Authors: Md. Najmul Islam, Md. Miner Hossain Rimon, Shah Sadek-E-Akbor Shamim, Zarif Mohaimen Fahad, Md. Jehadul Islam Mony, Md. Jalal Uddin Chowdhury
Abstract: Diabetes is a serious worldwide health issue, and successful intervention depends on early detection. However, overlapping risk factors and data asymmetry make prediction difficult. To use extensive health survey data to create a machine learning framework for diabetes classification that is both accurate and comprehensible, to produce results that will aid in clinical decision-making. Using the BRFSS dataset, we assessed a number of supervised learning techniques. SMOTE and Tomek Links were used to correct class imbalance. To improve prediction performance, both individual models and ensemble techniques such as stacking were investigated. The 2015 BRFSS dataset, which includes roughly 253,680 records with 22 numerical features, is used in this study. Strong ROC-AUC performance of approximately 0.96 was attained by the individual models Random Forest, XGBoost, CatBoost, and LightGBM.The stacking ensemble with XGBoost and KNN yielded the best overall results with 94.82\% accuracy, ROC-AUC of 0.989, and PR-AUC of 0.991, indicating a favourable balance between recall and precision. In our study, we proposed and developed a React Native-based application with a Python Flask backend to support early diabetes prediction, providing users with an accessible and efficient health monitoring tool.
Authors: Omar Mahmood, Pedro O. Pinheiro, Richard Bonneau, Saeed Saremi, Vishnu Sresht
Abstract: Ligand-based drug discovery (LBDD) relies on making use of known binders to a protein target to find structurally diverse molecules similarly likely to bind. This process typically involves a brute force search of the known binder (query) against a molecular library using some metric of molecular similarity. One popular approach overlays the pharmacophore-shape profile of the known binder to 3D conformations enumerated for each of the library molecules, computes overlaps, and picks a set of diverse library molecules with high overlaps. While this virtual screening workflow has had considerable success in hit diversification, scaffold hopping, and patent busting, it scales poorly with library sizes and restricts candidate generation to existing library compounds. Leveraging recent advances in voxel-based generative modelling, we propose a pharmacophore-based generative model and workflows that address the scaling and fecundity issues of conventional pharmacophore-based virtual screening. We introduce \emph{VoxCap}, a voxel captioning method for generating SMILES strings from voxelised molecular representations. We propose two workflows as practical use cases as well as benchmarks for pharmacophore-based generation: \emph{de-novo} design, in which we aim to generate new molecules with high pharmacophore-shape similarities to query molecules, and fast search, which aims to combine generative design with a cheap 2D substructure similarity search for efficient hit identification. Our results show that VoxCap significantly outperforms previous methods in generating diverse \textit{de-novo} hits. When combined with our fast search workflow, VoxCap reduces computational time by orders of magnitude while returning hits for all query molecules, enabling the search of large libraries that are intractable to search by brute force.
Authors: R\'emi Marsal, St\'ephanie Chaillat
Abstract: The boundary element method (BEM) provides an efficient numerical framework for solving multiple scattering problems in unbounded homogeneous domains, since it reduces the discretization to the domain boundaries, thereby condensing the computational complexity. The procedure first consists in determining the solution trace on the boundaries of the domain by solving a boundary integral equation, after which the volumetric solution can be recovered at low computational cost with a boundary integral representation. As the first step of the BEM represents the main computational bottleneck, we introduce PIBNet, a learning-based approach designed to approximate the solution trace. The method leverages a physics-inspired graph-based strategy to model obstacles and their long-range interactions efficiently. Then, we introduce a novel multiscale graph neural network architecture for simulating the multiple scattering. To train and evaluate our network, we present a benchmark consisting of several datasets of different types of multiple scattering problems. The results indicate that our approach not only surpasses existing state-of-the-art learning-based methods on the considered tasks but also exhibits superior generalization to settings with an increased number of obstacles. github.com/ENSTA-U2IS-AI/pibnet
Authors: M. A. Gameiro
Abstract: This study introduces a significant architectural advancement in feature fusion for lyrical content classification by integrating auxiliary structural features directly into the self-attention mechanism of a pre-trained Transformer. I propose the SFL Transformer, a novel deep learning model that utilizes a Contextual Gating mechanism (an Intermediate SFL) to modulate the sequence of hidden states within the BERT encoder stack, rather than fusing features at the final output layer. This approach modulates the deep, contextualized semantic features (Hseq) using low-dimensional structural cues (Fstruct). The model is applied to a challenging binary classification task derived from UMAP-reduced lyrical embeddings. The SFL Transformer achieved an Accuracy of 0.9910 and a Macro F1 score of 0.9910, significantly improving the state-of-the-art established by the previously published SFL model (Accuracy 0.9894). Crucially, this Contextual Gating strategy maintained exceptional reliability, with a low Expected Calibration Error (ECE = 0.0081) and Log Loss (0.0489). This work validates the hypothesis that injecting auxiliary context mid-stack is the most effective means of synergistically combining structural and semantic information, creating a model with both superior discriminative power and high-fidelity probability estimates.
Authors: Haoxiang Zhang, Ruihao Yuan, Lihui Zhang, Yushi Luo, Qiang Zhang, Pan Ding, Xiaodong Ren, Weijie Xing, Niu Gao, Jishan Chen, Chubo Zhang
Abstract: The industrial adoption of Artificial Intelligence for Engineering (AI4E) faces two fundamental bottlenecks: scarce high-quality data and the lack of interpretability in black-box models-particularly critical in safety-sensitive sectors like aerospace. We present an explainable, few-shot AI4E framework that is systematically informed by physics and expert knowledge throughout its architecture. Starting from only 32 experimental samples in an aerial K439B superalloy castings repair welding case, we first augment physically plausible synthetic data through a three-stage protocol: differentiated noise injection calibrated to process variabilities, enforcement of hard physical constraints, and preservation of inter-parameter relationships. We then employ a nested optimization strategy for constitutive model discovery, where symbolic regression explores equation structures while differential evolution optimizes parameters, followed by intensive parameter refinement using hybrid global-local optimization. The resulting interpretable constitutive equation achieves 88% accuracy in predicting hot-cracking tendency. This equation not only provides quantitative predictions but also delivers explicit physical insight, revealing how thermal, geometric, and metallurgical mechanisms couple to drive cracking-thereby advancing engineers' cognitive understanding of the process. Furthermore, the constitutive equation serves as a multi-functional tool for process optimization and high-fidelity virtual data generation, enabling accuracy improvements in other data-driven models. Our approach provides a general blueprint for developing trustworthy AI systems that embed engineering domain knowledge directly into their architecture, enabling reliable adoption in high-stakes industrial applications where data is limited but physical understanding is available.
Authors: Zhenliang Ni, Xiaowen Ma, Zhenkai Wu, Shuai Xiao, Han Shu, Xinghao Chen
Abstract: Multivariate time series forecasts are widely used, such as industrial, transportation and financial forecasts. However, the dominant frequencies in time series may shift with the evolving spectral distribution of the data. Traditional Mixture of Experts (MoE) models, which employ a fixed number of experts, struggle to adapt to these changes, resulting in frequency coverage imbalance issue. Specifically, too few experts can lead to the overlooking of critical information, while too many can introduce noise. To this end, we propose Ada-MoGE, an adaptive Gaussian Mixture of Experts model. Ada-MoGE integrates spectral intensity and frequency response to adaptively determine the number of experts, ensuring alignment with the input data's frequency distribution. This approach prevents both information loss due to an insufficient number of experts and noise contamination from an excess of experts. Additionally, to prevent noise introduction from direct band truncation, we employ Gaussian band-pass filtering to smoothly decompose the frequency domain features, further optimizing the feature representation. The experimental results show that our model achieves state-of-the-art performance on six public benchmarks with only 0.2 million parameters.
Authors: Li Qianyang, Zhang Xingjun, Wang Shaoxun, Wei Jia
Abstract: Long-term time series forecasting (LTSF) is a critical task in computational intelligence. While Transformer-based models effectively capture long-range dependencies, they often suffer from quadratic complexity and overfitting due to data sparsity. Conversely, efficient linear models struggle to depict complex non-linear local dynamics. Furthermore, existing multi-scale frameworks typically rely on average pooling, which acts as a non-ideal low-pass filter, leading to spectral aliasing and the irreversible loss of high-frequency transients. In response, this paper proposes DPWMixer, a computationally efficient Dual-Path architecture. The framework is built upon a Lossless Haar Wavelet Pyramid that replaces traditional pooling, utilizing orthogonal decomposition to explicitly disentangle trends and local fluctuations without information loss. To process these components, we design a Dual-Path Trend Mixer that integrates a global linear mapping for macro-trend anchoring and a flexible patch-based MLP-Mixer for micro-dynamic evolution. Finally, An adaptive multi-scale fusion module then integrates predictions from diverse scales, weighted by channel stationarity to optimize synthesis. Extensive experiments on eight public benchmarks demonstrate that our method achieves a consistent improvement over state-of-the-art baselines. The code is available at https://github.com/hit636/DPWMixer.
Authors: Qirui Ji, Bin Qin, Yifan Jin, Yunze Zhao, Chuxiong Sun, Changwen Zheng, Jianwen Cao, Jiangmeng Li
Abstract: Graph contrastive learning (GCL) aims to learn discriminative semantic invariance by contrasting different views of the same graph that share critical topological patterns. However, existing GCL approaches with structural augmentations often struggle to identify task-relevant topological structures, let alone adapt to the varying coarse-to-fine topological granularities required across different downstream tasks. To remedy this issue, we introduce Hierarchical Topological Granularity Graph Contrastive Learning (HTG-GCL), a novel framework that leverages transformations of the same graph to generate multi-scale ring-based cellular complexes, embodying the concept of topological granularity, thereby generating diverse topological views. Recognizing that a certain granularity may contain misleading semantics, we propose a multi-granularity decoupled contrast and apply a granularity-specific weighting mechanism based on uncertainty estimation. Comprehensive experiments on various benchmarks demonstrate the effectiveness of HTG-GCL, highlighting its superior performance in capturing meaningful graph representations through hierarchical topological information.
Authors: Haozhe Wu
Abstract: This study focuses on the feature extraction problem in multi-modal data regression. To address three core challenges in real-world scenarios: limited and non-IID data, effective extraction and fusion of multi-modal information, and susceptibility to catastrophic forgetting in model learning, a task-driven supervised multi-modal federated feature extraction method is proposed. The method integrates multi-modal information extraction and contrastive learning mechanisms, and can adapt to different neural network structures as the latent mapping functions for data of each modality. It supports each client to independently learn low-dimensional representations of multi-modal data, and can flexibly control the degree of retention of effective information about the response variable in the predictive variables within the low-dimensional features through parameter tuning. The multi-constraint learning framework constructed by the method guarantees regression accuracy using Mean Squared Error loss. Through the synergistic effect of mutual information preservation constraint, symmetric Kullback-Leibler divergence constraint, and inter-model contrastive constraint, it achieves the retention of task-related information, the extraction, fusion, and alignment of multi-modal features, and the mitigation of representation drift and catastrophic forgetting in non-IID scenarios, respectively. This ensures that the feature extraction process always centers on improving the performance of downstream regression tasks. Experimental results from simulations and real-world data analysis demonstrate that the proposed method achieves more significant performance improvement on downstream regression tasks compared with classical feature extraction techniques.
