Gaussian process (GP) hyperparameters. Specifically, we first utilize first-order linearization to derive a Bayesian update equation for the joint distribution between the system state and the GP model, enabling closed-form and domain-independent learning. Second, an online selection algorithm for inducing points is developed based on informative criteria to achieve lightweight learning. Third, to support online hyperparameter optimization, we recover historical measurement information from the current filtering distribution. Comprehensive evaluations on both synthetic and real-world datasets demonstrate the superior accuracy, computational efficiency, and adaptability of our method compared to state-of-the-art online GPSSM techniques.
Authors: Haeyong Kang, Chang D. Yoo
Abstract: Inspired by the Well-initialized Lottery Ticket Hypothesis (WLTH), which provides suboptimal fine-tuning solutions, we propose a novel fully fine-tuned continual learning (CL) method referred to as Soft-TransFormers (Soft-TF). Soft-TF sequentially learns and selects an optimal soft-network for each task. During sequential training in CL, a well-initialized Soft-TF mask optimizes the weights of sparse layers to obtain task-adaptive soft (real-valued) networks, while keeping the well-pre-trained layer parameters frozen. In inference, the identified task-adaptive network of Soft-TF masks the parameters of the pre-trained network, mapping to an optimal solution for each task and minimizing Catastrophic Forgetting (CF) - the soft-masking preserves the knowledge of the pre-trained network. Extensive experiments on the Vision Transformer (ViT) and the Language Transformer (Bert) demonstrate the effectiveness of Soft-TF, achieving state-of-the-art performance across Vision and Language Class Incremental Learning (CIL) scenarios.
Authors: Masanori Ishikura, Masayuki Karasuyama
Abstract: This study considers multi-objective Bayesian optimization (MOBO) through the information gain of the Pareto-frontier. To calculate the information gain, a predictive distribution conditioned on the Pareto-frontier plays a key role, which is defined as a distribution truncated by the Pareto-frontier. However, it is usually impossible to obtain the entire Pareto-frontier in a continuous domain, and therefore, the complete truncation cannot be known. We consider an approximation of the truncate distribution by using a mixture distribution consisting of two possible approximate truncation obtainable from a subset of the Pareto-frontier, which we call over- and under-truncation. Since the optimal balance of the mixture is unknown beforehand, we propose optimizing the balancing coefficient through the variational lower bound maximization framework, by which the approximation error of the information gain can be minimized. Our empirical evaluation demonstrates the effectiveness of the proposed method particularly when the number of objective functions is large.
Authors: Srinitish Srinivasan, Omkumar CU
Abstract: Graph representation learning has emerged as a cornerstone for tasks like node classification and link prediction, yet prevailing self-supervised learning (SSL) methods face challenges such as computational inefficiency, reliance on contrastive objectives, and representation collapse. Existing approaches often depend on feature reconstruction, negative sampling, or complex decoders, which introduce training overhead and hinder generalization. Further, current techniques which address such limitations fail to account for the contribution of node embeddings to a certain prediction in the absence of labeled nodes. To address these limitations, we propose a novel joint embedding predictive framework for graph SSL that eliminates contrastive objectives and negative sampling while preserving semantic and structural information. Additionally, we introduce a semantic-aware objective term that incorporates pseudo-labels derived from Gaussian Mixture Models (GMMs), enhancing node discriminability by evaluating latent feature contributions. Extensive experiments demonstrate that our framework outperforms state-of-the-art graph SSL methods across benchmarks, achieving superior performance without contrastive loss or complex decoders. Key innovations include (1) a non-contrastive, view-invariant joint embedding predictive architecture, (2) Leveraging single context and multiple targets relationship between subgraphs, and (3) GMM-based pseudo-label scoring to capture semantic contributions. This work advances graph SSL by offering a computationally efficient, collapse-resistant paradigm that bridges spatial and semantic graph features for downstream tasks. The code for our paper can be found at https://github.com/Deceptrax123/JPEB-GSSL
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: Abhilash Neog, Arka Daw, Sepideh Fatemi Khorasgani, Medha Sawhney, Aanish Pradhan, Mary E. Lofton, Bennett J. McAfee, Adrienne Breef-Pilz, Heather L. Wander, Dexter W Howard, Cayelan C. Carey, Paul Hanson, Anuj Karpatne
Abstract: Modeling Irregularly-sampled and Multivariate Time Series (IMTS) is crucial across a variety of applications where different sets of variates may be missing at different time-steps due to sensor malfunctions or high data acquisition costs. Existing approaches for IMTS either consider a two-stage impute-then-model framework or involve specialized architectures specific to a particular model and task. We perform a series of experiments to derive novel insights about the performance of IMTS methods on a variety of semi-synthetic and real-world datasets for both classification and forecasting. We also introduce Missing Feature-aware Time Series Modeling (MissTSM) or MissTSM, a novel model-agnostic and imputation-free approach for IMTS modeling. We show that MissTSM shows competitive performance compared to other IMTS approaches, especially when the amount of missing values is large and the data lacks simplistic periodic structures - conditions common to real-world IMTS applications.
Authors: Andy Gray, Alma Rahat, Tom Crick, Stephen Lindsay
Abstract: Comparative Judgement (CJ) provides an alternative assessment approach by evaluating work holistically rather than breaking it into discrete criteria. This method leverages human ability to make nuanced comparisons, yielding more reliable and valid assessments. CJ aligns with real-world evaluations, where overall quality emerges from the interplay of various elements. However, rubrics remain widely used in education, offering structured criteria for grading and detailed feedback. This creates a gap between CJ's holistic ranking and the need for criterion-based performance breakdowns. This paper addresses this gap using a Bayesian approach. We build on Bayesian CJ (BCJ) by Gray et al., which directly models preferences instead of using likelihoods over total scores, allowing for expected ranks with uncertainty estimation. Their entropy-based active learning method selects the most informative pairwise comparisons for assessors. We extend BCJ to handle multiple independent learning outcome (LO) components, defined by a rubric, enabling both holistic and component-wise predictive rankings with uncertainty estimates. Additionally, we propose a method to aggregate entropies and identify the most informative comparison for assessors. Experiments on synthetic and real data demonstrate our method's effectiveness. Finally, we address a key limitation of BCJ, which is the inability to quantify assessor agreement. We show how to derive agreement levels, enhancing transparency in assessment.
Authors: Yifei He, Yang Liu, Chen Liang, Hany Hassan Awadalla
Abstract: Mixture-of-Experts (MoE) models have become a key approach for scaling large language models efficiently by activating only a subset of experts during training and inference. Typically, the number of activated experts presents a trade-off: fewer experts reduce computational costs, while more experts improve performance. Recent studies reveal that not all activated experts contribute equally to model performance, with some providing minimal utility, particularly when finetuning pretrained MoE models for specialized downstream tasks. The co-existence of significant and redundant parameters in experts provides us an opportunity to reduce the number of activated experts while maintaining model performance. In this work, we propose the concept of compressed experts, lightweight modules that serve as compact representations of full experts. Our approach preserves the most important experts while replacing other auxiliary activated experts with compressed experts. The reduction of active parameters significantly lowers inference costs while achieving comparable performance. Extensive experiments on models including Phi-MoE and OLMoE demonstrate that compressed experts recover over 90% of full expert performance across various tasks while reducing more than 30% active parameters and saving 20% in inference costs. This approach enables efficient deployment of MoE models in resource-constrained settings and facilitates scaling to larger models with manageable overhead. Our code is available at https://github.com/yifei-he/Compressed-Experts.
Authors: Raphael Trumpp, Ansgar Sch\"afftlein, Mirco Theile, Marco Caccamo
Abstract: As image-based deep reinforcement learning tackles more challenging tasks, increasing model size has become an important factor in improving performance. Recent studies achieved this by focusing on the parameter efficiency of scaled networks, typically using Impala-CNN, a 15-layer ResNet-inspired network, as the image encoder. However, while Impala-CNN evidently outperforms older CNN architectures, potential advancements in network design for deep reinforcement learning-specific image encoders remain largely unexplored. We find that replacing the flattening of output feature maps in Impala-CNN with global average pooling leads to a notable performance improvement. This approach outperforms larger and more complex models in the Procgen Benchmark, particularly in terms of generalization. We call our proposed encoder model Impoola-CNN. A decrease in the network's translation sensitivity may be central to this improvement, as we observe the most significant gains in games without agent-centered observations. Our results demonstrate that network scaling is not just about increasing model size - efficient network design is also an essential factor. We make our code available at https://github.com/raphajaner/impoola.
Authors: Naoufal El Bekri, Lucas Drumetz, Franck Vermet
Abstract: Solving the Fokker-Planck equation for high-dimensional complex dynamical systems remains a pivotal yet challenging task due to the intractability of analytical solutions and the limitations of traditional numerical methods. In this work, we present FlowKac, a novel approach that reformulates the Fokker-Planck equation using the Feynman-Kac formula, allowing to query the solution at a given point via the expected values of stochastic paths. A key innovation of FlowKac lies in its adaptive stochastic sampling scheme which significantly reduces the computational complexity while maintaining high accuracy. This sampling technique, coupled with a time-indexed normalizing flow, designed for capturing time-evolving probability densities, enables robust sampling of collocation points, resulting in a flexible and mesh-free solver. This formulation mitigates the curse of dimensionality and enhances computational efficiency and accuracy, which is particularly crucial for applications that inherently require dimensions beyond the conventional three. We validate the robustness and scalability of our method through various experiments on a range of stochastic differential equations, demonstrating significant improvements over existing techniques.
Authors: Ertu\u{g}rul Ke\c{c}eci, M\"ujde G\"uzelkaya, Tufan Kumbasar
Abstract: This paper presents FedAlign, a Federated Learning (FL) framework particularly designed for System Identification (SYSID) tasks by aligning state representations. Local workers can learn State-Space Models (SSMs) with equivalent representations but different dynamics. We demonstrate that directly aggregating these local SSMs via FedAvg results in a global model with altered system dynamics. FedAlign overcomes this problem by employing similarity transformation matrices to align state representations of local SSMs, thereby establishing a common parameter basin that retains the dynamics of local SSMs. FedAlign computes similarity transformation matrices via two distinct approaches: FedAlign-A and FedAlign-O. In FedAlign-A, we represent the global SSM in controllable canonical form (CCF). We apply control theory to analytically derive similarity transformation matrices that convert each local SSM into this form. Yet, establishing global SSM in CCF brings additional alignment challenges in multi input - multi output SYSID as CCF representation is not unique, unlike in single input - single output SYSID. In FedAlign-O, we address these alignment challenges by reformulating the local parameter basin alignment problem as an optimization task. We determine the parameter basin of a local worker as the common parameter basin and solve least square problems to obtain similarity transformation matrices needed to align the remaining local SSMs. Through the experiments conducted on synthetic and real-world datasets, we show that FedAlign outperforms FedAvg, converges faster, and provides improved stability of the global SSM thanks to the efficient alignment of local parameter basins.
Authors: Nathan P. Lawrence, Thomas Banker, Ali Mesbah
Abstract: The reinforcement learning (RL) and model predictive control (MPC) communities have developed vast ecosystems of theoretical approaches and computational tools for solving optimal control problems. Given their conceptual similarities but differing strengths, there has been increasing interest in synergizing RL and MPC. However, existing approaches tend to be limited for various reasons, including computational cost of MPC in an RL algorithm and software hurdles towards seamless integration of MPC and RL tools. These challenges often result in the use of "simple" MPC schemes or RL algorithms, neglecting the state-of-the-art in both areas. This paper presents MPCritic, a machine learning-friendly architecture that interfaces seamlessly with MPC tools. MPCritic utilizes the loss landscape defined by a parameterized MPC problem, focusing on "soft" optimization over batched training steps; thereby updating the MPC parameters while avoiding costly minimization and parametric sensitivities. Since the MPC structure is preserved during training, an MPC agent can be readily used for online deployment, where robust constraint satisfaction is paramount. We demonstrate the versatility of MPCritic, in terms of MPC architectures and RL algorithms that it can accommodate, on classic control benchmarks.