Authors: Ahmet Sami Korkmaz, Selim Coskunuzer, Md Joshem Uddin
Abstract: Graph classification has gained significant attention due to its applications in chemistry, social networks, and bioinformatics. While Graph Neural Networks (GNNs) effectively capture local structural patterns, they often overlook global topological features that are critical for robust representation learning. In this work, we propose GraphTCL, a dual-view contrastive learning framework that integrates structural embeddings from GNNs with topological embeddings derived from persistent homology. By aligning these complementary views through a cross-view contrastive loss, our method enhances representation quality and improves classification performance. Extensive experiments on benchmark datasets, including TU and OGB molecular graphs, demonstrate that GraphTCL consistently outperforms state-of-the-art baselines. This study highlights the importance of topology-aware contrastive learning for advancing graph representation methods.
Authors: Aheer Sravon, Md. Ibrahim, Devdyuti Mazumder, Ridwan Al Aziz
Abstract: Autonomous pricing algorithms are increasingly influencing competition in digital markets; however, their behavior under realistic demand conditions remains largely unexamined. This paper offers a thorough analysis of four pricing algorithms -- Q-Learning, PSO, Double DQN, and DDPG -- across three classic duopoly models (Logit, Hotelling, Linear) and under various demand-shock regimes created by auto-regressive processes. By utilizing profit- and price-based collusion indices, we investigate how the interactions among algorithms, market structure, and stochastic demand collaboratively influence competitive outcomes. Our findings reveal that reinforcement-learning algorithms often sustain supra-competitive prices under stable demand, with DDPG demonstrating the most pronounced collusive tendencies. Demand shocks produce notably varied effects: Logit markets suffer significant performance declines, Hotelling markets remain stable, and Linear markets experience shock-induced profit inflation. Despite marked changes in absolute performance, the relative rankings of the algorithms are consistent across different environments. These results underscore the critical importance of market structure and demand uncertainty in shaping algorithmic competition, while also contributing to the evolving policy discussions surrounding autonomous pricing behavior.
Authors: Yuxuan Shu, Peter H. Charlton, Fahim Kawsar, Jussi Hernesniemi, Mohammad Malekzadeh
Abstract: The electrocardiogram (ECG) is a key diagnostic tool in cardiovascular health. Single-lead ECG recording is integrated into both clinical-grade and consumer wearables. While self-supervised pretraining of foundation models on unlabeled ECGs improves diagnostic performance, existing approaches do not incorporate domain knowledge from clinical metadata. We introduce a novel contrastive learning approach that utilizes an established clinical risk score to adaptively weight negative pairs: clinically-guided contrastive learning. It aligns the similarities of ECG embeddings with clinically meaningful differences between subjects, with an explicit mechanism to handle missing metadata. On 12-lead ECGs from 161K patients in the MIMIC-IV dataset, we pretrain single-lead ECG foundation models at three scales, collectively called CLEF, using only routinely collected metadata without requiring per-sample ECG annotations. We evaluate CLEF on 18 clinical classification and regression tasks across 7 held-out datasets, and benchmark against 5 foundation model baselines and 3 self-supervised algorithms. When pretrained on 12-lead ECG data and tested on lead-I data, CLEF outperforms self-supervised foundation model baselines: the medium-sized CLEF achieves average AUROC improvements of at least 2.6% in classification and average reductions in MAEs of at least 3.2% in regression. Comparing with existing self-supervised learning algorithms, CLEF improves the average AUROC by at least 1.8%. Moreover, when pretrained only on lead-I data for classification tasks, CLEF performs comparably to the state-of-the-art ECGFounder, which was trained in a supervised manner. Overall, CLEF enables more accurate and scalable single-lead ECG analysis, advancing remote health monitoring. Code and pretrained CLEF models are available at: github.com/Nokia-Bell-Labs/ecg-foundation-model.
Authors: Sophie L. Wang, Alex Quach, Nithin Parsan, John J. Yang
Abstract: Sparse autoencoders (SAEs) have been widely used for interpretability of neural networks, but their learned features often vary across seeds and hyperparameter settings. We introduce Ordered Sparse Autoencoders (OSAE), which extend Matryoshka SAEs by (1) establishing a strict ordering of latent features and (2) deterministically using every feature dimension, avoiding the sampling-based approximations of prior nested SAE methods. Theoretically, we show that OSAEs resolve permutation non-identifiability in settings of sparse dictionary learning where solutions are unique (up to natural symmetries). Empirically on Gemma2-2B and Pythia-70M, we show that OSAEs can help improve consistency compared to Matryoshka baselines.
Authors: Stefano Vrizzi, Daniel W. O'Neill
Abstract: The 'Doughnut' of social and planetary boundaries has emerged as a popular framework for assessing environmental and social sustainability. Here, we provide a proof-of-concept analysis that shows how machine learning (ML) methods can be applied to a simple macroeconomic model of the Doughnut. First, we show how ML methods can be used to find policy parameters that are consistent with 'living within the Doughnut'. Second, we show how a reinforcement learning agent can identify the optimal trajectory towards desired policies in the parameter space. The approaches we test, which include a Random Forest Classifier and $Q$-learning, are frugal ML methods that are able to find policy parameter combinations that achieve both environmental and social sustainability. The next step is the application of these methods to a more complex ecological macroeconomic model.
Authors: Orr Paradise, Pranav Muralikrishnan, Liangyuan Chen, Hugo Flores Garc\'ia, Bryan Pardo, Roee Diamant, David F. Gruber, Shane Gero, Shafi Goldwasser
Abstract: Sperm whales communicate in short sequences of clicks known as codas. We present WhAM (Whale Acoustics Model), the first transformer-based model capable of generating synthetic sperm whale codas from any audio prompt. WhAM is built by finetuning VampNet, a masked acoustic token model pretrained on musical audio, using 10k coda recordings collected over the past two decades. Through iterative masked token prediction, WhAM generates high-fidelity synthetic codas that preserve key acoustic features of the source recordings. We evaluate WhAM's synthetic codas using Fr\'echet Audio Distance and through perceptual studies with expert marine biologists. On downstream classification tasks including rhythm, social unit, and vowel classification, WhAM's learned representations achieve strong performance, despite being trained for generation rather than classification. Our code is available at https://github.com/Project-CETI/wham
Authors: Mamadou K. Keita, Sebastien Diarra, Christopher Homan, Seydou Diallo
Abstract: Effective text generation and chat interfaces for low-resource languages (LRLs) remain a challenge for state-of-the-art large language models (LLMs) to support. This is mainly due to the difficulty of curating high-quality instruction datasets for LRLs, a limitation prevalent in the languages spoken across the African continent and other regions. Current approaches, such as automated translation and synthetic data generation, frequently yield outputs that lack fluency or even orthographic consistency. In this paper, we introduce InstructLR, a novel framework designed to generate high-quality instruction datasets for LRLs. Our approach integrates LLM-driven text generation with a dual-layer quality filtering mechanism: an automated filtering layer based on retrieval-augmented-generation (RAG)-based n-shot prompting, and a human-in-the-loop validation layer. Drawing inspiration from benchmarks such as MMLU in task definition, InstructLR has facilitated the creation of three multi-domain instruction benchmarks: ZarmaInstruct-50k, BambaraInstruct-50k, and FulfuldeInstruct-50k.
Authors: Aida Afshar, Aldo Pacchiano
Abstract: We study the problem of online model selection in reinforcement learning, where the selector has access to a class of reinforcement learning agents and learns to adaptively select the agent with the right configuration. Our goal is to establish the improved efficiency and performance gains achieved by integrating online model selection methods into reinforcement learning training procedures. We examine the theoretical characterizations that are effective for identifying the right configuration in practice, and address three practical criteria from a theoretical perspective: 1) Efficient resource allocation, 2) Adaptation under non-stationary dynamics, and 3) Training stability across different seeds. Our theoretical results are accompanied by empirical evidence from various model selection tasks in reinforcement learning, including neural architecture selection, step-size selection, and self model selection.
Authors: F. Br\"uckerhoff-Pl\"uckelmann, H. Borras, S. U. Hulyal, L. Meyer, X. Ji, J. Hu, J. Sun, B. Klein, F. Ebert, J. Dijkstra, L. McRae, P. Schmidt, T. J. Kippenberg, H. Fr\"oning, W. Pernice
Abstract: Artificial intelligence (AI) systems increasingly influence safety-critical aspects of society, from medical diagnosis to autonomous mobility, making uncertainty awareness a central requirement for trustworthy AI. We present a photonic Bayesian machine that leverages the inherent randomness of chaotic light sources to enable uncertainty reasoning within the framework of Bayesian Neural Networks. The analog processor features a 1.28 Tbit/s digital interface compatible with PyTorch, enabling probabilistic convolutions processing within 37.5 ps per convolution. We use the system for simultaneous classification and out-of-domain detection of blood cell microscope images and demonstrate reasoning between aleatoric and epistemic uncertainties. The photonic Bayesian machine removes the bottleneck of pseudo random number generation in digital systems, minimizes the cost of sampling for probabilistic models, and thus enables high-speed trustworthy AI systems.
Authors: Benjamin K. Rosenzweig, Matthew W. Hahn
Abstract: Inferring the phylogenetic relationships among a sample of organisms is a fundamental problem in modern biology. While distance-based hierarchical clustering algorithms achieved early success on this task, these have been supplanted by Bayesian and maximum likelihood search procedures based on complex models of molecular evolution. In this work we describe minimal neural network architectures that can approximate classic phylogenetic distance functions and the properties required to learn distances under a variety of molecular evolutionary models. In contrast to model-based inference (and recently proposed model-free convolutional and transformer networks), these architectures have a small computational footprint and are scalable to large numbers of taxa and molecular characters. The learned distance functions generalize well and, given an appropriate training dataset, achieve results comparable to state-of-the art inference methods.
Authors: Joshua Wolff Anderson, Shyam Visweswaran
Abstract: Trustworthy machine learning in healthcare requires strong predictive performance, fairness, and explanations. While it is known that improving fairness can affect predictive performance, little is known about how fairness improvements influence explainability, an essential ingredient for clinical trust. Clinicians may hesitate to rely on a model whose explanations shift after fairness constraints are applied. In this study, we examine how enhancing fairness through bias mitigation techniques reshapes Shapley-based feature rankings. We quantify changes in feature importance rankings after applying fairness constraints across three datasets: pediatric urinary tract infection risk, direct anticoagulant bleeding risk, and recidivism risk. We also evaluate multiple model classes on the stability of Shapley-based rankings. We find that increasing model fairness across racial subgroups can significantly alter feature importance rankings, sometimes in different ways across groups. These results highlight the need to jointly consider accuracy, fairness, and explainability in model assessment rather than in isolation.
Authors: Jane Lange, Mingda Qiao
Abstract: Membership queries (MQ) often yield speedups for learning tasks, particularly in the distribution-specific setting. We show that in the \emph{testable learning} model of Rubinfeld and Vasilyan [RV23], membership queries cannot decrease the time complexity of testable learning algorithms beyond the complexity of sample-only distribution-specific learning. In the testable learning model, the learner must output a hypothesis whenever the data distribution satisfies a desired property, and if it outputs a hypothesis, the hypothesis must be near-optimal. We give a general reduction from sample-based \emph{refutation} of boolean concept classes, as presented in [Vadhan17, KL18], to testable learning with queries (TL-Q). This yields lower bounds for TL-Q via the reduction from learning to refutation given in [KL18]. The result is that, relative to a concept class and a distribution family, no $m$-sample TL-Q algorithm can be super-polynomially more time-efficient than the best $m$-sample PAC learner. Finally, we define a class of ``statistical'' MQ algorithms that encompasses many known distribution-specific MQ learners, such as those based on influence estimation or subcube-conditional statistical queries. We show that TL-Q algorithms in this class imply efficient statistical-query refutation and learning algorithms. Thus, combined with known SQ dimension lower bounds, our results imply that these efficient membership query learners cannot be made testable.
Authors: Max W. Shen, Ewa Nowara, Michael Maser, Kyunghyun Cho
Abstract: While data augmentation is widely used to train symmetry-agnostic models, it remains unclear how quickly and effectively they learn to respect symmetries. We investigate this by deriving a principled measure of equivariance error that, for convex losses, calculates the percent of total loss attributable to imperfections in learned symmetry. We focus our empirical investigation to 3D-rotation equivariance on high-dimensional molecular tasks (flow matching, force field prediction, denoising voxels) and find that models reduce equivariance error quickly to $\leq$2\% held-out loss within 1k-10k training steps, a result robust to model and dataset size. This happens because learning 3D-rotational equivariance is an easier learning task, with a smoother and better-conditioned loss landscape, than the main prediction task. For 3D rotations, the loss penalty for non-equivariant models is small throughout training, so they may achieve lower test loss than equivariant models per GPU-hour unless the equivariant ``efficiency gap'' is narrowed. We also experimentally and theoretically investigate the relationships between relative equivariance error, learning gradients, and model parameters.
Authors: Kentaro Kubo, Hayato Goto
Abstract: Boltzmann machines (BMs) are powerful energy-based generative models, but their heavy training cost has largely confined practical use to Restricted BMs (RBMs) trained with an efficient learning method called contrastive divergence. More accurate learning typically requires Markov chain Monte Carlo (MCMC) Boltzmann sampling, but it is time-consuming due to the difficulty of parallelization for more expressive models. To address this limitation, we first propose a new Boltzmann sampler inspired by a quantum-inspired combinatorial optimization called simulated bifurcation (SB). This SB-inspired approach, which we name Langevin SB (LSB), enables parallelized sampling while maintaining accuracy comparable to MCMC. Furthermore, this is applicable not only to RBMs but also to BMs with general couplings. However, LSB cannot control the inverse temperature of the output Boltzmann distribution, which hinders learning and degrades performance. To overcome this limitation, we also developed an efficient method for estimating the inverse temperature during the learning process, which we call conditional expectation matching (CEM). By combining LSB and CEM, we establish an efficient learning framework for BMs with greater expressive power than RBMs. We refer to this framework as sampler-adaptive learning (SAL). SAL opens new avenues for energy-based generative modeling beyond RBMs.
Authors: Wenzhang Du
Abstract: Retrieval-augmented models couple parametric predictors with non-parametric memories, but their use in streaming supervised learning with concept drift is not well understood. We study online classification in non-stationary environments and propose Retrieval-Augmented Memory for Online Learning (RAM-OL), a simple extension of stochastic gradient descent that maintains a small buffer of past examples. At each time step, RAM-OL retrieves a few nearest neighbours of the current input in the hidden representation space and updates the model jointly on the current example and the retrieved neighbours. We compare a naive replay variant with a gated replay variant that constrains neighbours using a time window, similarity thresholds, and gradient reweighting, in order to balance fast reuse of relevant past data against robustness to outdated regimes. From a theoretical perspective, we interpret RAM-OL under a bounded drift model and discuss how retrieval can reduce adaptation cost and improve regret constants when patterns recur over time. Empirically, we instantiate RAM-OL on a simple online multilayer perceptron and evaluate it on three real-world data streams derived from electricity pricing, electricity load, and airline delay data. On strongly and periodically drifting streams, RAM-OL improves prequential accuracy by up to about seven percentage points and greatly reduces variance across random seeds, while on a noisy airline stream the gated variant closely matches the purely online baseline. These results show that retrieval-augmented memory is a practical and robust tool for online learning under concept drift.
Authors: Sai Siddharth Nalamalpu, Kaining Yuan, Aiden Zhou, Eugene Pinsky
Abstract: The Massachusetts Bay Transportation Authority (MBTA) is the main public transit provider in Boston, operating multiple means of transport, including trains, subways, and buses. However, the system often faces delays and fluctuations in ridership volume, which negatively affect efficiency and passenger satisfaction. To further understand this phenomenon, this paper compares the performance of existing and unique methods to determine the best approach in predicting gated station entries in the subway system (a proxy for subway usage) and the number of delays in the overall MBTA system. To do so, this research considers factors that tend to affect public transportation, such as day of week, season, pressure, wind speed, average temperature, and precipitation. This paper evaluates the performance of 10 statistical and machine learning models on predicting next-day subway usage. On predicting delay count, the number of models is extended to 11 per day by introducing a self-exciting point process model, representing a unique application of a point-process framework for MBTA delay modeling. This research involves experimenting with the selective inclusion of features to determine feature importance, testing model accuracy via Root Mean Squared Error (RMSE). Remarkably, it is found that providing either day of week or season data has a more substantial benefit to predictive accuracy compared to weather data; in fact, providing weather data generally worsens performance, suggesting a tendency of models to overfit.
Authors: Zhendong Tan, Xingjun Zhang, Chaoyi Hu, Junjie Peng, Kun Xia
Abstract: Growing demands from tasks like code generation, deep reasoning, and long-document understanding have made long-context generation a crucial capability for large language models (LLMs). Speculative decoding is one of the most direct and effective approaches for accelerating generation. It follows a draft-verify paradigm, where a lightweight draft model proposes several candidate tokens and the target model verifies them. However, we find that as the context length grows, verification becomes the dominant bottleneck. To further accelerate speculative decoding in long-context generation, we introduce SpecPV, a self-speculative decoding approach that performs fast verification using partial key-value states (KV) and periodically applies full verification to eliminate accumulated errors. We validate SpecPV across multiple long-context benchmarks and models, including LLaMA-3.1-8B-Instruct and Qwen3-series. Experimental results show that SpecPV achieves up to 6x decoding speedup over standard autoregressive decoding with minor degradation.
Authors: Nan Qiao, Sheng Yue, Ju Ren, Yaoxue Zhang
Abstract: Offline Federated Reinforcement Learning (FRL), a marriage of federated learning and offline reinforcement learning, has attracted increasing interest recently. Albeit with some advancement, we find that the performance of most existing offline FRL methods drops dramatically when provided with mixed-quality data, that is, the logging behaviors (offline data) are collected by policies with varying qualities across clients. To overcome this limitation, this paper introduces a new vote-based offline FRL framework, named FOVA. It exploits a \emph{vote mechanism} to identify high-return actions during local policy evaluation, alleviating the negative effect of low-quality behaviors from diverse local learning policies. Besides, building on advantage-weighted regression (AWR), we construct consistent local and global training objectives, significantly enhancing the efficiency and stability of FOVA. Further, we conduct an extensive theoretical analysis and rigorously show that the policy learned by FOVA enjoys strict policy improvement over the behavioral policy. Extensive experiments corroborate the significant performance gains of our proposed algorithm over existing baselines on widely used benchmarks.
Authors: Jonathan Baxter, Peter L. Bartlett
Abstract: This paper discusses theoretical and experimental aspects of gradient-based approaches to the direct optimization of policy performance in controlled POMDPs. We introduce GPOMDP, a REINFORCE-like algorithm for estimating an approximation to the gradient of the average reward as a function of the parameters of a stochastic policy. The algorithm's chief advantages are that it requires only a single sample path of the underlying Markov chain, it uses only one free parameter $\beta\in [0,1)$, which has a natural interpretation in terms of bias-variance trade-off, and it requires no knowledge of the underlying state. We prove convergence of GPOMDP and show how the gradient estimates produced by GPOMDP can be used in a conjugate-gradient procedure to find local optima of the average reward.
Authors: Chuhan Xie
Abstract: This paper studies the problem of risk-sensitive reinforcement learning (RSRL) in continuous time, where the environment is characterized by a controllable stochastic differential equation (SDE) and the objective is a potentially nonlinear functional of cumulative rewards. We prove that when the functional is an optimized certainty equivalent (OCE), the optimal policy is Markovian with respect to an augmented environment. We also propose \textit{CT-RS-q}, a risk-sensitive q-learning algorithm based on a novel martingale characterization approach. Finally, we run a simulation study on a dynamic portfolio selection problem and illustrate the effectiveness of our algorithm.
Authors: Hongxiang Liu, Zhifang Deng, Tong Pu, Shengli Lu
Abstract: Transformers, composed of QKV generation, attention computation, and FFNs, have become the dominant model across various domains due to their outstanding performance. However, their high computational cost hinders efficient hardware deployment. Sparsity offers a promising solution, yet most existing accelerators exploit only intra-row sparsity in attention, while few consider inter-row sparsity. Approaches leveraging inter-row sparsity often rely on costly global similarity estimation, which diminishes the acceleration benefits of sparsity, and typically apply sparsity to only one or two transformer components. Through careful analysis of the attention distribution and computation flow, we observe that local similarity allows end-to-end sparse acceleration with lower computational overhead. Motivated by this observation, we propose ESACT, an end-to-end sparse accelerator for compute-intensive Transformers. ESACT centers on the Sparsity Prediction with Local Similarity (SPLS) mechanism, which leverages HLog quantization to accurately predict local attention sparsity prior to QK generation, achieving efficient sparsity across all transformer components. To support efficient hardware realization, we introduce three architectural innovations. Experimental results on 26 benchmarks demonstrate that SPLS reduces total computation by 52.03% with less than 1% accuracy loss. ESACT achieves an end-to-end energy efficiency of 3.29 TOPS/W, and improves attention-level energy efficiency by 2.95x and 2.26x over SOTA attention accelerators SpAtten and Sanger, respectively.