Authors: Zinuo You, John Cartlidge, Karen Elliott, Menghan Ge, Daniel Gold
Abstract: Existing black-box portfolio management systems are prevalent in the financial industry due to commercial and safety constraints, though their performance can fluctuate dramatically with changing market regimes. Evaluating these non-transparent systems is computationally expensive, as fixed budgets limit the number of possible observations. Therefore, achieving stable and sample-efficient optimization for these systems has become a critical challenge. This work presents a novel Bayesian optimization framework (TPE-AS) that improves search stability and efficiency for black-box portfolio models under these limited observation budgets. Standard Bayesian optimization, which solely maximizes expected return, can yield erratic search trajectories and misalign the surrogate model with the true objective, thereby wasting the limited evaluation budget. To mitigate these issues, we propose a weighted Lagrangian estimator that leverages an adaptive schedule and importance sampling. This estimator dynamically balances exploration and exploitation by incorporating both the maximization of model performance and the minimization of the variance of model observations. It guides the search from broad, performance-seeking exploration towards stable and desirable regions as the optimization progresses. Extensive experiments and ablation studies, which establish our proposed method as the primary approach and other configurations as baselines, demonstrate its effectiveness across four backtest settings with three distinct black-box portfolio management models.
Authors: Hans Hohenfeld, Marius Beuerle, Elie Mounzer
Abstract: Tensor networks, a class of variational quantum many-body wave functions have attracted considerable research interest across many disciplines, including classical machine learning. Recently, Aizpurua et al. demonstrated explainable anomaly detection with matrix product states on a discrete-valued cyber-security task, using quantum-inspired methods to gain insight into the learned model and detected anomalies. Here, we extend this framework to real-valued data domains. We furthermore introduce tree tensor networks for the task of explainable anomaly detection. We demonstrate these methods with three benchmark problems, show adequate predictive performance compared to several baseline models and both tensor network architectures' ability to explain anomalous samples. We thereby extend the application of tensor networks to a broader class of potential problems and open a pathway for future extensions to more complex tensor network architectures.
Authors: Lang Feng, Zhenghai Xue, Tingcong Liu, Bo An
Abstract: Recent advances in group-based reinforcement learning (RL) have driven frontier large language models (LLMs) in single-turn tasks like mathematical reasoning. However, their scalability to long-horizon LLM agent training remains limited. Unlike static tasks, agent-environment interactions unfold over many steps and often yield sparse or delayed rewards, making credit assignment across individual steps significantly more challenging. In this work, we propose Group-in-Group Policy Optimization (GiGPO), a novel RL algorithm that achieves fine-grained credit assignment for LLM agents while preserving the appealing properties of group-based RL: critic-free, low memory, and stable convergence. GiGPO introduces a two-level structure for estimating relative advantage: (i) At the episode-level, GiGPO computes macro relative advantages based on groups of complete trajectories; (ii) At the step-level, GiGPO introduces an anchor state grouping mechanism that retroactively constructs step-level groups by identifying repeated environment states across trajectories. Actions stemming from the same state are grouped together, enabling micro relative advantage estimation. This hierarchical structure effectively captures both global trajectory quality and local step effectiveness without relying on auxiliary models or additional rollouts. We evaluate GiGPO on two challenging agent benchmarks, ALFWorld and WebShop, using Qwen2.5-1.5B-Instruct and Qwen2.5-7B-Instruct. Crucially, GiGPO delivers fine-grained per-step credit signals and achieves performance gains of > 12\% on ALFWorld and > 9\% on WebShop over the GRPO baseline: all while maintaining the same GPU memory overhead, identical LLM rollout, and incurring little to no additional time cost.
Authors: Juntian Zhu, Miguel de Carvalho, Zhouwang Yang, Fengxiang He
Abstract: An AI agent might surprisingly find she has reached an unknown state which she has never been aware of -- an unknown unknown. We mathematically ground this scenario in reinforcement learning: an agent, after taking an action calculated from value functions $Q$ and $V$ defined on the {\it {aware domain}}, reaches a state out of the domain. To enable the agent to handle this scenario, we propose an {\it episodic Markov decision {process} with growing awareness} (EMDP-GA) model, taking a new {\it noninformative value expansion} (NIVE) approach to expand value functions to newly aware areas: when an agent arrives at an unknown unknown, value functions $Q$ and $V$ whereon are initialised by noninformative beliefs -- the averaged values on the aware domain. This design is out of respect for the complete absence of knowledge in the newly discovered state. The upper confidence bound momentum Q-learning is then adapted to the growing awareness for training the EMDP-GA model. We prove that (1) the regret of our approach is asymptotically consistent with the state of the art (SOTA) without exposure to unknown unknowns in an extremely uncertain environment, and (2) our computational complexity and space complexity are comparable with the SOTA -- these collectively suggest that though an unknown unknown is surprising, it will be asymptotically properly discovered with decent speed and an affordable cost.
Authors: Devendra Parkar, Anya Chaturvedi, Joshua J. Daymude
Abstract: We present the first unsupervised learning model for finding Maximum Independent Sets (MaxIS) in dynamic graphs where edges change over time. Our method combines structural learning from graph neural networks (GNNs) with a learned distributed update mechanism that, given an edge addition or deletion event, modifies nodes' internal memories and infers their MaxIS membership in a single, parallel step. We parameterize our model by the update mechanism's radius and investigate the resulting performance-runtime tradeoffs for various dynamic graph topologies. We evaluate our model against a mixed integer programming solver and the state-of-the-art learning-based methods for MaxIS on static graphs (ICML 2020; NeurIPS 2020, 2023). Across synthetic and empirical dynamic graphs of 50-1,000 nodes, our model achieves competitive approximation ratios with excellent scalability; on large graphs, it significantly outperforms the state-of-the-art learning methods in solution quality, runtime, and memory usage. When generalizing to graphs of 10,000 nodes (100x larger than the ones used for training), our model produces MaxIS solutions 1.05-1.18x larger than any other learning method, even while maintaining competitive runtimes.
Authors: Dong Liu, Yanxuan Yu, Jiayi Zhang, Yifan Li, Ben Lengerich, Ying Nian Wu
Abstract: Diffusion Transformers (DiT) are powerful generative models but remain computationally intensive due to their iterative structure and deep transformer stacks. To alleviate this inefficiency, we propose FastCache, a hidden-state-level caching and compression framework that accelerates DiT inference by exploiting redundancy within the model's internal representations. FastCache introduces a dual strategy: (1) a spatial-aware token selection mechanism that adaptively filters redundant tokens based on hidden state saliency, and (2) a transformer-level cache that reuses latent activations across timesteps when changes are statistically insignificant. These modules work jointly to reduce unnecessary computation while preserving generation fidelity through learnable linear approximation. Theoretical analysis shows that FastCache maintains bounded approximation error under a hypothesis-testing-based decision rule. Empirical evaluations across multiple DiT variants demonstrate substantial reductions in latency and memory usage, with best generation output quality compared to other cache methods, as measured by FID and t-FID. Code implementation of FastCache is available on GitHub at https://github.com/NoakLiu/FastCache-xDiT.
Authors: Jun Tian, He Wang, Jibo He, Yu Pan, Shuo Cao, Qingquan Jiang
Abstract: Convolutional neural networks (CNNs) have become widely adopted in gravitational wave (GW) detection pipelines due to their ability to automatically learn hierarchical features from raw strain data. However, the physical meaning of these learned features remains underexplored, limiting the interpretability of such models. In this work, we propose a hybrid architecture that combines a CNN-based feature extractor with a random forest (RF) classifier to improve both detection performance and interpretability. Unlike prior approaches that directly connect classifiers to CNN outputs, our method introduces four physically interpretable metrics - variance, signal-to-noise ratio (SNR), waveform overlap, and peak amplitude - computed from the final convolutional layer. These are jointly used with the CNN output in the RF classifier to enable more informed decision boundaries. Tested on long-duration strain datasets, our hybrid model outperforms a baseline CNN model, achieving a relative improvement of 21\% in sensitivity at a fixed false alarm rate of 10 events per month. Notably, it also shows improved detection of low-SNR signals (SNR $\le$ 10), which are especially vulnerable to misclassification in noisy environments. Feature attribution via the RF model reveals that both CNN-extracted and handcrafted features contribute significantly to classification decisions, with learned variance and CNN outputs ranked among the most informative. These findings suggest that physically motivated post-processing of CNN feature maps can serve as a valuable tool for interpretable and efficient GW detection, bridging the gap between deep learning and domain knowledge.
Authors: Dhanesh Ramachandram, Ananya Raval
Abstract: Competing risks are crucial considerations in survival modelling, particularly in healthcare domains where patients may experience multiple distinct event types. We propose CRISP-NAM (Competing Risks Interpretable Survival Prediction with Neural Additive Models), an interpretable neural additive model for competing risks survival analysis which extends the neural additive architecture to model cause-specific hazards while preserving feature-level interpretability. Each feature contributes independently to risk estimation through dedicated neural networks, allowing for visualization of complex non-linear relationships between covariates and each competing risk. We demonstrate competitive performance on multiple datasets compared to existing approaches.
Authors: Minh Luu, Surya Jasper, Khoi Le, Evan Pan, Michael Quinn, Aakash Tyagi, Jiang Hu
Abstract: Failure triage in design functional verification is critical but time-intensive, relying on manual specification reviews, log inspections, and waveform analyses. While machine learning (ML) has improved areas like stimulus generation and coverage closure, its application to RTL-level simulation failure triage, particularly for large designs, remains limited. VCDiag offers an efficient, adaptable approach using VCD data to classify failing waveforms and pinpoint likely failure locations. In the largest experiment, VCDiag achieves over 94% accuracy in identifying the top three most likely modules. The framework introduces a novel signal selection and statistical compression approach, achieving over 120x reduction in raw data size while preserving features essential for classification. It can also be integrated into diverse Verilog/SystemVerilog designs and testbenches.
Authors: Jiajun He, Jos\'e Miguel Hern\'andez-Lobato, Yuanqi Du, Francisco Vargas
Abstract: Diffusion models generate data by removing noise gradually, which corresponds to the time-reversal of a noising process. However, access to only the denoising kernels is often insufficient. In many applications, we need the knowledge of the marginal densities along the generation trajectory, which enables tasks such as inference-time control. To address this gap, in this paper, we introduce the Radon-Nikodym Estimator (RNE). Based on the concept of the density ratio between path distributions, it reveals a fundamental connection between marginal densities and transition kernels, providing a flexible plug-and-play framework that unifies diffusion density estimation, inference-time control, and energy-based diffusion training under a single perspective. Experiments demonstrated that RNE delivers strong results in inference-time control applications, such as annealing and model composition, with promising inference-time scaling performance. Moreover, RNE provides a simple yet efficient regularisation for training energy-based diffusion.