Authors: Oshada Jayasinghe, Farhana Choudhury, Egemen Tanin, Shanika Karunasekera
Abstract: With increased travelling needs more than ever, traffic congestion has become a major concern in most urban areas. Allocating spaces for on-street parking, further hinders traffic flow, by limiting the effective road width available for driving. With the advancement of vehicle-to-infrastructure connectivity technologies, we explore how the impact of on-street parking on traffic congestion could be minimized, by dynamically configuring on-street parking spaces. Towards that end, we formulate dynamic on-street parking space configuration as an optimization problem, and we follow a data driven approach, considering the nature of our problem. Our proposed solution comprises a two-layer multi agent reinforcement learning based framework, which is inherently scalable to large road networks. The lane level agents are responsible for deciding the optimal parking space configuration for each lane, and we introduce a novel Deep Q-learning architecture which effectively utilizes long short term memory networks and graph attention networks to capture the spatio-temporal correlations evident in the given problem. The block level agents control the actions of the lane level agents and maintain a sufficient level of parking around the block. We conduct a set of comprehensive experiments using SUMO, on both synthetic data as well as real-world data from the city of Melbourne. Our experiments show that the proposed framework could reduce the average travel time loss of vehicles significantly, reaching upto 47%, with a negligible increase in the walking distance for parking.
Authors: Yizhou Zhang, Lun Du
Abstract: Large-scale neural models are increasingly trained with data pruning, synthetic data generation, cross-model distillation, reinforcement learning from human feedback (RLHF), and difficulty-based sampling. While several of these data-centric strategies reliably improve training efficiency and downstream performance, others fail to provide meaningful gains -- most notably self-generated synthetic data, which often increases dataset volume without enhancing model capability. We formalize data curation as reweighting the sampling distribution and map its effect onto the eigenstructure of the data-induced operator. Our first main result shows that \textbf{static pruning induces a bounded operator and therefore cannot change the spectral tail exponent}; it provides at most finite-region improvements and cannot alter asymptotic neural scaling. Our second result analyzes \textbf{time-dependent data curation}, showing that an ideal oracle capable of tracking spectral residuals and continuously re-normalizing the tail can provably accelerate learning -- although practical systems can only approximate this behavior.
Authors: Zhongjian Qiao, Rui Yang, Jiafei Lyu, Chenjia Bai, Xiu Li, Zhuoran Yang, Siyang Gao, Shuang Qiu
Abstract: Cross-Domain Offline Reinforcement Learning aims to train an agent deployed in the target environment, leveraging both a limited target domain dataset and a source domain dataset with (possibly) sufficient data coverage. Due to the underlying dynamics misalignment between the source and target domain, simply merging the data from two datasets may incur inferior performance. Recent advances address this issue by selectively sharing source domain samples that exhibit dynamics alignment with the target domain. However, these approaches focus solely on dynamics alignment and overlook \textit{value alignment}, i.e., selecting high-quality, high-value samples from the source domain. In this paper, we first demonstrate that both dynamics alignment and value alignment are essential for policy learning, by examining the limitations of the current theoretical framework for cross-domain RL and establishing a concrete sub-optimality gap of a policy trained on the source domain and evaluated on the target domain. Motivated by the theoretical insights, we propose to selectively share those source domain samples with both high dynamics and value alignment and present our \textbf{\underline{D}}ynamics- and \textbf{\underline{V}}alue-aligned \textbf{\underline{D}}ata \textbf{\underline{F}}iltering (DVDF) method. We design a range of dynamics shift settings, including kinematic and morphology shifts, and evaluate DVDF on various tasks and datasets, as well as in challenging extremely low-data settings where the target domain dataset contains only 5,000 transitions. Extensive experiments demonstrate that DVDF consistently outperforms prior strong baselines and delivers exceptional performance across multiple tasks and datasets.
Authors: Tsimur Hadeliya, Mohammad Ali Jauhar, Nidhi Sakpal, Diogo Cruz
Abstract: Solving complex or long-horizon problems often requires large language models (LLMs) to use external tools and operate over a significantly longer context window. New LLMs enable longer context windows and support tool calling capabilities. Prior works have focused mainly on evaluation of LLMs on long-context prompts, leaving agentic setup relatively unexplored, both from capability and safety perspectives. Our work addresses this gap. We find that LLM agents could be sensitive to length, type, and placement of the context, exhibiting unexpected and inconsistent shifts in task performance and in refusals to execute harmful requests. Models with 1M-2M token context windows show severe degradation already at 100K tokens, with performance drops exceeding 50\% for both benign and harmful tasks. Refusal rates shift unpredictably: GPT-4.1-nano increases from $\sim$5\% to $\sim$40\% while Grok 4 Fast decreases from $\sim$80\% to $\sim$10\% at 200K tokens. Our work shows potential safety issues with agents operating on longer context and opens additional questions on the current metrics and paradigm for evaluating LLM agent safety on long multi-step tasks. In particular, our results on LLM agents reveal a notable divergence in both capability and safety performance compared to prior evaluations of LLMs on similar criteria.
Authors: Xingwang Li (C. Z.), Mengyun Chen (C. Z.), Jiamou Liu (C. Z.), Sijie Wang (C. Z.), Shuanggen Jin (C. Z.), Jafet C. M. Andersson (C. Z.), Jonas Olsson (C. Z.), Remco (C. Z.), van de Beek, Hai Victor Habi, Congzheng Han
Abstract: In the face of accelerating global urbanization and the increasing frequency of extreme weather events, highresolution urban rainfall monitoring is crucial for building resilient smart cities. Commercial Microwave Links (CMLs) are an emerging data source with great potential for this task.While traditional rainfall retrieval from CMLs relies on physicsbased models, these often struggle with real-world complexities like signal noise and nonlinear attenuation. To address these limitations, this paper proposes a novel hybrid deep learning architecture based on the Transformer and a Bidirectional Gated Recurrent Unit (BiGRU), which we name TabGRU. This design synergistically captures both long-term dependencies and local sequential features in the CML signal data. The model is further enhanced by a learnable positional embedding and an attention pooling mechanism to improve its dynamic feature extraction and generalization capabilities. The model was validated on a public benchmark dataset from Gothenburg, Sweden (June-September 2015). The evaluation used 12 sub-links from two rain gauges (Torp and Barl) over a test period (August 22-31) covering approximately 10 distinct rainfall events. The proposed TabGRU model demonstrated consistent advantages, outperforming deep learning baselines and achieving high coefficients of determination (R2) at both the Torp site (0.91) and the Barl site (0.96). Furthermore, compared to the physics-based approach, TabGRU maintained higher accuracy and was particularly effective in mitigating the significant overestimation problem observed in the PL model during peak rainfall events. This evaluation confirms that the TabGRU model can effectively overcome the limitations of traditional methods, providing a robust and accurate solution for CML-based urban rainfall monitoring under the tested conditions.
Authors: Zhongjian Qiao, Rui Yang, Jiafei Lyu, Xiu Li, Zhongxiang Dai, Zhuoran Yang, Siyang Gao, Shuang Qiu
Abstract: Single-domain offline reinforcement learning (RL) often suffers from limited data coverage, while cross-domain offline RL handles this issue by leveraging additional data from other domains with dynamics shifts. However, existing studies primarily focus on train-time robustness (handling dynamics shifts from training data), neglecting the test-time robustness against dynamics perturbations when deployed in practical scenarios. In this paper, we investigate dual (both train-time and test-time) robustness against dynamics shifts in cross-domain offline RL. We first empirically show that the policy trained with cross-domain offline RL exhibits fragility under dynamics perturbations during evaluation, particularly when target domain data is limited. To address this, we introduce a novel robust cross-domain Bellman (RCB) operator, which enhances test-time robustness against dynamics perturbations while staying conservative to the out-of-distribution dynamics transitions, thus guaranteeing the train-time robustness. To further counteract potential value overestimation or underestimation caused by the RCB operator, we introduce two techniques, the dynamic value penalty and the Huber loss, into our framework, resulting in the practical \textbf{D}ual-\textbf{RO}bust \textbf{C}ross-domain \textbf{O}ffline RL (DROCO) algorithm. Extensive empirical results across various dynamics shift scenarios show that DROCO outperforms strong baselines and exhibits enhanced robustness to dynamics perturbations.
Authors: Mithra D K
Abstract: I present an application of established machine learning techniques to NHANES health survey data for predicting diabetes status. I compare baseline models (logistic regression, random forest, XGBoost) with a hybrid approach that uses an XGBoost feature encoder and a lightweight multilayer perceptron (MLP) head. Experiments show the hybrid model attains improved AUC and balanced accuracy compared to baselines on the processed NHANES subset. I release code and reproducible scripts to encourage replication.
Authors: Zihao Zhao, Kai-Chia Mo, Shing-Hei Ho, Brandon Amos, Kai Wang
Abstract: Differentiable optimization layers enable learning systems to make decisions by solving embedded optimization problems. However, computing gradients via implicit differentiation requires solving a linear system with Hessian terms, which is both compute- and memory-intensive. To address this challenge, we propose a novel algorithm that computes the gradient using only first-order information. The key insight is to rewrite the differentiable optimization as a bilevel optimization problem and leverage recent advances in bilevel methods. Specifically, we introduce an active-set Lagrangian hypergradient oracle that avoids Hessian evaluations and provides finite-time, non-asymptotic approximation guarantees. We show that an approximate hypergradient can be computed using only first-order information in $\tilde{\oo}(1)$ time, leading to an overall complexity of $\tilde{\oo}(\delta^{-1}\epsilon^{-3})$ for constrained bilevel optimization, which matches the best known rate for non-smooth non-convex optimization. Furthermore, we release an open-source Python library that can be easily adapted from existing solvers. Our code is available here: https://github.com/guaguakai/FFOLayer.
Authors: Marco Cardia, Stefano Chessa, Alessio Micheli, Antonella Giuliana Luminare, Francesca Gambineri
Abstract: The quality of water is key for the quality of agrifood sector. Water is used in agriculture for fertigation, for animal husbandry, and in the agrifood processing industry. In the context of the progressive digitalization of this sector, the automatic assessment of the quality of water is thus becoming an important asset. In this work, we present the integration of Ultraviolet-Visible (UV-Vis) spectroscopy with Machine Learning in the context of water quality assessment aiming at ensuring water safety and the compliance of water regulation. Furthermore, we emphasize the importance of model interpretability by employing SHapley Additive exPlanations (SHAP) to understand the contribution of absorbance at different wavelengths to the predictions. Our approach demonstrates the potential for rapid, accurate, and interpretable assessment of key water quality parameters.
Authors: Krishnendu S. Tharakan, Carlo Fischione
Abstract: Wireless connectivity promises to unshackle virtual reality (VR) experiences, allowing users to engage from anywhere, anytime. However, delivering seamless, high-quality, real-time VR video wirelessly is challenging due to the stringent quality of experience requirements, low latency constraints, and limited VR device capabilities. This paper addresses these challenges by introducing a novel decentralized multi task fair federated learning (DMTFL) based caching that caches and prefetches each VR user's field of view (FOV) at base stations (BSs) based on the caching strategies tailored to each BS. In federated learning (FL) in its naive form, often biases toward certain users, and a single global model fails to capture the statistical heterogeneity across users and BSs. In contrast, the proposed DMTFL algorithm personalizes content delivery by learning individual caching models at each BS. These models are further optimized to perform well under any target distribution, while providing theoretical guarantees via Rademacher complexity and a probably approximately correct (PAC) bound on the loss. Using a realistic VR head-tracking dataset, our simulations demonstrate the superiority of our proposed DMTFL algorithm compared to baseline algorithms.
Authors: Vishnu Sarukkai, Asanshay Gupta, James Hong, Micha\"el Gharbi, Kayvon Fatahalian
Abstract: The world currently has an abundance of ideas for how to use new LLM agents, and developers seek to rapidly prototype and test new agentic designs. However, executing agents at scale using high-capacity LLMs incurs high inference costs. We propose a simple method for reducing LLM agent inference costs without incurring the development friction costs associated with LLM fine-tuning (long training cycles, optimization hyperparameter tweaking loops) or manual prompt engineering (laborious trial and error). Most importantly, we introduce $\textit{in-context distillation}$, which adapts the idea of knowledge distillation (training a low cost-student model to mimic a high-cost teacher) to an in-context learning setting. Our approach retrieves relevant teacher demonstrations at each agent step and provides them to the student as in-context examples, enabling the student to imitate teacher behavior on-the-fly. We combine in-context distillation with the established idea of $\textit{self-consistency cascades}$ to know when the trust the student. This adaptive strategy realizes the cost benefits of model specialization while preserving the productivity of working with frozen models. On the multi-step embodied reasoning benchmark ALFWorld, our method matches teacher-level accuracy at $\textbf{2.5$\times$ lower cost}$, reducing per-episode costs from \$0.059 to \$0.024. The upfront demonstration cost amortizes after just 843 episodes, yielding cumulative savings exceeding \$34,900 at deployment scale (1M episodes). On AppWorld, a complex agent benchmark requiring multi-step API workflows, we shift the Pareto frontier by achieving a $\textbf{2$\times$ cost reduction}$ at iso-accuracy. By reducing operational costs while maintaining rapid experimentation cycles with frozen models, our approach makes advanced agentic systems economically viable for a broader range of applications.
Authors: Albert Saiapin, Kim Batselier
Abstract: Many approximations were suggested to circumvent the cubic complexity of kernel-based algorithms, allowing their application to large-scale datasets. One strategy is to consider the primal formulation of the learning problem by mapping the data to a higher-dimensional space using tensor-product structured polynomial and Fourier features. The curse of dimensionality due to these tensor-product features was effectively solved by a tensor network reparameterization of the model parameters. However, another important aspect of model training - identifying optimal feature hyperparameters - has not been addressed and is typically handled using the standard cross-validation approach. In this paper, we introduce the Feature Learning (FL) model, which addresses this issue by representing tensor-product features as a learnable Canonical Polyadic Decomposition (CPD). By leveraging this CPD structure, we efficiently learn the hyperparameters associated with different features alongside the model parameters using an Alternating Least Squares (ALS) optimization method. We prove the effectiveness of the FL model through experiments on real data of various dimensionality and scale. The results show that the FL model can be consistently trained 3-5 times faster than and have the prediction quality on par with a standard cross-validated model.
Authors: Songqiao Su, Xiaofei Sun, Xiaoya Li, Albert Wang, Jiwei Li, Chris Shum
Abstract: In this paper, we propose CUDA-L2, a system that combines large language models (LLMs) and reinforcement learning (RL) to automatically optimize Half-precision General Matrix Multiply (HGEMM) CUDA kernels. Using CUDA execution speed as the RL reward, CUDA-L2 automatically optimizes HGEMM kernels across 1,000 configurations. CUDA-L2 systematically outperforms major matmul baselines to date, from the widely-used {\it torch.matmul} to state-of-the-art Nvidia's closed-source libraries, i.e., {\it cuBLAS}, {\it cuBLASLt}. In offline mode, where kernels are executed consecutively without time intervals, CUDA-L2 yields +22.0\% over {\it torch.matmul} on average; +19.2\% over {\it cuBLAS} using the optimal layout configuration (normal-normal NN and transposed-normal TN); +16.8\% over {\it cuBLASLt-heuristic}, which queries {\it cuBLASLt} library and selects the algorithm based on the heuristic's suggestion; and +11.4\% over the most competitive {\it cuBLASLt-AutoTuning} model, which selects the fastest algorithm from up to 100 candidates from {\it cuBLASLt}'s suggestions. In server mode, where kernels are executed at random intervals simulating real-time inference, the speedups further increase to +28.7\%, +26.0\%, +22.4\%, and +15.9\% for {\it torch.matmul}, {\it cuBLAS}, {\it cuBLASLt-heuristic}, and {\it cuBLASLt-AutoTuning} respectively. CUDA-L2 shows that even the most performance-critical, heavily-optimized kernels like HGEMM can be improved through LLM-guided RL automation by systematically exploring configuration spaces at scales impractical for humans. Project and code can be found at github.com/deepreinforce-ai/CUDA-L2
Authors: Chubin Zhang, Zhenglin Wan, Feng Chen, Xingrui Yu, Ivor Tsang, Bo An
Abstract: Reinforcement learning (RL) faces a persistent tension: policies that are stable to optimize are often too simple to represent the multimodal action distributions needed for complex control. Gaussian policies provide tractable likelihoods and smooth gradients, but their unimodal form limits expressiveness. Conversely, generative policies based on diffusion or flow matching can model rich multimodal behaviors; however, in online RL, they are frequently unstable due to intractable likelihoods and noisy gradients propagating through deep sampling chains. We address this tension with a key structural principle: decoupling optimization from generation. Building on this insight, we introduce GoRL (Generative Online Reinforcement Learning), a framework that optimizes a tractable latent policy while utilizing a conditional generative decoder to synthesize actions. A two-timescale update schedule enables the latent policy to learn stably while the decoder steadily increases expressiveness, without requiring tractable action likelihoods. Across a range of continuous-control tasks, GoRL consistently outperforms both Gaussian policies and recent generative-policy baselines. Notably, on the HopperStand task, it reaches a normalized return above 870, more than 3 times that of the strongest baseline. These results demonstrate that separating optimization from generation provides a practical path to policies that are both stable and highly expressive.
Authors: Wenhao Sha, Tienchong Chang
Abstract: Heat transfer in semiconductor devices is dominated by chip and substrate assemblies, where heat generated within a finite chip layer dissipates into a semi-infinite substrate with much higher thermophysical properties. This mismatch produces steep interfacial temperature gradients, making the transient thermal response highly sensitive to the interface. Conventional numerical solvers require excessive discretization to resolve these dynamics, while physics-informed neural networks (PINNs) often exhibit unstable convergence and loss of physical consistency near the material interface. To address these challenges, we introduce HeatTransFormer, a physics-guided Transformer architecture for interface-dominated diffusion problems. The framework integrates physically informed spatiotemporal sampling, a Laplace-based activation emulating analytical diffusion solutions, and a mask-free attention mechanism supporting bidirectional spatiotemporal coupling. These components enable the model to resolve steep gradients, maintain physical consistency, and remain stable where PINNs typically fail. HeatTransFormer produces coherent temperature fields across the interface when applied to a finite layer and semi-infinite substrate configuration. Coupled with a physics-constrained inverse strategy, it further enables reliable identification of three unknown thermal properties simultaneously using only external measurements. Overall, this work demonstrates that physics-guided Transformer architectures provide a unified framework for forward and inverse modeling in interface-dominated thermal systems.
Authors: Seline J. S. de Rooij, Borb\'ala Hunyadi
Abstract: Transfer learning aims to optimize performance in a target task by learning from a related source problem. In this work, we propose an efficient transfer learning method using a tensor kernel machine. Our method takes inspiration from the adaptive SVM and hence transfers 'knowledge' from the source to the 'adapted' model via regularization. The main advantage of using tensor kernel machines is that they leverage low-rank tensor networks to learn a compact non-linear model in the primal domain. This allows for a more efficient adaptation without adding more parameters to the model. To demonstrate the effectiveness of our approach, we apply the adaptive tensor kernel machine (Adapt-TKM) to seizure detection on behind-the-ear EEG. By personalizing patient-independent models with a small amount of patient-specific data, the patient-adapted model (which utilizes the Adapt-TKM), achieves better performance compared to the patient-independent and fully patient-specific models. Notably, it is able to do so while requiring around 100 times fewer parameters than the adaptive SVM model, leading to a correspondingly faster inference speed. This makes the Adapt-TKM especially useful for resource-constrained wearable devices.
Authors: Zhengcheng Wang, Zichuan Lin, Yijun Yang, Haobo Fu, Deheng Ye
Abstract: Existing Vision-Language Navigation (VLN) agents based on Large Vision-Language Models (LVLMs) often suffer from perception errors, reasoning errors, and planning errors, which significantly hinder their navigation performance. To address these limitations, a novel VLN agent framework, named SeeNav-Agent, is proposed in this work. First, to reduce perception hallucinations of the visual module of the VLN agent, a dual-view Visual Prompt (VP) technique is introduced in the input space, which can also improve the agent's understanding of current spatial states. Subsequently, a novel step-level Reinforcement Fine-Tuning (RFT) method, Step Reward Group Policy Optimization (SRGPO), is designed for the post-training of VLN agents. In SRGPO, we first define verifiable process rewards for the navigation task, and then perform efficient step-level advantage estimation by randomly grouping different navigation steps. SRGPO provides dense reward signals for the reinforcement learning process of the VLN agent and enhances its planning capability. Experimental results on the EmbodiedBench Navigation benchmark indicate that by introducing the zero-shot VP module, the GPT-4.1 achieves a navigation success rate of 86.7%, surpassing the current best LVLM by approximately 20 percentage points (pp). Through post-training based on SRGPO, the Qwen2.5-VL-3B model reaches a navigation success rate of 72.3%, outperforming the best existing LVLM model by 5.6 pp. Moreover, compared to RFT algorithms such as GRPO and GiGPO, the proposed SRGPO demonstrates significant improvements in training stability, convergence efficiency, and generalization capability.
Authors: Xinyue Ai, Yutong He, Albert Gu, Ruslan Salakhutdinov, J Zico Kolter, Nicholas Matthew Boffi, Max Simchowitz
Abstract: Log-likelihood evaluation enables important capabilities in generative models, including model comparison, certain fine-tuning objectives, and many downstream applications. Yet paradoxically, some of today's best generative models -- diffusion and flow-based models -- still require hundreds to thousands of neural function evaluations (NFEs) to compute a single likelihood. While recent distillation methods have successfully accelerated sampling to just a few steps, they achieve this at the cost of likelihood tractability: existing approaches either abandon likelihood computation entirely or still require expensive integration over full trajectories. We present fast flow joint distillation (F2D2), a framework that simultaneously reduces the number of NFEs required for both sampling and likelihood evaluation by two orders of magnitude. Our key insight is that in continuous normalizing flows, the coupled ODEs for sampling and likelihood are computed from a shared underlying velocity field, allowing us to jointly distill both the sampling trajectory and cumulative divergence using a single model. F2D2 is modular, compatible with existing flow-based few-step sampling models, and requires only an additional divergence prediction head. Experiments demonstrate F2D2's capability of achieving accurate log-likelihood with few-step evaluations while maintaining high sample quality, solving a long-standing computational bottleneck in flow-based generative models. As an application of our approach, we propose a lightweight self-guidance method that enables a 2-step MeanFlow model to outperform a 1024 step teacher model with only a single additional backward NFE.