Authors: Alexander Marusov, Aleksandr Yugay, Alexey Zaytsev
Abstract: Self-supervised learning (SSL) has emerged as a powerful approach to learning representations, particularly in the field of computer vision. However, its application to dependent data, such as temporal and spatio-temporal domains, remains underexplored. Besides, traditional contrastive SSL methods often assume \emph{semantic independence between samples}, which does not hold for dependent data exhibiting complex correlations. We propose a novel theoretical framework for contrastive SSL tailored to \emph{continuous dependent data}, which allows the nearest samples to be semantically close to each other. In particular, we propose two possible \textit{ground truth similarity measures} between objects -- \emph{hard} and \emph{soft} closeness. Under it, we derive an analytical form for the \textit{estimated similarity matrix} that accommodates both types of closeness between samples, thereby introducing dependency-aware loss functions. We validate our approach, \emph{Dependent TS2Vec}, on temporal and spatio-temporal downstream problems. Given the dependency patterns presented in the data, our approach surpasses modern ones for dependent data, highlighting the effectiveness of our theoretically grounded loss functions for SSL in capturing spatio-temporal dependencies. Specifically, we outperform TS2Vec on the standard UEA and UCR benchmarks, with accuracy improvements of $4.17$\% and $2.08$\%, respectively. Furthermore, on the drought classification task, which involves complex spatio-temporal patterns, our method achieves a $7$\% higher ROC-AUC score.
Authors: Zhiyuan Su, Sunhao Dai, Xiao Zhang
Abstract: Clustering of Bandits (CB) methods enhance sequential decision-making by grouping bandits into clusters based on similarity and incorporating cluster-level contextual information, demonstrating effectiveness and adaptability in applications like personalized streaming recommendations. However, when extending CB algorithms to their neural version (commonly referred to as Clustering of Neural Bandits, or CNB), they suffer from loss of plasticity, where neural network parameters become rigid and less adaptable over time, limiting their ability to adapt to non-stationary environments (e.g., dynamic user preferences in recommendation). To address this challenge, we propose Selective Reinitialization (SeRe), a novel bandit learning framework that dynamically preserves the adaptability of CNB algorithms in evolving environments. SeRe leverages a contribution utility metric to identify and selectively reset underutilized units, mitigating loss of plasticity while maintaining stable knowledge retention. Furthermore, when combining SeRe with CNB algorithms, the adaptive change detection mechanism adjusts the reinitialization frequency according to the degree of non-stationarity, ensuring effective adaptation without unnecessary resets. Theoretically, we prove that SeRe enables sublinear cumulative regret in piecewise-stationary environments, outperforming traditional CNB approaches in long-term performances. Extensive experiments on six real-world recommendation datasets demonstrate that SeRe-enhanced CNB algorithms can effectively mitigate the loss of plasticity with lower regrets, improving adaptability and robustness in dynamic settings.
Authors: Ronald Katende
Abstract: We present a unified theoretical framework for analyzing the stability and consistency of Physics-Informed Neural Networks (PINNs), grounded in operator coercivity, variational formulations, and non-asymptotic perturbation theory. PINNs approximate solutions to partial differential equations (PDEs) by minimizing residual losses over sampled collocation and boundary points. We formalize both operator-level and variational notions of consistency, proving that residual minimization in Sobolev norms leads to convergence in energy and uniform norms under mild regularity. Deterministic stability bounds quantify how bounded perturbations to the network outputs propagate through the full composite loss, while probabilistic concentration results via McDiarmid's inequality yield sample complexity guarantees for residual-based generalization. A unified generalization bound links residual consistency, projection error, and perturbation sensitivity. Empirical results on elliptic, parabolic, and nonlinear PDEs confirm the predictive accuracy of our theoretical bounds across regimes. The framework identifies key structural principles, such as operator coercivity, activation smoothness, and sampling admissibility, that underlie robust and generalizable PINN training, offering principled guidance for the design and analysis of PDE-informed learning systems.
Authors: Wenyu Luo, Yikai Hou, Peng Tang
Abstract: Modern intelligent transportation systems rely on accurate spatiotemporal traffic analysis to optimize urban mobility and infrastructure resilience. However, pervasive missing data caused by sensor failures and heterogeneous sensing gaps fundamentally hinders reliable traffic modeling. This paper proposes a Neural Canonical Polyadic Factorization (NCPF) model that synergizes low-rank tensor algebra with deep representation learning for robust traffic data imputation. The model innovatively embeds CP decomposition into neural architecture through learnable embedding projections, where sparse traffic tensors are encoded into dense latent factors across road segments, time intervals, and mobility metrics. A hierarchical feature fusion mechanism employs Hadamard products to explicitly model multilinear interactions, while stacked multilayer perceptron layers nonlinearly refine these representations to capture complex spatiotemporal couplings. Extensive evaluations on six urban traffic datasets demonstrate NCPF's superiority over six state-of-the-art baselines. By unifying CP decomposition's interpretable factor analysis with neural network's nonlinear expressive power, NCPF provides a principled yet flexible approaches for high-dimensional traffic data imputation, offering critical support for next-generation transportation digital twins and adaptive traffic control systems.
Authors: Gabor Petnehazi, Bernadett Aradi
Abstract: The explosive growth of complex datasets across various modalities necessitates advanced analytical tools that not only group data effectively but also provide human-understandable insights into the discovered structures. We introduce HERCULES (Hierarchical Embedding-based Recursive Clustering Using LLMs for Efficient Summarization), a novel algorithm and Python package designed for hierarchical k-means clustering of diverse data types, including text, images, and numeric data (processed one modality per run). HERCULES constructs a cluster hierarchy by recursively applying k-means clustering, starting from individual data points at level 0. A key innovation is its deep integration of Large Language Models (LLMs) to generate semantically rich titles and descriptions for clusters at each level of the hierarchy, significantly enhancing interpretability. The algorithm supports two main representation modes: `direct' mode, which clusters based on original data embeddings or scaled numeric features, and `description' mode, which clusters based on embeddings derived from LLM-generated summaries. Users can provide a `topic\_seed' to guide LLM-generated summaries towards specific themes. An interactive visualization tool facilitates thorough analysis and understanding of the clustering results. We demonstrate HERCULES's capabilities and discuss its potential for extracting meaningful, hierarchical knowledge from complex datasets.
Authors: Yiming Li, Shuo Shao, Yu He, Junfeng Guo, Tianwei Zhang, Zhan Qin, Pin-Yu Chen, Michael Backes, Philip Torr, Dacheng Tao, Kui Ren
Abstract: The (generative) artificial intelligence (AI) era has profoundly reshaped the meaning and value of data. No longer confined to static content, data now permeates every stage of the AI lifecycle from the training samples that shape model parameters to the prompts and outputs that drive real-world model deployment. This shift renders traditional notions of data protection insufficient, while the boundaries of what needs safeguarding remain poorly defined. Failing to safeguard data in AI systems can inflict societal and individual, underscoring the urgent need to clearly delineate the scope of and rigorously enforce data protection. In this perspective, we propose a four-level taxonomy, including non-usability, privacy preservation, traceability, and deletability, that captures the diverse protection needs arising in modern (generative) AI models and systems. Our framework offers a structured understanding of the trade-offs between data utility and control, spanning the entire AI pipeline, including training datasets, model weights, system prompts, and AI-generated content. We analyze representative technical approaches at each level and reveal regulatory blind spots that leave critical assets exposed. By offering a structured lens to align future AI technologies and governance with trustworthy data practices, we underscore the urgency of rethinking data protection for modern AI techniques and provide timely guidance for developers, researchers, and regulators alike.
Authors: Claudiu Leoveanu-Condrei
Abstract: Generative models, particularly Large Language Models (LLMs), produce fluent outputs yet lack verifiable guarantees. We adapt Design by Contract (DbC) and type-theoretic principles to introduce a contract layer that mediates every LLM call. Contracts stipulate semantic and type requirements on inputs and outputs, coupled with probabilistic remediation to steer generation toward compliance. The layer exposes the dual view of LLMs as semantic parsers and probabilistic black-box components. Contract satisfaction is probabilistic and semantic validation is operationally defined through programmer-specified conditions on well-typed data structures. More broadly, this work postulates that any two agents satisfying the same contracts are \emph{functionally equivalent} with respect to those contracts.
Authors: Xiang Li, Shanshan Wang, Chenglong Xiao
Abstract: Extensive experiments and prior studies show that no single maximum clique algorithm consistently performs best across all instances, highlighting the importance of selecting suitable algorithms based on instance features. Through an extensive analysis of relevant studies, it is found that there is a lack of research work concerning algorithm selection oriented toward the Maximum Clique Problem (MCP). In this work, we propose a learning-based framework that integrates both traditional machine learning and graph neural networks to address this gap. We construct a labeled dataset by running four exact MCP algorithms on a diverse collection of graph instances, accompanied by structural and global statistical features extracted from each graph. We first evaluate four conventional classifiers: Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), and K-Nearest Neighbors (KNN), across multiple dataset variants. Experimental results show that RF consistently shows strong performance across metrics and dataset variants, making it a reliable baseline. In addition, feature importance analysis indicates that connectivity and topological structure are strong predictors of algorithm performance. Building on these findings, we develop a dual-channel model named GAT-MLP, which combines a Graph Attention Network (GAT) for local structural encoding with a Multilayer Perceptron (MLP) for global feature modeling. The GAT-MLP model shows strong and consistent performance across all metrics. Our results highlight the effectiveness of dual-channel architectures and the promise of graph neural networks in combinatorial algorithm selection.
Authors: Maozhen Zhang, Mengnan Zhao, Bo Wang
Abstract: Prompt-based tuning has emerged as a lightweight alternative to full fine-tuning in large vision-language models, enabling efficient adaptation via learned contextual prompts. This paradigm has recently been extended to federated learning settings (e.g., PromptFL), where clients collaboratively train prompts under data privacy constraints. However, the security implications of prompt-based aggregation in federated multimodal learning remain largely unexplored, leaving a critical attack surface unaddressed. In this paper, we introduce \textbf{BadPromptFL}, the first backdoor attack targeting prompt-based federated learning in multimodal contrastive models. In BadPromptFL, compromised clients jointly optimize local backdoor triggers and prompt embeddings, injecting poisoned prompts into the global aggregation process. These prompts are then propagated to benign clients, enabling universal backdoor activation at inference without modifying model parameters. Leveraging the contextual learning behavior of CLIP-style architectures, BadPromptFL achieves high attack success rates (e.g., \(>90\%\)) with minimal visibility and limited client participation. Extensive experiments across multiple datasets and aggregation protocols validate the effectiveness, stealth, and generalizability of our attack, raising critical concerns about the robustness of prompt-based federated learning in real-world deployments.
Authors: Adarsh Jamadandi, Jing Xu, Adam Dziedzic, Franziska Boenisch
Abstract: Deep neural networks (DNNs) have been shown to memorize their training data, yet similar analyses for graph neural networks (GNNs) remain largely under-explored. We introduce NCMemo (Node Classification Memorization), the first framework to quantify label memorization in semi-supervised node classification. We first establish an inverse relationship between memorization and graph homophily, i.e., the property that connected nodes share similar labels/features. We find that lower homophily significantly increases memorization, indicating that GNNs rely on memorization to learn less homophilic graphs. Secondly, we analyze GNN training dynamics. We find that the increased memorization in low homophily graphs is tightly coupled to the GNNs' implicit bias on using graph structure during learning. In low homophily regimes, this structure is less informative, hence inducing memorization of the node labels to minimize training loss. Finally, we show that nodes with higher label inconsistency in their feature-space neighborhood are significantly more prone to memorization. Building on our insights into the link between graph homophily and memorization, we investigate graph rewiring as a means to mitigate memorization. Our results demonstrate that this approach effectively reduces memorization without compromising model performance. Moreover, we show that it lowers the privacy risk for previously memorized data points in practice. Thus, our work not only advances understanding of GNN learning but also supports more privacy-preserving GNN deployment.
Authors: Amit Sinha, Matthieu Geist, Aditya Mahajan
Abstract: In this paper, we present a framework to understand the convergence of commonly used Q-learning reinforcement learning algorithms in practice. Two salient features of such algorithms are: (i)~the Q-table is recursively updated using an agent state (such as the state of a recurrent neural network) which is not a belief state or an information state and (ii)~policy regularization is often used to encourage exploration and stabilize the learning algorithm. We investigate the simplest form of such Q-learning algorithms which we call regularized agent-state-based Q-learning (RASQL) and show that it converges under mild technical conditions to the fixed point of an appropriately defined regularized MDP, which depends on the stationary distribution induced by the behavioral policy. We also show that a similar analysis continues to work for a variant of RASQL that learns periodic policies. We present numerical examples to illustrate that the empirical convergence behavior matches with the proposed theoretical limit.
Authors: Yao Fu, Runchao Li, Xianxuan Long, Haotian Yu, Xiaotian Han, Yu Yin, Pan Li
Abstract: Neural network pruning has emerged as a promising approach for deploying LLMs in low-resource scenarios while preserving downstream task performance. However, for the first time, we reveal that such pruning disrupts LLMs' internal activation features crucial for lie detection, where probing classifiers (typically small logistic regression models) trained on these features assess the truthfulness of LLM-generated statements. This discovery raises a crucial open question: how can we prune LLMs without sacrificing these critical lie detection capabilities? Our investigation further reveals that naively adjusting layer-wise pruning sparsity based on importance inadvertently removes crucial weights, failing to improve lie detection performance despite its reliance on the most crucial LLM layer. To address this issue, we propose Truthful Pruning aligned by Layer-wise Outliers (TPLO), which places greater emphasis on layers with more activation outliers and stronger discriminative features simultaneously. This preserves LLMs' original performance while retaining critical features of inner states needed for robust lie detection. Moreover, we introduce a prompting rule to enrich the TruthfulQA benchmark for better calibrating LLM pruning. Empirical results show that our approach improves the hallucination detection for pruned LLMs (achieving 88% accuracy at 50% sparsity) and enhances their performance on TruthfulQA.
Authors: Xiao Xue, M. F. P. ten Eikelder, Tianyue Yang, Yiqing Li, Kan He, Shuo Wang, Peter V. Coveney
Abstract: Phase separation in binary mixtures, governed by the Cahn-Hilliard equation, plays a central role in interfacial dynamics across materials science and soft matter. While numerical solvers are accurate, they are often computationally expensive and lack flexibility across varying initial conditions and geometries. Neural operators provide a data-driven alternative by learning solution operators between function spaces, but current architectures often fail to capture multiscale behavior and neglect underlying physical symmetries. Here we show that an equivariant U-shaped neural operator (E-UNO) can learn the evolution of the phase-field variable from short histories of past dynamics, achieving accurate predictions across space and time. The model combines global spectral convolution with a multi-resolution U-shaped architecture and regulates translation equivariance to align with the underlying physics. E-UNO outperforms standard Fourier neural operator and U-shaped neural operator baselines, particularly on fine-scale and high-frequency structures. By encoding symmetry and scale hierarchy, the model generalizes better, requires less training data, and yields physically consistent dynamics. This establishes E-UNO as an efficient surrogate for complex phase-field systems.
Authors: Shijie Wang, Li Zhang, Xinyan Liang, Yuhua Qian, Shen Hu
Abstract: Multimodal learning typically utilizes multimodal joint loss to integrate different modalities and enhance model performance. However, this joint learning strategy can induce modality imbalance, where strong modalities overwhelm weaker ones and limit exploitation of individual information from each modality and the inter-modality interaction information. Existing strategies such as dynamic loss weighting, auxiliary objectives and gradient modulation mitigate modality imbalance based on joint loss. These methods remain fundamentally reactive, detecting and correcting imbalance after it arises, while leaving the competitive nature of the joint loss untouched. This limitation drives us to explore a new strategy for multimodal imbalance learning that does not rely on the joint loss, enabling more effective interactions between modalities and better utilization of information from individual modalities and their interactions. In this paper, we introduce Unidirectional Dynamic Interaction (UDI), a novel strategy that abandons the conventional joint loss in favor of a proactive, sequential training scheme. UDI first trains the anchor modality to convergence, then uses its learned representations to guide the other modality via unsupervised loss. Furthermore, the dynamic adjustment of modality interactions allows the model to adapt to the task at hand, ensuring that each modality contributes optimally. By decoupling modality optimization and enabling directed information flow, UDI prevents domination by any single modality and fosters effective cross-modal feature learning. Our experimental results demonstrate that UDI outperforms existing methods in handling modality imbalance, leading to performance improvement in multimodal learning tasks.
Authors: Zhenghai Xue, Longtao Zheng, Qian Liu, Yingru Li, Xiaosen Zheng, Zejun Ma, Bo An
Abstract: Large Language Models (LLMs) can significantly improve their reasoning capabilities by interacting with external tools, a paradigm known as Tool-Integrated Reasoning (TIR). However, extending TIR to multi-turn scenarios using Reinforcement Learning (RL) is often hindered by training instability and performance collapse. We identify that such instability is primarily caused by a distributional drift from external tool feedback, leading to the generation of low-probability tokens. This issue compounds over successive turns, causing catastrophic gradient norm explosions that derail the training process. To address this challenge, we introduce SimpleTIR , a plug-and-play algorithm that stabilizes multi-turn TIR training. Its core strategy is to identify and filter out trajectories containing void turns, i.e., turns that yield neither a code block nor a final answer. By removing these problematic trajectories from the policy update, SimpleTIR effectively blocks the harmful, high-magnitude gradients, thus stabilizing the learning dynamics. Extensive experiments show that SimpleTIR achieves state-of-the-art performance on challenging math reasoning benchmarks, notably elevating the AIME24 score from a text-only baseline of 22.1 to 50.5 when starting from the Qwen2.5-7B base model. Furthermore, by avoiding the constraints of supervised fine-tuning, SimpleTIR encourages the model to discover diverse and sophisticated reasoning patterns, such as self-correction and cross-validation.
Authors: Xu Yuan, Chen Xu, Qiwei Chen, Chao Li, Junfeng Ge, Wenwu Ou
Abstract: In this era of information explosion, a personalized recommendation system is convenient for users to get information they are interested in. To deal with billions of users and items, large-scale online recommendation services usually consist of three stages: candidate generation, coarse-grained ranking, and fine-grained ranking. The success of each stage depends on whether the model accurately captures the interests of users, which are usually hidden in users' behavior data. Previous research shows that users' interests are diverse, and one vector is not sufficient to capture users' different preferences. Therefore, many methods use multiple vectors to encode users' interests. However, there are two unsolved problems: (1) The similarity of different vectors in existing methods is too high, with too much redundant information. Consequently, the interests of users are not fully represented. (2) Existing methods model the long-term and short-term behaviors together, ignoring the differences between them. This paper proposes a Hierarchical Multi-Interest Co-Network (HCN) to capture users' diverse interests in the coarse-grained ranking stage. Specifically, we design a hierarchical multi-interest extraction layer to update users' diverse interest centers iteratively. The multiple embedded vectors obtained in this way contain more information and represent the interests of users better in various aspects. Furthermore, we develop a Co-Interest Network to integrate users' long-term and short-term interests. Experiments on several real-world datasets and one large-scale industrial dataset show that HCN effectively outperforms the state-of-the-art methods. We deploy HCN into a large-scale real world E-commerce system and achieve extra 2.5\% improvements on GMV (Gross Merchandise Value).
Authors: Guillaume Wang, L\'ena\"ic Chizat
Abstract: We consider the problem of computing mixed Nash equilibria of two-player zero-sum games with continuous sets of pure strategies and with first-order access to the payoff function. This problem arises for example in game-theory-inspired machine learning applications, such as distributionally-robust learning. In those applications, the strategy sets are high-dimensional and thus methods based on discretisation cannot tractably return high-accuracy solutions. In this paper, we introduce and analyze a particle-based method that enjoys guaranteed local convergence for this problem. This method consists in parametrizing the mixed strategies as atomic measures and applying proximal point updates to both the atoms' weights and positions. It can be interpreted as a time-implicit discretization of the "interacting" Wasserstein-Fisher-Rao gradient flow. We prove that, under non-degeneracy assumptions, this method converges at an exponential rate to the exact mixed Nash equilibrium from any initialization satisfying a natural notion of closeness to optimality. We illustrate our results with numerical experiments and discuss applications to max-margin and distributionally-robust classification using two-layer neural networks, where our method has a natural interpretation as a simultaneous training of the network's weights and of the adversarial distribution.
Authors: Anran Hu, Junzi Zhang
Abstract: Reinforcement learning for multi-agent games has attracted lots of attention recently. However, given the challenge of solving Nash equilibria for large population games, existing works with guaranteed polynomial complexities either focus on variants of zero-sum and potential games, or aim at solving (coarse) correlated equilibria, or require access to simulators, or rely on certain assumptions that are hard to verify. This work proposes MF-OML (Mean-Field Occupation-Measure Learning), an online mean-field reinforcement learning algorithm for computing approximate Nash equilibria of large population sequential symmetric games. MF-OML is the first fully polynomial multi-agent reinforcement learning algorithm for provably solving Nash equilibria (up to mean-field approximation gaps that vanish as the number of players $N$ goes to infinity) beyond variants of zero-sum and potential games. When evaluated by the cumulative deviation from Nash equilibria, the algorithm is shown to achieve a high probability regret bound of $\tilde{O}(M^{3/4}+N^{-1/2}M)$ for games with the strong Lasry-Lions monotonicity condition, and a regret bound of $\tilde{O}(M^{11/12}+N^{- 1/6}M)$ for games with only the Lasry-Lions monotonicity condition, where $M$ is the total number of episodes and $N$ is the number of agents of the game. As a byproduct, we also obtain the first tractable globally convergent computational algorithm for computing approximate Nash equilibria of monotone mean-field games.
Authors: Andra\v{z} Jelin\v{c}i\v{c}, James Foster, Patrick Kidger
Abstract: Despite the success of adaptive time-stepping in ODE simulation, it has so far seen few applications for Stochastic Differential Equations (SDEs). To simulate SDEs adaptively, methods such as the Virtual Brownian Tree (VBT) have been developed, which can generate Brownian motion (BM) non-chronologically. However, in most applications, knowing only the values of Brownian motion is not enough to achieve a high order of convergence; for that, we must compute time-integrals of BM such as $\int_s^t W_r \, dr$. With the aim of using high order SDE solvers adaptively, we extend the VBT to generate these integrals of BM in addition to the Brownian increments. A JAX-based implementation of our construction is included in the popular Diffrax library (https://github.com/patrick-kidger/diffrax). Since the entire Brownian path produced by VBT is uniquely determined by a single PRNG seed, previously generated samples need not be stored, which results in a constant memory footprint and enables experiment repeatability and strong error estimation. Based on binary search, the VBT's time complexity is logarithmic in the tolerance parameter $\varepsilon$. Unlike the original VBT algorithm, which was only precise at some dyadic times, we prove that our construction exactly matches the joint distribution of the Brownian motion and its time integrals at any query times, provided they are at least $\varepsilon$ apart. We present two applications of adaptive high order solvers enabled by our new VBT. Using adaptive solvers to simulate a high-volatility CIR model, we achieve more than twice the convergence order of constant stepping. We apply an adaptive third order underdamped or kinetic Langevin solver to an MCMC problem, where our approach outperforms the No U-Turn Sampler, while using only a tenth of its function evaluations.