Authors: Farnaz Faramarzi Lighvan, Mehrdad Asadi, Lynn Houthuys
Abstract: Multi-view learning integrates diverse representations of the same instances to improve performance. Most existing kernel-based multi-view learning methods use fusion techniques without enforcing an explicit collaboration type across views or co-regularization which limits global collaboration. We propose AW-LSSVM, an adaptive weighted LS-SVM that promotes complementary learning by an iterative global coupling to make each view focus on hard samples of others from previous iterations. Experiments demonstrate that AW-LSSVM outperforms existing kernel-based multi-view methods on most datasets, while keeping raw features isolated, making it also suitable for privacy-preserving scenarios.
Authors: Naveen George, Naoki Murata, Yuhta Takida, Konda Reddy Mopuri, Yuki Mitsufuji
Abstract: The recent rapid growth of visual generative models trained on vast web-scale datasets has created significant tension with data privacy regulations and copyright laws, such as GDPR's ``Right to be Forgotten.'' This necessitates machine unlearning (MU) to remove specific concepts without the prohibitive cost of retraining. However, existing MU techniques are fundamentally ill-equipped for real-world scenarios where deletion requests arrive sequentially, a setting known as continual unlearning (CUL). Naively applying one-shot methods in a continual setting triggers a stability crisis, leading to a cascade of degradation characterized by retention collapse, compounding collateral damage to related concepts, and a sharp decline in generative quality. To address this critical challenge, we introduce a novel generative distillation based continual unlearning framework that ensures targeted and stable unlearning under sequences of deletion requests. By reframing each unlearning step as a multi-objective, teacher-student distillation process, the framework leverages principles from continual learning to maintain model integrity. Experiments on a 10-step sequential benchmark demonstrate that our method unlearns forget concepts with better fidelity and achieves this without significant interference to the performance on retain concepts or the overall image quality, substantially outperforming baselines. This framework provides a viable pathway for the responsible deployment and maintenance of large-scale generative models, enabling industries to comply with ongoing data removal requests in a practical and effective manner.
Authors: Haozhuo Zheng, Cheng Wang, Yang Liu
Abstract: The de novo generation of molecules with desirable properties is a critical challenge, where diffusion models are computationally intensive and autoregressive models struggle with error propagation. In this work, we introduce the Graph VQ-Transformer (GVT), a two-stage generative framework that achieves both high accuracy and efficiency. The core of our approach is a novel Graph Vector Quantized Variational Autoencoder (VQ-VAE) that compresses molecular graphs into high-fidelity discrete latent sequences. By synergistically combining a Graph Transformer with canonical Reverse Cuthill-McKee (RCM) node ordering and Rotary Positional Embeddings (RoPE), our VQ-VAE achieves near-perfect reconstruction rates. An autoregressive Transformer is then trained on these discrete latents, effectively converting graph generation into a well-structured sequence modeling problem. Crucially, this mapping of complex graphs to high-fidelity discrete sequences bridges molecular design with the powerful paradigm of large-scale sequence modeling, unlocking potential synergies with Large Language Models (LLMs). Extensive experiments show that GVT achieves state-of-the-art or highly competitive performance across major benchmarks like ZINC250k, MOSES, and GuacaMol, and notably outperforms leading diffusion models on key distribution similarity metrics such as FCD and KL Divergence. With its superior performance, efficiency, and architectural novelty, GVT not only presents a compelling alternative to diffusion models but also establishes a strong new baseline for the field, paving the way for future research in discrete latent-space molecular generation.
Authors: Senrong Xu, Tianyu Wang, Zenan Li, Yuan Yao, Taolue Chen, Feng Xu, Xiaoxing Ma
Abstract: Conformal prediction (CP) provides a comprehensive framework to produce statistically rigorous uncertainty sets for black-box machine learning models. To further improve the efficiency of CP, conformal correction is proposed to fine-tune or wrap the base model with an extra module using a conformal-aware inefficiency loss. In this work, we empirically and theoretically identify a trade-off between the CP efficiency and the entropy of model prediction. We then propose an entropy-constrained conformal correction method, exploring a better Pareto optimum between efficiency and entropy. Extensive experimental results on both computer vision and graph datasets demonstrate the efficacy of the proposed method. For instance, it can significantly improve the efficiency of state-of-the-art CP methods by up to 34.4%, given an entropy threshold.
Authors: Junpeng Wu, Pinheng Zong
Abstract: Graph-based Anomaly Detection models have gained widespread adoption in recent years, identifying suspicious nodes by aggregating neighborhood information. However, most existing studies overlook the pervasive issues of missing and adversarially obscured node attributes, which can undermine aggregation stability and prediction reliability. To mitigate this, we propose FGC-Comp, a lightweight, classifier-agnostic, and deployment-friendly attribute completion module-designed to enhance neighborhood aggregation under incomplete attributes. We partition each node's neighbors into three label-based groups, apply group-specific transforms to the labeled groups while a node-conditioned gate handles unknowns, fuse messages via residual connections, and train end-to-end with a binary classification objective to improve aggregation stability and prediction reliability under missing attributes. Experiments on two real-world fraud datasets validate the effectiveness of the approach with negligible computational overhead.
Authors: Matteo Tolloso, Davide Bacciu
Abstract: Uncertainty quantification is essential for deploying reliable Graph Neural Networks (GNNs), where existing approaches primarily rely on Bayesian inference or ensembles. In this paper, we introduce the first credal graph neural networks (CGNNs), which extend credal learning to the graph domain by training GNNs to output set-valued predictions in the form of credal sets. To account for the distinctive nature of message passing in GNNs, we develop a complementary approach to credal learning that leverages different aspects of layer-wise information propagation. We assess our approach on uncertainty quantification in node classification under out-of-distribution conditions. Our analysis highlights the critical role of the graph homophily assumption in shaping the effectiveness of uncertainty estimates. Extensive experiments demonstrate that CGNNs deliver more reliable representations of epistemic uncertainty and achieve state-of-the-art performance under distributional shift on heterophilic graphs.
Authors: Waleed El-Geresy, Deniz G\"und\"uz
Abstract: We propose the use of adversarial wireless jamming to regularise the latent space of an autoencoder to match a diagonal Gaussian distribution. We consider the minimisation of a mean squared error distortion, where a jammer attempts to disrupt the recovery of a Gaussian source encoded and transmitted over the adversarial channel. A straightforward consequence of existing theoretical results is the fact that the saddle point of a minimax game - involving such an encoder, its corresponding decoder, and an adversarial jammer - consists of diagonal Gaussian noise output by the jammer. We use this result as inspiration for a novel approach to distribution matching in the latent space, utilising jamming as an auxiliary objective to encourage the aggregated latent posterior to match a diagonal Gaussian distribution. Using this new technique, we achieve distribution matching comparable to standard variational autoencoders and to Wasserstein autoencoders. This approach can also be generalised to other latent distributions.
Authors: Anton Emelyanov, Sergei Kudriashov, Alena Fenogenova
Abstract: Membership Inference Attacks (MIAs) aim to determine whether a specific data point was included in the training set of a target model. Although there are have been numerous methods developed for detecting data contamination in large language models (LLMs), their performance on multimodal LLMs (MLLMs) falls short due to the instabilities introduced through multimodal component adaptation and possible distribution shifts across multiple inputs. In this work, we investigate multimodal membership inference and address two issues: first, by identifying distribution shifts in the existing datasets, and second, by releasing an extended baseline pipeline to detect them. We also generalize the perturbation-based membership inference methods to MLLMs and release \textbf{FiMMIA} -- a modular \textbf{F}ramework for \textbf{M}ultimodal \textbf{MIA}.\footnote{The source code and framework have been made publicly available under the MIT license via \href{https://github.com/ai-forever/data_leakage_detect}{link}.The video demonstration is available on \href{https://youtu.be/a9L4-H80aSg}{YouTube}.} Our approach trains a neural network to analyze the target model's behavior on perturbed inputs, capturing distributional differences between members and non-members. Comprehensive evaluations on various fine-tuned multimodal models demonstrate the effectiveness of our perturbation-based membership inference attacks in multimodal domains.
URLs: https://github.com/ai-forever/data_leakage_detect, https://youtu.be/a9L4-H80aSg
Authors: Haoming Liu, Jinnuo Liu, Yanhao Li, Liuyang Bai, Yunkai Ji, Yuanhe Guo, Shenji Wan, Hongyi Wen
Abstract: Flow-based diffusion models have emerged as a leading paradigm for training generative models across images and videos. However, their memorization-generalization behavior remains poorly understood. In this work, we revisit the flow matching (FM) objective and study its marginal velocity field, which admits a closed-form expression, allowing exact computation of the oracle FM target. Analyzing this oracle velocity field reveals that flow-based diffusion models inherently formulate a two-stage training target: an early stage guided by a mixture of data modes, and a later stage dominated by the nearest data sample. The two-stage objective leads to distinct learning behaviors: the early navigation stage generalizes across data modes to form global layouts, whereas the later refinement stage increasingly memorizes fine-grained details. Leveraging these insights, we explain the effectiveness of practical techniques such as timestep-shifted schedules, classifier-free guidance intervals, and latent space design choices. Our study deepens the understanding of diffusion model training dynamics and offers principles for guiding future architectural and algorithmic improvements.
Authors: Ihab Ahmed, Denis Krompa{\ss}, Cheng Feng, Volker Tresp
Abstract: We investigate input normalization methods for Time-Series Foundation Models (TSFMs). While normalization is well-studied in dataset-specific time-series models, it remains overlooked in TSFMs where generalization is critical. Time-series data, unlike text or images, exhibits significant scale variation across domains and channels, coupled with non-stationarity, can undermine TSFM performance regardless of architectural complexity. Through systematic evaluation across four architecturally diverse TSFMs, we empirically establish REVIN as the most efficient approach, reducing zero-shot MASE by 89\% relative to an un-normalized baseline and by 44\% versus other normalization methods, while matching the best in-domain accuracy (0.84 MASE) without any dataset-level preprocessing -- yielding the highest accuracy-efficiency trade-off. Yet its effect utilization depends on architectural design choices and optimization objective, particularly with respect to training loss scale sensitivity and model type (probabilistic, point-forecast, or LLM-based models).