Authors: Taiming Lu, Philipp Koehn
Abstract: This paper investigates the propagation of harmful information in multilingual large language models (LLMs) and evaluates the efficacy of various unlearning methods. We demonstrate that fake information, regardless of the language it is in, once introduced into these models through training data, can spread across different languages, compromising the integrity and reliability of the generated content. Our findings reveal that standard unlearning techniques, which typically focus on English data, are insufficient in mitigating the spread of harmful content in multilingual contexts and could inadvertently reinforce harmful content across languages. We show that only by addressing harmful responses in both English and the original language of the harmful data can we effectively eliminate generations for all languages. This underscores the critical need for comprehensive unlearning strategies that consider the multilingual nature of modern LLMs to enhance their safety and reliability across diverse linguistic landscapes.
Authors: Qianchao Zhu, Jiangfei Duan, Chang Chen, Siran Liu, Guanyu Feng, Xin Lv, Xiao Chuanfu, Dahua Lin, Chao Yang
Abstract: Large language models (LLMs) now support extremely long context windows, but the quadratic complexity of vanilla attention results in significantly long Time-to-First-Token (TTFT) latency. Existing approaches to address this complexity require additional pretraining or finetuning, and often sacrifice model accuracy. In this paper, we first provide both theoretical and empirical foundations for near-lossless sparse attention. We find dynamically capturing head-specific sparse patterns at runtime with low overhead is crucial. To address this, we propose SampleAttention, an adaptive structured and near-lossless sparse attention. Leveraging observed significant sparse patterns, SampleAttention attends to a fixed percentage of adjacent tokens to capture local window patterns, and employs a two-stage query-guided key-value filtering approach, which adaptively select a minimum set of key-values with low overhead, to capture column stripe patterns. Comprehensive evaluations show that SampleAttention can seamlessly replace vanilla attention in off-the-shelf LLMs with nearly no accuracy loss, and reduces TTFT by up to $2.42\times$ compared with FlashAttention.
Authors: Lukas Muttenthaler, Klaus Greff, Frieda Born, Bernhard Spitzer, Simon Kornblith, Michael C. Mozer, Klaus-Robert M\"uller, Thomas Unterthiner, Andrew K. Lampinen
Abstract: Deep neural networks have achieved success across a wide range of applications, including as models of human behavior and neural representations in vision tasks. However, neural network training and human learning differ in fundamental ways, and neural networks often fail to generalize as robustly as humans do raising questions regarding the similarity of their underlying representations. What is missing for modern learning systems to exhibit more human-aligned behavior? We highlight a key misalignment between vision models and humans: whereas human conceptual knowledge is hierarchically organized from fine- to coarse-scale distinctions, model representations do not accurately capture all these levels of abstraction. To address this misalignment, we first train a teacher model to imitate human judgments, then transfer human-aligned structure from its representations to refine the representations of pretrained state-of-the-art vision foundation models via finetuning. These human-aligned models more accurately approximate human behavior and uncertainty across a wide range of similarity tasks, including a new dataset of human judgments spanning multiple levels of semantic abstractions. They also perform better on a diverse set of machine learning tasks, increasing generalization and out-of-distribution robustness. Thus, infusing neural networks with additional human knowledge yields a best-of-both-worlds representation that is both more consistent with human cognitive judgments and more practically useful, thus paving the way toward more robust, interpretable, and human-aligned artificial intelligence systems.
Authors: Han Zhou, Jordy Van Landeghem, Teodora Popordanoska, Matthew B. Blaschko
Abstract: The selective classifier (SC) has been proposed for rank based uncertainty thresholding, which could have applications in safety critical areas such as medical diagnostics, autonomous driving, and the justice system. The Area Under the Risk-Coverage Curve (AURC) has emerged as the foremost evaluation metric for assessing the performance of SC systems. In this work, we present a formal statistical formulation of population AURC, presenting an equivalent expression that can be interpreted as a reweighted risk function. Through Monte Carlo methods, we derive empirical AURC plug-in estimators for finite sample scenarios. The weight estimators associated with these plug-in estimators are shown to be consistent, with low bias and tightly bounded mean squared error (MSE). The plug-in estimators are proven to converge at a rate of $\mathcal{O}(\sqrt{\ln(n)/n})$ demonstrating statistical consistency. We empirically validate the effectiveness of our estimators through experiments across multiple datasets, model architectures, and confidence score functions (CSFs), demonstrating consistency and effectiveness in fine-tuning AURC performance.
Authors: Jiahao Qiu, Yifu Lu, Yifan Zeng, Jiacheng Guo, Jiayi Geng, Chenhao Zhu, Xinzhe Juan, Ling Yang, Huazheng Wang, Kaixuan Huang, Yue Wu, Mengdi Wang
Abstract: Inference-time alignment enhances the performance of large language models without requiring additional training or fine-tuning but presents challenges due to balancing computational efficiency with high-quality output. Best-of-N (BoN) sampling, as a simple yet powerful approach, generates multiple responses and selects the best one, achieving improved performance but with a high computational cost. We propose TreeBoN, a novel framework that integrates a speculative tree-search strategy into Best-of-N (BoN) Sampling. TreeBoN maintains a set of parent nodes, iteratively branching and pruning low-quality responses, thereby reducing computational overhead while maintaining high output quality. Our approach also leverages token-level rewards from Direct Preference Optimization (DPO) to guide tree expansion and prune low-quality paths. We evaluate TreeBoN using AlpacaFarm, HH-RLHF, UltraFeedback, GSM8K, and TutorEval datasets, demonstrating consistent improvements. Specifically, TreeBoN achieves the highest win rate of 65% on TutorEval and around 60% win rates across other different datasets, outperforming standard BoN with the same computational cost and showcasing its scalability and alignment efficacy.
Authors: Johann Brehmer, V\'ictor Bres\'o, Pim de Haan, Tilman Plehn, Huilin Qu, Jonas Spinner, Jesse Thaler
Abstract: We show that the Lorentz-Equivariant Geometric Algebra Transformer (L-GATr) yields state-of-the-art performance for a wide range of machine learning tasks at the Large Hadron Collider. L-GATr represents data in a geometric algebra over space-time and is equivariant under Lorentz transformations. The underlying architecture is a versatile and scalable transformer, which is able to break symmetries if needed. We demonstrate the power of L-GATr for amplitude regression and jet classification, and then benchmark it as the first Lorentz-equivariant generative network. For all three LHC tasks, we find significant improvements over previous architectures.
Authors: Ruoqi Wang, Haitao Wang, Qiong Luo
Abstract: Galaxy morphology analysis involves studying galaxies based on their shapes and structures. For such studies, fundamental tasks include identifying and classifying galaxies in astronomical images, as well as retrieving visually or structurally similar galaxies through similarity search. Existing methods either directly train domain-specific foundation models on large, annotated datasets or fine-tune vision foundation models on a smaller set of images. The former is effective but costly, while the latter is more resource-efficient but often yields lower accuracy. To address these challenges, we introduce GalaxAlign, a multimodal approach inspired by how citizen scientists identify galaxies in astronomical images by following textual descriptions and matching schematic symbols. Specifically, GalaxAlign employs a tri-modal alignment framework to align three types of data during fine-tuning: (1) schematic symbols representing galaxy shapes and structures, (2) textual labels for these symbols, and (3) galaxy images. By incorporating multimodal instructions, GalaxAlign eliminates the need for expensive pretraining and enhances the effectiveness of fine-tuning. Experiments on galaxy classification and similarity search demonstrate that our method effectively fine-tunes general pre-trained models for astronomical tasks by incorporating domain-specific multi-modal knowledge. Code is available at https://github.com/RapidsAtHKUST/GalaxAlign.
Authors: Xiaoyan Xing, Konrad Groh, Sezer Karaoglu, Theo Gevers, Anand Bhattad
Abstract: We introduce LumiNet, a novel architecture that leverages generative models and latent intrinsic representations for effective lighting transfer. Given a source image and a target lighting image, LumiNet synthesizes a relit version of the source scene that captures the target's lighting. Our approach makes two key contributions: a data curation strategy from the StyleGAN-based relighting model for our training, and a modified diffusion-based ControlNet that processes both latent intrinsic properties from the source image and latent extrinsic properties from the target image. We further improve lighting transfer through a learned adaptor (MLP) that injects the target's latent extrinsic properties via cross-attention and fine-tuning. Unlike traditional ControlNet, which generates images with conditional maps from a single scene, LumiNet processes latent representations from two different images - preserving geometry and albedo from the source while transferring lighting characteristics from the target. Experiments demonstrate that our method successfully transfers complex lighting phenomena including specular highlights and indirect illumination across scenes with varying spatial layouts and materials, outperforming existing approaches on challenging indoor scenes using only images as input.
Authors: Mengze Hong, Wailing Ng, Chen Jason Zhang, Yuanfeng Song, Di Jiang
Abstract: Discovering customer intentions is crucial for automated service agents, yet existing intent clustering methods often fall short due to their reliance on embedding distance metrics and neglect of underlying semantic structures. To address these limitations, we propose an LLM-in-the-loop (LLM-ITL) intent clustering framework, integrating the language understanding capabilities of LLMs into conventional clustering algorithms. Specifically, this paper (1) examines the effectiveness of fine-tuned LLMs in semantic coherence evaluation and intent cluster naming, achieving over 95% accuracy aligned with human judgments; (2) designs an LLM-ITL framework that facilitates the iterative discovery of coherent intent clusters and the optimal number of clusters; and (3) introduces context-aware techniques tailored for customer service dialogue. Since existing English benchmarks lack sufficient semantic diversity and intent coverage, we further present a comprehensive Chinese dialogue intent dataset comprising over 100k real customer service calls with 1,507 human-annotated clusters. The proposed approaches significantly outperform LLM-guided baselines, achieving notable improvements in clustering quality, cost efficiency, and downstream applications. Combined with several best practices, our findings highlight the prominence of LLM-in-the-loop techniques for scalable dialogue data mining.
Authors: Carlos G\"uemes-Palau, Miquel Ferriol-Galm\'es, Jordi Paillisse-Vilanova, Albert L\'opez-Bresc\'o, Pere Barlet-Ros, Albert Cabellos-Aparicio
Abstract: Network simulation is pivotal in network modeling, assisting with tasks ranging from capacity planning to performance estimation. Traditional approaches such as Discrete Event Simulation (DES) face limitations in terms of computational cost and accuracy. This paper introduces RouteNet-Gauss, a novel integration of a testbed network with a Machine Learning (ML) model to address these challenges. By using the testbed as a hardware accelerator, RouteNet-Gauss generates training datasets rapidly and simulates network scenarios with high fidelity to real-world conditions. Experimental results show that RouteNet-Gauss significantly reduces prediction errors by up to 95% and achieves a 488x speedup in inference time compared to state-of-the-art DES-based methods. RouteNet-Gauss's modular architecture is dynamically constructed based on the specific characteristics of the network scenario, such as topology and routing. This enables it to understand and generalize to different network configurations beyond those seen during training, including networks up to 10x larger. Additionally, it supports Temporal Aggregated Performance Estimation (TAPE), providing configurable temporal granularity and maintaining high accuracy in flow performance metrics. This approach shows promise in improving both simulation efficiency and accuracy, offering a valuable tool for network operators.
Authors: Bhavna Bose, Saurav Verma
Abstract: The development of quantum computers has been the stimulus that enables the realization of Quantum Machine Learning (QML), an area that integrates the calculational framework of quantum mechanics with the adaptive properties of classical machine learning. This article suggests a broad architecture that allows the connection between classical data pipelines and quantum algorithms, hybrid quantum-classical models emerge as a promising route to scalable and near-term quantum benefit. At the core of this paradigm lies the Classical-Quantum (CQ) paradigm, in which the qubit states of high-dimensional classical data are encoded using sophisticated classical encoding strategies which encode the data in terms of amplitude and angle of rotation, along with superposition mapping. These techniques allow compression of information exponentially into Hilbert space representations, which, together with reduced sample complexity, allows greater feature expressivity. We also examine variational quantum circuits, quantum gates expressed as trainable variables that run with classical optimizers to overcome decoherence, noise, and gate-depth constraints of the existing Noisy Intermediate-Scale Quantum (NISQ) devices. Experimental comparisons with a Quantum Naive Bayes classifier prove that even small quantum circuits can approximate probabilistic inference with competitive accuracy compared to classical benchmarks, and have much better robustness to noisy data distributionsThis model does not only explain the algorithmic and architectural design of QML, it also offers a roadmap to the implementation of quantum kernels, variational algorithms, and hybrid feedback loops into practice, including optimization, computer vision, and medical diagnostics. The results support the idea that hybrid architectures with strong data encoding and adaptive error protection are key to moving QML out of theory to practice.
Authors: Wentao Wang, Guangyuan Jiang, Tal Linzen, Brenden M. Lake
Abstract: Humans can quickly learn a new word from a few illustrative examples, and then systematically and flexibly use it in novel contexts. Yet the abilities of current language models for few-shot word learning, and methods for improving these abilities, are underexplored. In this study, we introduce a novel method, Meta-training for IN-context learNing Of Words (Minnow). This method trains language models to generate new examples of a word's usage given a few in-context examples, using a special placeholder token to represent the new word. This training is repeated on many new words to develop a general word-learning ability. We find that training models from scratch with Minnow on human-scale child-directed language enables strong few-shot word learning, comparable to a large language model (LLM) pre-trained on orders of magnitude more data. Furthermore, through discriminative and generative evaluations, we demonstrate that finetuning pre-trained LLMs with Minnow improves their ability to discriminate between new words, identify syntactic categories of new words, and generate reasonable new usages and definitions for new words, based on one or a few in-context examples. These findings highlight the data efficiency of Minnow and its potential to improve language model performance in word learning tasks.
Authors: Namiko Matsumoto, Arya Mazumdar
Abstract: In statistics, generalized linear models (GLMs) are widely used for modeling data and can expressively capture potential nonlinear dependence of the model's outcomes on its covariates. Within the broad family of GLMs, those with binary outcomes, which include logistic and probit regressions, are motivated by common tasks such as binary classification with (possibly) non-separable data. In addition, in modern machine learning and statistics, data is often high-dimensional yet has a low intrinsic dimension, making sparsity constraints in models another reasonable consideration. In this work, we propose to use and analyze an iterative hard thresholding (projected gradient descent on the ReLU loss) algorithm, called binary iterative hard thresholding (BIHT), for parameter estimation in sparse GLMs with binary outcomes. We establish that BIHT is statistically efficient and converges to the correct solution for parameter estimation in a general class of sparse binary GLMs. Unlike many other methods for learning GLMs, including maximum likelihood estimation, generalized approximate message passing, and GLM-tron (Kakade et al. 2011; Bahmani et al. 2016), BIHT does not require knowledge of the GLM's link function, offering flexibility and generality in allowing the algorithm to learn arbitrary binary GLMs. As two applications, logistic and probit regression are additionally studied. In this regard, it is shown that in logistic regression, the algorithm is in fact statistically optimal in the sense that the order-wise sample complexity matches (up to logarithmic factors) the lower bound obtained previously. To the best of our knowledge, this is the first work achieving statistical optimality for logistic regression in all noise regimes with a computationally efficient algorithm. Moreover, for probit regression, our sample complexity is on the same order as that obtained for logistic regression.
Authors: Andrea Montanari, Pierfrancesco Urbani
Abstract: Understanding the inductive bias and generalization properties of large overparametrized machine learning models requires to characterize the dynamics of the training algorithm. We study the learning dynamics of large two-layer neural networks via dynamical mean field theory, a well established technique of non-equilibrium statistical physics. We show that, for large network width, the training dynamics exhibits a separation of timescales which implies: $(i)$ The emergence of a slow time scale associated with the growth in Gaussian/Rademacher complexity of the network; $(ii)$ Inductive bias towards small complexity if the initialization has small enough complexity; $(iii)$ A dynamical decoupling between feature learning and overfitting regimes; $(iv)$ A non-monotone behavior of the test error, associated `feature unlearning' regime at large times.
Authors: Irched Chafaa, Giacomo Bacci, Luca Sanguinetti
Abstract: Power allocation is an important task in wireless communication networks. Classical optimization algorithms and deep learning methods, while effective in small and static scenarios, become either computationally demanding or unsuitable for large and dynamic networks with varying user loads. This letter explores the potential of transformer-based deep learning models to address these challenges. We propose a transformer neural network to jointly predict optimal uplink and downlink power using only user and access point positions. The max-min fairness problem in cell-free massive multiple input multiple output systems is considered. Numerical results show that the trained model provides near-optimal performance and adapts to varying numbers of users and access points without retraining, additional processing, or updating its neural network architecture. This demonstrates the effectiveness of the proposed model in achieving robust and flexible power allocation for dynamic networks.
Authors: Alessio Spagnoletti, Jean Prost, Andr\'es Almansa, Nicolas Papadakis, Marcelo Pereyra
Abstract: Text-to-image latent diffusion models (LDMs) have recently emerged as powerful generative models with great potential for solving inverse problems in imaging. However, leveraging such models in a Plug & Play (PnP), zero-shot manner remains challenging because it requires identifying a suitable text prompt for the unknown image of interest. Also, existing text-to-image PnP approaches are highly computationally expensive. We herein address these challenges by proposing a novel PnP inference paradigm specifically designed for embedding generative models within stochastic inverse solvers, with special attention to Latent Consistency Models (LCMs), which distill LDMs into fast generators. We leverage our framework to propose LAtent consisTency INverse sOlver (LATINO), the first zero-shot PnP framework to solve inverse problems with priors encoded by LCMs. Our conditioning mechanism avoids automatic differentiation and reaches SOTA quality in as little as 8 neural function evaluations. As a result, LATINO delivers remarkably accurate solutions and is significantly more memory and computationally efficient than previous approaches. We then embed LATINO within an empirical Bayesian framework that automatically calibrates the text prompt from the observed measurements by marginal maximum likelihood estimation. Extensive experiments show that prompt self-calibration greatly improves estimation, allowing LATINO with PRompt Optimization to define new SOTAs in image reconstruction quality and computational efficiency. The code is available at https://latino-pro.github.io
Authors: Ron Campos, Ashmal Vayani, Parth Parag Kulkarni, Rohit Gupta, Aizan Zafar, Aritra Dutta, Mubarak Shah
Abstract: Image geolocalization, in which an AI model traditionally predicts the precise GPS coordinates of an image, is a challenging task with many downstream applications. However, the user cannot utilize the model to further their knowledge beyond the GPS coordinates; the model lacks an understanding of the location and the conversational ability to communicate with the user. In recent days, with the tremendous progress of large multimodal models (LMMs) -- proprietary and open-source -- researchers have attempted to geolocalize images via LMMs. However, the issues remain unaddressed; beyond general tasks, for more specialized downstream tasks, such as geolocalization, LMMs struggle. In this work, we propose solving this problem by introducing a conversational model, GAEA, that provides information regarding the location of an image as the user requires. No large-scale dataset enabling the training of such a model exists. Thus, we propose GAEA-1.4M, a comprehensive dataset comprising over 800k images and approximately 1.4M question-answer pairs, constructed by leveraging OpenStreetMap (OSM) attributes and geographical context clues. For quantitative evaluation, we propose a diverse benchmark, GAEA-Bench, comprising 3.5k image-text pairs to evaluate conversational capabilities equipped with diverse question types. We consider 11 state-of-the-art open-source and proprietary LMMs and demonstrate that GAEA significantly outperforms the best open-source model, LLaVA-OneVision, by 18.2% and the best proprietary model, GPT-4o, by 7.2%. Our dataset, model and codes are available.
Authors: Davide Piras, Francesco Sorrenti, Ruth Durrer, Martin Kunz
Abstract: We develop a novel approach to constrain the Hubble parameter $H_0$ and the primordial power spectrum amplitude $A_\mathrm{s}$ using type Ia supernovae (SNIa) data. By considering SNIa as tracers of the peculiar velocity field, we can model their distance and their covariance as a function of cosmological parameters without the need of calibrators like Cepheids; this yields a new independent probe of the large-scale structure based on SNIa data without distance anchors. Crucially, we implement a differentiable pipeline in JAX, including efficient emulators and affine sampling, reducing inference time from years to hours on a single GPU. We first validate our method on mock datasets, demonstrating that we can constrain $H_0$ and $\log 10^{10}A_\mathrm{s}$ within $10\%$ and $15\%$, respectively, using $\mathcal{O}(10^3)$ SNIa. We then test our pipeline with SNIa from an $N$-body simulation, obtaining $6\%$-level unbiased constraints on $H_0$ with a moderate noise level. We finally apply our method to Pantheon+ data, constraining $H_0$ at the $15\%$ level without Cepheids when fixing $A_\mathrm{s}$ to its $\it{Planck}$ value. On the other hand, we obtain $20\%$-level constraints on $\log 10^{10}A_\mathrm{s}$ in agreement with $\it{Planck}$ when including Cepheids in the analysis. In light of upcoming observations of low redshift SNIa from the Zwicky Transient Facility and the Vera Rubin Legacy Survey of Space and Time, surveys for which our method will develop its full potential, we make our code publicly available.
Authors: Debarati Das, Khanh Chi Le, Ritik Sachin Parkar, Karin De Langis, Brendan Madson, Chad M. Berryman, Robin M. Willis, Daniel H. Moses, Brett McDonnell, Daniel Schwarcz, Dongyeop Kang
Abstract: Legal practitioners, particularly those early in their careers, face complex, high-stakes tasks that require adaptive, context-sensitive reasoning. While AI holds promise in supporting legal work, current datasets and models are narrowly focused on isolated subtasks and fail to capture the end-to-end decision-making required in real-world practice. To address this gap, we introduce LawFlow, a dataset of complete end-to-end legal workflows collected from trained law students, grounded in real-world business entity formation scenarios. Unlike prior datasets focused on input-output pairs or linear chains of thought, LawFlow captures dynamic, modular, and iterative reasoning processes that reflect the ambiguity, revision, and client-adaptive strategies of legal practice. Using LawFlow, we compare human and LLM-generated workflows, revealing systematic differences in structure, reasoning flexibility, and plan execution. Human workflows tend to be modular and adaptive, while LLM workflows are more sequential, exhaustive, and less sensitive to downstream implications. Our findings also suggest that legal professionals prefer AI to carry out supportive roles, such as brainstorming, identifying blind spots, and surfacing alternatives, rather than executing complex workflows end-to-end. Our results highlight both the current limitations of LLMs in supporting complex legal workflows and opportunities for developing more collaborative, reasoning-aware legal AI systems. All data and code are available on our project page (https://minnesotanlp.github.io/LawFlow-website/).