Authors: Miko{\l}aj Sacha, Hammad Jafri, Mattie Terzolo, Ayan Sinha, Andrew Rabinovich
Abstract: Recommending matches in a text-rich, dynamic two-sided marketplace presents unique challenges due to evolving content and interaction graphs. We introduce GraphMatch, a new large-scale recommendation framework that fuses pre-trained language models with graph neural networks to overcome these challenges. Unlike prior approaches centered on standalone models, GraphMatch is a comprehensive recipe built on powerful text encoders and GNNs working in tandem. It employs adversarial negative sampling alongside point-in-time subgraph training to learn representations that capture both the fine-grained semantics of evolving text and the time-sensitive structure of the graph. We evaluated extensively on interaction data from Upwork, a leading labor marketplace, at large scale, and discuss our approach towards low-latency inference suitable for real-time use. In our experiments, GraphMatch outperforms language-only and graph-only baselines on matching tasks while being efficient at runtime. These results demonstrate that unifying language and graph representations yields a highly effective solution to text-rich, dynamic two-sided recommendations, bridging the gap between powerful pretrained LMs and large-scale graphs in practice.
Authors: Shuyuan Wu, Feifei Wang, Yuan Gao, Hansheng Wang
Abstract: In decentralized federated learning (DFL), the presence of abnormal clients, often caused by noisy or poisoned data, can significantly disrupt the learning process and degrade the overall robustness of the model. Previous methods on this issue often require a sufficiently large number of normal neighboring clients or prior knowledge of reliable clients, which reduces the practical applicability of DFL. To address these limitations, we develop here a novel adaptive DFL (aDFL) approach for robust estimation. The key idea is to adaptively adjust the learning rates of clients. By assigning smaller rates to suspicious clients and larger rates to normal clients, aDFL mitigates the negative impact of abnormal clients on the global model in a fully adaptive way. Our theory does not put any stringent conditions on neighboring nodes and requires no prior knowledge. A rigorous convergence analysis is provided to guarantee the oracle property of aDFL. Extensive numerical experiments demonstrate the superior performance of the aDFL method.
Authors: Cristian J. Villatoro, Gianluca Geraci, Daniele E. Schiavazzi
Abstract: Outer loop tasks such as optimization, uncertainty quantification or inference can easily become intractable when the underlying high-fidelity model is computationally expensive. Similarly, data-driven architectures typically require large datasets to perform predictive tasks with sufficient accuracy. A possible approach to mitigate these challenges is the development of multi-fidelity emulators, leveraging potentially biased, inexpensive low-fidelity information while correcting and refining predictions using scarce, accurate high-fidelity data. This study investigates the performance of multi-fidelity neural emulators, neural networks designed to learn the input-to-output mapping by integrating limited high-fidelity data with abundant low-fidelity model solutions. We investigate the performance of such emulators for low and high-dimensional functions, with oscillatory character, in the presence of discontinuities, for collections of models with equal and dissimilar parametrization, and for a possibly large number of potentially corrupted low-fidelity sources. In doing so, we consider a large number of architectural, hyperparameter, and dataset configurations including networks with a different amount of spectral bias (Multi-Layered Perceptron, Siren and Kolmogorov Arnold Network), various mechanisms for coordinate encoding, exact or learnable low-fidelity information, and for varying training dataset size. We further analyze the added value of the multi-fidelity approach by conducting equivalent single-fidelity tests for each case, quantifying the performance gains achieved through fusing multiple sources of information.
Authors: Youkang Wang, Jian Wang, Rubing Chen, Tianyi Zeng, Xiao-Yong Wei, Qing Li
Abstract: Test-time policy optimization enables large language models (LLMs) to adapt to distribution shifts by leveraging feedback from self-generated rollouts. However, existing methods rely on fixed-budget majority voting to estimate rewards, incurring substantial computational redundancy. We propose Optimal Rollout Allocation for Test-time Policy Optimization (OptPO), a principled framework that adaptively allocates inference budgets. By formulating the voting process as a Bayesian sequential probability ratio test, OptPO dynamically halts sampling once the posterior confidence in a consensus answer exceeds a specified threshold. Crucially, it utilizes the retained rollouts for on-policy updates, seamlessly integrating with algorithms like PPO or GRPO without requiring ground-truth labels. Across diverse reasoning benchmarks, OptPO significantly reduces rollout overhead compared to fixed-sample baselines while preserving or improving accuracy. By unifying statistically optimal stopping with test-time learning, OptPO offers a computationally efficient paradigm for test-time adaptation. The source code will be open upon acceptance at https://open-upon-acceptance.
Authors: Feiyu Wang, Xinyu Tan, Bokai Huang, Yihao Zhang, Guoan Wang, Peizhuang Cong, Tong Yang
Abstract: Large language models (LLMs) have revolutionized artificial intelligence, yet their massive memory and computational demands necessitate aggressive quantization, increasingly pushing representations toward the theoretical limit of a single bit. While complex-valued LLMs, such as iFairy, offer a superior chance for low-bit representation compared to real-valued counterparts, they require training from scratch, preventing the utilization of the vast ecosystem of pre-trained real-valued foundation models. Here we present Fairy2i, a universal framework that transforms pre-trained real-valued layers into an equivalent widely-linear complex form, enabling extremely low-bit quantization while reusing existing checkpoints. By proving a lossless mathematical equivalence between real and widely-linear maps, we convert standard Transformers into the complex domain and employ a phase-aware quantization scheme with a highly efficient codebook of fourth roots of unity. Furthermore, we introduce a recursive residual quantization mechanism that iteratively minimizes quantization error, allowing inference to proceed via efficient multiplication-free accumulation. We demonstrate that Fairy2i restores the performance of LLaMA-2 7B at an effective 2-bit precision to levels nearly comparable with full-precision baselines, significantly outperforming state-of-the-art real-valued binary and ternary quantization methods. This work bridges the gap between the representational efficiency of complex-valued arithmetic and the practical utility of pre-trained models, paving a new way for efficient inference on commodity hardware.
Authors: Anuran Makur, Japneet Singh
Abstract: In this work, we develop a hypothesis testing framework to determine whether pairwise comparison data is generated by an underlying \emph{generalized Thurstone model} $\mathcal{T}_F$ for a given choice function $F$. While prior work has predominantly focused on parameter estimation and uncertainty quantification for such models, we address the fundamental problem of minimax hypothesis testing for $\mathcal{T}_F$ models. We formulate this testing problem by introducing a notion of separation distance between general pairwise comparison models and the class of $\mathcal{T}_F$ models. We then derive upper and lower bounds on the critical threshold for testing that depend on the topology of the observation graph. For the special case of complete observation graphs, this threshold scales as $\Theta((nk)^{-1/2})$, where $n$ is the number of agents and $k$ is the number of comparisons per pair. Furthermore, we propose a hypothesis test based on our separation distance, construct confidence intervals, establish time-uniform bounds on the probabilities of type I and II errors using reverse martingale techniques, and derive minimax lower bounds using information-theoretic methods. Finally, we validate our results through experiments on synthetic and real-world datasets.
Authors: Ziniu Zhang, Minxuan Duan, Haris N. Koutsopoulos, Hongyang R. Zhang
Abstract: We consider analyzing traffic accident patterns using both road network data and satellite images aligned to road graph nodes. Previous work for predicting accident occurrences relies primarily on road network structural features while overlooking physical and environmental information from the road surface and its surroundings. In this work, we construct a large multimodal dataset across six U.S. states, containing nine million traffic accident records from official sources, and one million high-resolution satellite images for each node of the road network. Additionally, every node is annotated with features such as the region's weather statistics and road type (e.g., residential vs. motorway), and each edge is annotated with traffic volume information (i.e., Average Annual Daily Traffic). Utilizing this dataset, we conduct a comprehensive evaluation of multimodal learning methods that integrate both visual and network embeddings. Our findings show that integrating both data modalities improves prediction accuracy, achieving an average AUROC of $90.1\%$, which is a $3.7\%$ gain over graph neural network models that only utilize graph structures. With the improved embeddings, we conduct a causal analysis based on a matching estimator to estimate the key contributing factors influencing traffic accidents. We find that accident rates rise by $24\%$ under higher precipitation, by $22\%$ on higher-speed roads such as motorways, and by $29\%$ due to seasonal patterns, after adjusting for other confounding factors. Ablation studies confirm that satellite imagery features are essential for achieving accurate prediction.
Authors: Ahmadreza Chokhachian, Matthias Katzfuss, Yu Ding
Abstract: This paper is concerned with the problem of how to speed up computation for Gaussian process models trained on autocorrelated data. The Gaussian process model is a powerful tool commonly used in nonlinear regression applications. Standard regression modeling assumes random samples and an independently, identically distributed noise. Various fast approximations that speed up Gaussian process regression work under this standard setting. But for autocorrelated data, failing to account for autocorrelation leads to a phenomenon known as temporal overfitting that deteriorates model performance on new test instances. To handle autocorrelated data, existing fast Gaussian process approximations have to be modified; one such approach is to segment the originally correlated data points into blocks in which the blocked data are de-correlated. This work explains how to make some of the existing Gaussian process approximations work with blocked data. Numerical experiments across diverse application datasets demonstrate that the proposed approaches can remarkably accelerate computation for Gaussian process regression on autocorrelated data without compromising model prediction performance.
Authors: Jennifer Zvonek, Andrew Gillette
Abstract: An implicit neural representation (INR) is a neural network that approximates a spatiotemporal function. Many memory-intensive visualization tasks, including modern 4D CT scanning methods, represent data natively as INRs. While INRs are prized for being more memory-efficient than traditional data stored on a lattice, many visualization tasks still require discretization to a regular grid. We present PruningAMR, an algorithm that builds a mesh with resolution adapted to geometric features encoded by the INR. To identify these geometric features, we use an interpolative decomposition pruning method on the weight matrices of the INR. The resulting pruned network is used to guide adaptive mesh refinement, enabling automatic mesh generation tailored to the underlying resolution of the function. Starting from a pre-trained INR--without access to its training data--we produce a variable resolution visualization with substantial memory savings.
Authors: Louis McConnell, Jieran Sun, Theo Maffei, Raphael Gottardo, Marianna Rapsomaniki
Abstract: Understanding the spatial architecture of the tumor microenvironment (TME) is critical to advance precision oncology. We present ProteinPNet, a novel framework based on prototypical part networks that discovers TME motifs from spatial proteomics data. Unlike traditional post-hoc explanability models, ProteinPNet directly learns discriminative, interpretable, faithful spatial prototypes through supervised training. We validate our approach on synthetic datasets with ground truth motifs, and further test it on a real-world lung cancer spatial proteomics dataset. ProteinPNet consistently identifies biologically meaningful prototypes aligned with different tumor subtypes. Through graphical and morphological analyses, we show that these prototypes capture interpretable features pointing to differences in immune infiltration and tissue modularity. Our results highlight the potential of prototype-based learning to reveal interpretable spatial biomarkers within the TME, with implications for mechanistic discovery in spatial omics.