Authors: Nay Klaimi (IETR, INSA Rennes), Amira Bedoui (IETR, INSA Rennes), Cl\'ement Elvira (IETR), Philippe Mary (INSA Rennes, IETR), Luc Le Magoarou (INSA Rennes, IETR)
Abstract: Hybrid precoding is a key ingredient of cost-effective massive multiple-input multiple-output transceivers. However, setting jointly digital and analog precoders to optimally serve multiple users is a difficult optimization problem. Moreover, it relies heavily on precise knowledge of the channels, which is difficult to obtain, especially when considering realistic systems comprising hardware impairments. In this paper, a joint channel estimation and hybrid precoding method is proposed, which consists in an end-to-end architecture taking received pilots as inputs and outputting pre-coders. The resulting neural network is fully model-based, making it lightweight and interpretable with very few learnable parameters. The channel estimation step is performed using the unfolded matching pursuit algorithm, accounting for imperfect knowledge of the antenna system, while the precoding step is done via unfolded projected gradient ascent. The great potential of the proposed method is empirically demonstrated on realistic synthetic channels.
Authors: Dhruvesh Patel, Aishwarya Sahoo, Avinash Amballa, Tahira Naseem, Tim G. J. Rudner, Andrew McCallum
Abstract: Autoregressive models (ARMs), which predict subsequent tokens one-by-one ``from left to right,'' have achieved significant success across a wide range of sequence generation tasks. However, they struggle to accurately represent sequences that require satisfying sophisticated constraints or whose sequential dependencies are better addressed by out-of-order generation. Masked Diffusion Models (MDMs) address some of these limitations, but the process of unmasking multiple tokens simultaneously in MDMs can introduce incoherences, and MDMs cannot handle arbitrary infilling constraints when the number of tokens to be filled in is not known in advance. In this work, we introduce Insertion Language Models (ILMs), which learn to insert tokens at arbitrary positions in a sequence -- that is, they select jointly both the position and the vocabulary element to be inserted. By inserting tokens one at a time, ILMs can represent strong dependencies between tokens, and their ability to generate sequences in arbitrary order allows them to accurately model sequences where token dependencies do not follow a left-to-right sequential structure. To train ILMs, we propose a tailored network parameterization and use a simple denoising objective. Our empirical evaluation demonstrates that ILMs outperform both ARMs and MDMs on common planning tasks. Furthermore, we show that ILMs outperform MDMs and perform on par with ARMs in an unconditional text generation task while offering greater flexibility than MDMs in arbitrary-length text infilling. The code is available at: https://dhruveshp.com/projects/ilm .
Authors: Wei Liu, Siya Qi, Xinyu Wang, Chen Qian, Yali Du, Yulan He
Abstract: Recent advances such as DeepSeek R1-Zero highlight the effectiveness of incentive training, a reinforcement learning paradigm that computes rewards solely based on the final answer part of a language model's output, thereby encouraging the generation of intermediate reasoning steps. However, these methods fundamentally rely on external verifiers, which limits their applicability to domains like mathematics and coding where such verifiers are readily available. Although reward models can serve as verifiers, they require high-quality annotated data and are costly to train. In this work, we propose NOVER, NO-VERifier Reinforcement Learning, a general reinforcement learning framework that requires only standard supervised fine-tuning data with no need for an external verifier. NOVER enables incentive training across a wide range of text-to-text tasks and outperforms the model of the same size distilled from large reasoning models such as DeepSeek R1 671B by 7.7 percent. Moreover, the flexibility of NOVER enables new possibilities for optimizing large language models, such as inverse incentive training.
Authors: Shuichi Nishino, Tomohiro Shiraishi, Teruyuki Katsuoka, Ichiro Takeuchi
Abstract: Graph Neural Networks (GNNs) have gained prominence for their ability to process graph-structured data across various domains. However, interpreting GNN decisions remains a significant challenge, leading to the adoption of saliency maps for identifying salient subgraphs composed of influential nodes and edges. Despite their utility, the reliability of GNN saliency maps has been questioned, particularly in terms of their robustness to input noise. In this study, we propose a statistical testing framework to rigorously evaluate the significance of saliency maps. Our main contribution lies in addressing the inflation of the Type I error rate caused by double-dipping of data, leveraging the framework of Selective Inference. Our method provides statistically valid $p$-values while controlling the Type I error rate, ensuring that identified salient subgraphs contain meaningful information rather than random artifacts. The method is applicable to a variety of saliency methods with piecewise linearity (e.g., Class Activation Mapping). We validate our method on synthetic and real-world datasets, demonstrating its capability in assessing the reliability of GNN interpretations.
Authors: Kristin Qi, Jiali Cheng, Youxiang Zhu, Hadi Amiri, Xiaohui Liang
Abstract: Detecting Mild Cognitive Impairment from picture descriptions is critical yet challenging, especially in multilingual and multiple picture settings. Prior work has primarily focused on English speakers describing a single picture (e.g., the 'Cookie Theft'). The TAUKDIAL-2024 challenge expands this scope by introducing multilingual speakers and multiple pictures, which presents new challenges in analyzing picture-dependent content. To address these challenges, we propose a framework with three components: (1) enhancing discriminative representation learning via supervised contrastive learning, (2) involving image modality rather than relying solely on speech and text modalities, and (3) applying a Product of Experts (PoE) strategy to mitigate spurious correlations and overfitting. Our framework improves MCI detection performance, achieving a +7.1% increase in Unweighted Average Recall (UAR) (from 68.1% to 75.2%) and a +2.9% increase in F1 score (from 80.6% to 83.5%) compared to the text unimodal baseline. Notably, the contrastive learning component yields greater gains for the text modality compared to speech. These results highlight our framework's effectiveness in multilingual and multi-picture MCI detection.
Authors: Bayu Adhi Tama, Mansa Krishna, Homayra Alam, Mostafa Cham, Omar Faruque, Gong Cheng, Jianwu Wang, Mathieu Morlighem, Vandana Janeja
Abstract: Understanding Greenland's subglacial topography is critical for projecting the future mass loss of the ice sheet and its contribution to global sea-level rise. However, the complex and sparse nature of observational data, particularly information about the bed topography under the ice sheet, significantly increases the uncertainty in model projections. Bed topography is traditionally measured by airborne ice-penetrating radar that measures the ice thickness directly underneath the aircraft, leaving data gap of tens of kilometers in between flight lines. This study introduces a deep learning framework, which we call as DeepTopoNet, that integrates radar-derived ice thickness observations and BedMachine Greenland data through a novel dynamic loss-balancing mechanism. Among all efforts to reconstruct bed topography, BedMachine has emerged as one of the most widely used datasets, combining mass conservation principles and ice thickness measurements to generate high-resolution bed elevation estimates. The proposed loss function adaptively adjusts the weighting between radar and BedMachine data, ensuring robustness in areas with limited radar coverage while leveraging the high spatial resolution of BedMachine predictions i.e. bed estimates. Our approach incorporates gradient-based and trend surface features to enhance model performance and utilizes a CNN architecture designed for subgrid-scale predictions. By systematically testing on the Upernavik Isstr{\o}m) region, the model achieves high accuracy, outperforming baseline methods in reconstructing subglacial terrain. This work demonstrates the potential of deep learning in bridging observational gaps, providing a scalable and efficient solution to inferring subglacial topography.
Authors: Anarya Ray
Abstract: Population synthesis simulations of compact binary coalescences~(CBCs) play a crucial role in extracting astrophysical insights from an ensemble of gravitational wave~(GW) observations. However, realistic simulations can be costly to implement for a dense grid of initial conditions. Normalizing flows can emulate population synthesis runs to enable simulation-based inference from observed catalogs and data augmentation for feature prediction in rarely synthesizable sub-populations. However, flow predictions can be wrought with uncertainties, especially for sparse training sets. In this work, we develop a method for quantifying and marginalizing uncertainties in the emulators by implementing the Bayesian Normalizing flow, a conditional density estimator constructed from Bayesian neural networks. Using the exact likelihood function naturally associated with density estimators, we sample the posterior distribution of flow parameters with suitably chosen priors to quantify and marginalize over flow uncertainties. We demonstrate the accuracy, calibration, inference, and data-augmentation impacts of the estimated uncertainties for simulations of binary black hole populations formed through common envelope evolution. We outline the applications of the proposed methodology in the context of simulation-based inference from growing GW catalogs and feature prediction, with state-of-the-art binary evolution simulators, now marginalized over model and data uncertainties.
Authors: Christopher Subia-Waud (Rayonlabs Team)
Abstract: Current AutoML platforms leave substantial performance untapped. Testing 180 fine-tuning tasks across models from 70M to 70B parameters, we found that HuggingFace AutoTrain, TogetherAI, Databricks, and Google Cloud consistently produce suboptimal configurations. Gradients, built on the Bittensor network, attacks this problem through competition. Independent miners race to find optimal hyperparameters, earning rewards proportional to their models' performance. This tournament drives exploration of configuration spaces that single-strategy methods never examine. In our experiments, Gradients achieved a 100\% win rate against TogetherAI, Databricks, and Google Cloud, and beat HuggingFace AutoTrain in 82.8\% of experiments. Mean improvements reached 42.1\% against commercial platforms. Retrieval-augmented generation tasks saw 30-40\% gains; diffusion models improved 23.4\% on person-specific generation. When miners compete for rewards, they develop optimization strategies that centralized approaches overlook. These findings demonstrate that decentralized systems with economic incentives can systematically outperform traditional AutoML, suggesting market dynamics may be key to achieving superior fine-tuning results. Code is available at https://github.com/rayonlabs/G.O.D.
Authors: Rohit Mohan, Julia Hindel, Florian Drews, Claudius Gl\"aser, Daniele Cattaneo, Abhinav Valada
Abstract: Autonomous vehicles that navigate in open-world environments may encounter previously unseen object classes. However, most existing LiDAR panoptic segmentation models rely on closed-set assumptions, failing to detect unknown object instances. In this work, we propose ULOPS, an uncertainty-guided open-set panoptic segmentation framework that leverages Dirichlet-based evidential learning to model predictive uncertainty. Our architecture incorporates separate decoders for semantic segmentation with uncertainty estimation, embedding with prototype association, and instance center prediction. During inference, we leverage uncertainty estimates to identify and segment unknown instances. To strengthen the model's ability to differentiate between known and unknown objects, we introduce three uncertainty-driven loss functions. Uniform Evidence Loss to encourage high uncertainty in unknown regions. Adaptive Uncertainty Separation Loss ensures a consistent difference in uncertainty estimates between known and unknown objects at a global scale. Contrastive Uncertainty Loss refines this separation at the fine-grained level. To evaluate open-set performance, we extend benchmark settings on KITTI-360 and introduce a new open-set evaluation for nuScenes. Extensive experiments demonstrate that ULOPS consistently outperforms existing open-set LiDAR panoptic segmentation methods.
Authors: Zhiyuan Wang, Bokui Chen
Abstract: Continual learning (CL) empowers pre-trained vision-language models to adapt effectively to novel or previously underrepresented data distributions without comprehensive retraining, enhancing their adaptability and efficiency. While vision-language models like CLIP show great promise, they struggle to maintain performance across domains in incremental learning scenarios. Existing prompt learning methods face two main limitations: 1) they primarily focus on class-incremental learning scenarios, lacking specific strategies for multi-domain task incremental learning; 2) most current approaches employ single-modal prompts, neglecting the potential benefits of cross-modal information exchange. To address these challenges, we propose the \ChordPrompt framework, which facilitates a harmonious interplay between visual and textual prompts. \ChordPrompt introduces cross-modal prompts to leverage interactions between visual and textual information. Our approach also employs domain-adaptive text prompts to select appropriate prompts for continual adaptation across multiple domains. Comprehensive experiments on multi-domain incremental learning benchmarks demonstrate that \ChordPrompt outperforms state-of-the-art methods in zero-shot generalization and downstream task performance.
Authors: Hongjie Wu, Mingqin Zhang, Linchao He, Ji-Zhe Zhou, Jiancheng Lv
Abstract: Diffusion models have shown remarkable promise for image restoration by leveraging powerful priors. Prominent methods typically frame the restoration problem within a Bayesian inference framework, which iteratively combines a denoising step with a likelihood guidance step. However, the interactions between these two components in the generation process remain underexplored. In this paper, we analyze the underlying gradient dynamics of these components and identify significant instabilities. Specifically, we demonstrate conflicts between the prior and likelihood gradient directions, alongside temporal fluctuations in the likelihood gradient itself. We show that these instabilities disrupt the generative process and compromise restoration performance. To address these issues, we propose Stabilized Progressive Gradient Diffusion (SPGD), a novel gradient management technique. SPGD integrates two synergistic components: (1) a progressive likelihood warm-up strategy to mitigate gradient conflicts; and (2) adaptive directional momentum (ADM) smoothing to reduce fluctuations in the likelihood gradient. Extensive experiments across diverse restoration tasks demonstrate that SPGD significantly enhances generation stability, leading to state-of-the-art performance in quantitative metrics and visually superior results. Code is available at https://github.com/74587887/SPGD.
Authors: Uri Cohen, M\'at\'e Lengyel
Abstract: Neural networks storing multiple discrete attractors are canonical models of biological memory. Previously, the dynamical stability of such networks could only be guaranteed under highly restrictive conditions. Here, we derive a theory of the local stability of discrete fixed points in a broad class of networks with graded neural activities and in the presence of noise. By directly analyzing the bulk and outliers of the Jacobian spectrum, we show that all fixed points are stable below a critical load that is distinct from the classical \textit{critical capacity} and depends on the statistics of neural activities in the fixed points as well as the single-neuron activation function. Our analysis highlights the computational benefits of threshold-linear activation and sparse-like patterns.
Authors: Takashi Hayakawa, Satoshi Asai
Abstract: Previous studies have shown that hazard ratios between treatment groups estimated with the Cox model are uninterpretable because the unspecified baseline hazard of the model fails to identify temporal change in the risk set composition due to treatment assignment and unobserved factors among multiple, contradictory scenarios. To alleviate this problem, especially in studies based on observational data with uncontrolled dynamic treatment and real-time measurement of many covariates, we propose abandoning the baseline hazard and using kernel-based machine learning to explicitly model the change in the risk set with or without latent variables. For this framework, we clarify the context in which hazard ratios can be causally interpreted, and then develop a method based on Neyman orthogonality to compute debiased maximum-likelihood estimators of hazard ratios, proving necessary convergence results. Numerical simulations confirm that the proposed method identifies the true hazard ratios with minimal bias. These results lay the foundation for developing a useful, alternative method for causal inference with uncontrolled, observational data in modern epidemiology.
Authors: Roopkatha Banerjee, Sampath Koti, Gyanendra Singh, Anirban Chakraborty, Gurunath Gurrala, Bhushan Jagyasi, Yogesh Simmhan
Abstract: Real-time monitoring of power consumption in cities and micro-grids through the Internet of Things (IoT) can help forecast future demand and optimize grid operations. But moving all consumer-level usage data to the cloud for predictions and analysis at fine time scales can expose activity patterns. Federated Learning~(FL) is a privacy-sensitive collaborative DNN training approach that retains data on edge devices, trains the models on private data locally, and aggregates the local models in the cloud. But key challenges exist: (i) clients can have non-independently identically distributed~(non-IID) data, and (ii) the learning should be computationally cheap while scaling to 1000s of (unseen) clients. In this paper, we develop and evaluate several optimizations to FL training across edge and cloud for time-series demand forecasting in micro-grids and city-scale utilities using DNNs to achieve a high prediction accuracy while minimizing the training cost. We showcase the benefit of using exponentially weighted loss while training and show that it further improves the prediction of the final model. Finally, we evaluate these strategies by validating over 1000s of clients for three states in the US from the OpenEIA corpus, and performing FL both in a pseudo-distributed setting and a Pi edge cluster. The results highlight the benefits of the proposed methods over baselines like ARIMA and DNNs trained for individual consumers, which are not scalable.
Authors: Qian-Rui Lee, Daw-Wei Wang
Abstract: We propose a machine learning framework based on Flow Matching to overcome the scaling limitations of Markov Chain Monte Carlo (MCMC) methods. We demonstrate its capability in the 2D XY model, where a single network, trained only on configurations from a small ($32\times 32$) lattice at sparse temperature points, generates reliable samples for a significantly larger system ($128\times 128$) across a continuous temperature range without retraining. The generated configurations show strong agreement with key thermodynamic observables and correctly capture the signatures of the Berezinskii-Kosterlitz-Thouless (BKT) transition. This dual generalization is enabled by the Flow Matching framework, which allows us to learn a continuous, temperature-conditioned mapping. At the same time, the inductive biases of the underlying CNN architecture ensure that the learned local physical rules are scale-invariant. This "train-small, generate-large" capability offers a powerful and efficient alternative for studying critical phenomena. The method can be directly applied to other classical or quantum many-body systems described by continuous fields on a lattice. Furthermore, this framework can serve as a powerful proposal generator in a hybrid scheme with MCMC, dramatically accelerating high-precision studies of the thermodynamic limit.
Authors: Xiancheng Gao, Yufeng Shi, Wengang Zhou, Houqiang Li
Abstract: Offline reinforcement learning refers to the process of learning policies from fixed datasets, without requiring additional environment interaction. However, it often relies on well-defined reward functions, which are difficult and expensive to design. Human feedback is an appealing alternative, but its two common forms, expert demonstrations and preferences, have complementary limitations. Demonstrations provide stepwise supervision, but they are costly to collect and often reflect limited expert behavior modes. In contrast, preferences are easier to collect, but it is unclear which parts of a behavior contribute most to a trajectory segment, leaving credit assignment unresolved. In this paper, we introduce a Search-Based Preference Weighting (SPW) scheme to unify these two feedback sources. For each transition in a preference labeled trajectory, SPW searches for the most similar state-action pairs from expert demonstrations and directly derives stepwise importance weights based on their similarity scores. These weights are then used to guide standard preference learning, enabling more accurate credit assignment that traditional approaches struggle to achieve. We demonstrate that SPW enables effective joint learning from preferences and demonstrations, outperforming prior methods that leverage both feedback types on challenging robot manipulation tasks.
Authors: Luca Ambrogioni
Abstract: Generative diffusion models have emerged as a powerful class of models in machine learning, yet a unified theoretical understanding of their operation is still developing. This perspective paper provides an integrated perspective on generative diffusion by connecting their dynamic, information-theoretic, and thermodynamic properties under a unified mathematical framework. We demonstrate that the rate of conditional entropy production during generation (i.e. the generative bandwidth) is directly governed by the expected divergence of the score function's vector field. This divergence, in turn, is linked to the branching of trajectories and generative bifurcations, which we characterize as symmetry-breaking phase transitions in the energy landscape. This synthesis offers a powerful insight: the process of generation is fundamentally driven by the controlled, noise-induced breaking of (approximate) symmetries, where peaks in information transfer correspond to critical transitions between possible outcomes. The score function acts as a dynamic non-linear filter that regulates the bandwidth of the noise by suppressing fluctuations that are incompatible with the data.
Authors: Malte L\"uken, Javier Garcia-Bernardo, Sreeparna Deb, Flavio Hafner, Megha Khosla
Abstract: Administrative registry data can be used to construct population-scale networks whose ties reflect shared social contexts between persons. With machine learning, such networks can be encoded into numerical representations -- embeddings -- that automatically capture individuals' position within the network. We created embeddings for all persons in the Dutch population from a population-scale network that represents five shared contexts: neighborhood, work, family, household, and school. To assess the informativeness of these embeddings, we used them to predict right-wing populist voting. Embeddings alone predicted right-wing populist voting above chance-level but performed worse than individual characteristics. Combining the best subset of embeddings with individual characteristics only slightly improved predictions. After transforming the embeddings to make their dimensions more sparse and orthogonal, we found that one embedding dimension was strongly associated with the outcome. Mapping this dimension back to the population network revealed differences in network structure related to right-wing populist voting between different school ties and achieved education levels. Our study contributes methodologically by demonstrating how population-scale network embeddings can be made interpretable, and substantively by linking structural network differences in education to right-wing populist voting.
Authors: Justin Jung
Abstract: Generating realistic sparse multi-category 3D voxel structures is difficult due to the cubic memory scaling of voxel structures and moreover the significant class imbalance caused by sparsity. We introduce Scaffold Diffusion, a generative model designed for sparse multi-category 3D voxel structures. By treating voxels as tokens, Scaffold Diffusion uses a discrete diffusion language model to generate 3D voxel structures. We show that discrete diffusion language models can be extended beyond inherently sequential domains such as text to generate spatially coherent 3D structures. We evaluate on Minecraft house structures from the 3D-Craft dataset and demonstrate that, unlike prior baselines and an auto-regressive formulation, Scaffold Diffusion produces realistic and coherent structures even when trained on data with over 98% sparsity. We provide an interactive viewer where readers can visualize generated samples and the generation process: https://scaffold.deepexploration.org/
Authors: I. Shavindra Jayasekera, Jacob Si, Filippo Valdettaro, Wenlong Chen, A. Aldo Faisal, Yingzhen Li
Abstract: As large language models (LLMs) gain popularity in conducting prediction tasks in-context, understanding the sources of uncertainty in in-context learning becomes essential to ensuring reliability. The recent hypothesis of in-context learning performing predictive Bayesian inference opens the avenue for Bayesian uncertainty estimation, particularly for decomposing uncertainty into epistemic uncertainty due to lack of in-context data and aleatoric uncertainty inherent in the in-context prediction task. However, the decomposition idea remains under-explored due to the intractability of the latent parameter posterior from the underlying Bayesian model. In this work, we introduce a variational uncertainty decomposition framework for in-context learning without explicitly sampling from the latent parameter posterior, by optimising auxiliary queries as probes to obtain an upper bound to the aleatoric uncertainty of an LLM's in-context learning procedure, which also induces a lower bound to the epistemic uncertainty. Through experiments on synthetic and real-world tasks, we show quantitatively and qualitatively that the decomposed uncertainties obtained from our method exhibit desirable properties of epistemic and aleatoric uncertainty.
Authors: Haichen Hu, David Simchi-Levi
Abstract: We study the problem of evaluating the excess risk of classical penalized empirical risk minimization (ERM) with Bregman losses. We show that by leveraging the recently proposed wild refitting procedure (Wainwright, 2025), one can efficiently upper bound the excess risk through the so-called "wild optimism," without relying on the global structure of the underlying function class. This property makes our approach inherently model-free. Unlike conventional analyses, our framework operates with just one dataset and black-box access to the training procedure. The method involves randomized vector-valued symmetrization with an appropriate scaling of the prediction residues and constructing artificially modified outcomes, upon which we retrain a second predictor for excess risk estimation. We establish high-probability performance guarantees both under the fixed design setting and the random design setting, demonstrating that wild refitting under Bregman losses, with an appropriately chosen wild noise scale, yields a valid upper bound on the excess risk. This work thus is promising for theoretically evaluating modern opaque ML and AI models such as deep neural networks and large language models, where the model class is too complex for classical learning theory and empirical process techniques to apply.