Authors: Hamid Dadkhahi, Firas Trabelsi, Parker Riley, Juraj Juraska, Mehdi Mirzazadeh
Abstract: Thinking Large Language Models (LLMs) used as judges for pairwise preferences remain noisy at the single-sample level, and common aggregation rules (majority vote, soft self-consistency, or instruction-based self-aggregation) are inconsistent when ties are allowed. We study inference-time compute (ITC) for evaluators that generate n independent thinking-rating samples per item, and propose a principled, distribution-calibrated aggregation scheme. Our method models three-way preferences with a Bradley-Terry-Davidson formulation on rating counts, leveraging both polarity (margin among non-ties) and decisiveness (non-tie rate) to distinguish narrow margins from strong consensus. Across various evaluation benchmarks, our approach consistently reduces MAE and increases pairwise accuracy versus standard baselines, and when evaluated against human-consensus meta-labels, matches or exceeds individual human raters. These results show that carefully allocating ITC and aggregating with distribution-aware methods turns noisy individual model judgments into reliable ratings for evaluation.
Authors: Chenxu Niu, Wei Zhang, Jie Li, Yongjian Zhao, Tongyang Wang, Xi Wang, Yong Chen
Abstract: Large language model (LLM) services now answer billions of queries per day, and industry reports show that inference, not training, accounts for more than 90% of total power consumption. However, existing benchmarks focus on either training/fine-tuning or performance of inference and provide little support for power consumption measurement and analysis of inference. We introduce TokenPowerBench, the first lightweight and extensible benchmark designed for LLM-inference power consumption studies. The benchmark combines (i) a declarative configuration interface covering model choice, prompt set, and inference engine, (ii) a measurement layer that captures GPU-, node-, and system-level power without specialized power meters, and (iii) a phase-aligned metrics pipeline that attributes energy to the prefill and decode stages of every request. These elements make it straight-forward to explore the power consumed by an LLM inference run; furthermore, by varying batch size, context length, parallelism strategy and quantization, users can quickly assess how each setting affects joules per token and other energy-efficiency metrics. We evaluate TokenPowerBench on four of the most widely used model series (Llama, Falcon, Qwen, and Mistral). Our experiments cover from 1 billion parameters up to the frontier-scale Llama3-405B model. Furthermore, we release TokenPowerBench as open source to help users to measure power consumption, forecast operating expenses, and meet sustainability targets when deploying LLM services.
Authors: Abdullah Al Asif, Farhana Chowdhury Tisha
Abstract: One of the deadliest pandemics is now happening in the current world due to COVID-19. This contagious virus is spreading like wildfire around the whole world. To minimize the spreading of this virus, World Health Organization (WHO) has made protocols mandatory for wearing face masks and maintaining 6 feet physical distance. In this paper, we have developed a system that can detect the proper maintenance of that distance and people are properly using masks or not. We have used the customized attention-inceptionv3 model in this system for the identification of those two components. We have used two different datasets along with 10,800 images including both with and without Face Mask images. The training accuracy has been achieved 98% and validation accuracy 99.5%. The system can conduct a precision value of around 98.2% and the frame rate per second (FPS) was 25.0. So, with this system, we can identify high-risk areas with the highest possibility of the virus spreading zone. This may help authorities to take necessary steps to locate those risky areas and alert the local people to ensure proper precautions in no time.
Authors: Pankaj Jalote, Y. Raghu Reddy, Vasudeva Varma
Abstract: Covid has made online teaching and learning acceptable and students, faculty, and industry professionals are all comfortable with this mode. This comfort can be leveraged to offer an online multi-institutional research-level course in an area where individual institutions may not have the requisite faculty to teach and/or research students to enroll. If the subject is of interest to industry, online offering also allows industry experts to contribute and participate with ease. Advanced topics in Software Engineering are ideally suited for experimenting with this approach as industry, which is often looking to incorporate advances in software engineering in their practices, is likely to agree to contribute and participate. In this paper we describe an experiment in teaching a course titled "AI in Software Engineering" jointly between two institutions with active industry participation, and share our and student's experience. We believe this collaborative teaching approach can be used for offering research level courses in any applied area of computer science by institutions who are small and find it difficult to offer research level courses on their own.
Authors: Aditya Sneh, Nilesh Kumar Sahu, Snehil Gupta, Haroon R. Lone
Abstract: Accurately recognizing human context from smartphone sensor data remains a significant challenge, especially in sedentary settings where activities such as studying, attending lectures, relaxing, and eating exhibit highly similar inertial patterns. Furthermore, social context plays a critical role in understanding user behavior, yet is often overlooked in mobile sensing research. To address these gaps, we introduce LogMe, a mobile sensing application that passively collects smartphone sensor data (accelerometer, gyroscope, magnetometer, and rotation vector) and prompts users for hourly self-reports capturing both sedentary activity and social context. Using this dual-label dataset, we propose DySTAN (Dynamic Cross-Stitch with Task Attention Network), a multi-task learning framework that jointly classifies both context dimensions from shared sensor inputs. It integrates task-specific layers with cross-task attention to model subtle distinctions effectively. DySTAN improves sedentary activity macro F1 scores by 21.8% over a single-task CNN-BiLSTM-GRU (CBG) model and by 8.2% over the strongest multi-task baseline, Sluice Network (SN). These results demonstrate the importance of modeling multiple, co-occurring context dimensions to improve the accuracy and robustness of mobile context recognition.
Authors: Kashaf Gulzar, Dominik Wagner, Sebastian P. Bayerl, Florian H\"onig, Tobias Bocklet, Korbinian Riedhammer
Abstract: Automatic transcription of stuttered speech remains a challenge, even for modern end-to-end (E2E) automatic speech recognition (ASR) frameworks. Dysfluencies and fluency-shaping artifacts are often overlooked, resulting in non-verbatim transcriptions with limited clinical and research value. We propose a parameter-efficient adaptation method to decode dysfluencies and fluency modifications as special tokens within transcriptions, evaluated on simulated (LibriStutter, English) and natural (KSoF, German) stuttered speech datasets. To mitigate ASR performance disparities and bias towards English, we introduce a multi-step fine-tuning strategy with language-adaptive pretraining. Tokenization analysis further highlights the tokenizer's English-centric bias, which poses challenges for improving performance on German data. Our findings demonstrate the effectiveness of lightweight adaptation techniques for dysfluency-aware ASR while exposing key limitations in multilingual E2E systems.
Authors: Yiping Wang, Peiren Wang, Zhenye Li, Fang Liu, Jinguo Huang
Abstract: Complex spatial connectivity patterns, such as interictal suppression and ictal propagation, complicate accurate drug-resistant epilepsy (DRE) seizure detection using stereotactic electroencephalography (SEEG) and traditional machine learning methods. Two critical challenges remain:(1)a low signal-to-noise ratio in functional connectivity estimates, making it difficult to learn seizure-related interactions; and (2)expert labels for spatial pathological connectivity patterns are difficult to obtain, meanwhile lacking the patterns' representation to improve seizure detection. To address these issues, we propose a novel node-graph dual contrastive learning framework, Seizure-NGCLNet, to learn SEEG interictal suppression and ictal propagation patterns for detecting DRE seizures with high precision. First, an adaptive graph augmentation strategy guided by centrality metrics is developed to generate seizure-related brain networks. Second, a dual-contrastive learning approach is integrated, combining global graph-level contrast with local node-graph contrast, to encode both spatial structural and semantic epileptogenic features. Third, the pretrained embeddings are fine-tuned via a top-k localized graph attention network to perform the final classification. Extensive experiments on a large-scale public SEEG dataset from 33 DRE patients demonstrate that Seizure-NGCLNet achieves state-of-the-art performance, with an average accuracy of 95.93%, sensitivity of 96.25%, and specificity of 94.12%. Visualizations confirm that the learned embeddings clearly separate ictal from interictal states, reflecting suppression and propagation patterns that correspond to the clinical mechanisms. These results highlight Seizure-NGCLNet's ability to learn interpretable spatial pathological patterns, enhancing both seizure detection and seizure onset zone localization.
Authors: Hao Qian, Pu You, Lin Zeng, Jingyuan Zhou, Dengdeng Huang, Kaicheng Li, Shikui Tu, Lei Xu
Abstract: Glioblastoma (GBM) remains the most aggressive tumor, urgently requiring novel therapeutic strategies. Here, we present a dry-to-wet framework combining generative modeling and experimental validation to optimize peptides targeting ATP5A, a potential peptide-binding protein for GBM. Our framework introduces the first lead-conditioned generative model, which focuses exploration on geometrically relevant regions around lead peptides and mitigates the combinatorial complexity of de novo methods. Specifically, we propose POTFlow, a \underline{P}rior and \underline{O}ptimal \underline{T}ransport-based \underline{Flow}-matching model for peptide optimization. POTFlow employs secondary structure information (e.g., helix, sheet, loop) as geometric constraints, which are further refined by optimal transport to produce shorter flow paths. With this design, our method achieves state-of-the-art performance compared with five popular approaches. When applied to GBM, our method generates peptides that selectively inhibit cell viability and significantly prolong survival in a patient-derived xenograft (PDX) model. As the first lead peptide-conditioned flow matching model, POTFlow holds strong potential as a generalizable framework for therapeutic peptide design.
Authors: Gaurav Rudravaram, Lianrui Zuo, Adam M. Saunders, Michael E. Kim, Praitayini Kanakaraj, Nancy R. Newlin, Aravind R. Krishnan, Elyssa M. McMaster, Chloe Cho, Susan M. Resnick, Lori L. Beason Held, Derek Archer, Timothy J. Hohman, Daniel C. Moyer, Bennett A. Landman
Abstract: Structural connectomes are detailed graphs that map how different brain regions are physically connected, offering critical insight into aging, cognition, and neurodegenerative diseases. However, these connectomes are high-dimensional and densely interconnected, which makes them difficult to interpret and analyze at scale. While low-dimensional spaces like PCA and autoencoders are often used to capture major sources of variation, their latent spaces are generally continuous and cannot fully reflect the mixed nature of variability in connectomes, which include both continuous (e.g., connectivity strength) and discrete factors (e.g., imaging site). Motivated by this, we propose a variational autoencoder (VAE) with a hybrid latent space that jointly models the discrete and continuous components. We analyze a large dataset of 5,761 connectomes from six Alzheimer's disease studies with ten acquisition protocols. Each connectome represents a single scan from a unique subject (3579 females, 2182 males), aged 22 to 102, with 4338 cognitively normal, 809 with mild cognitive impairment (MCI), and 614 with Alzheimer's disease (AD). Each connectome contains 121 brain regions defined by the BrainCOLOR atlas. We train our hybrid VAE in an unsupervised way and characterize what each latent component captures. We find that the discrete space is particularly effective at capturing subtle site-related differences, achieving an Adjusted Rand Index (ARI) of 0.65 with site labels, significantly outperforming PCA and a standard VAE followed by clustering (p < 0.05). These results demonstrate that the hybrid latent space can disentangle distinct sources of variability in connectomes in an unsupervised manner, offering potential for large-scale connectome analysis.
Authors: Juan C. King, Jose M. Amigo
Abstract: The aim of this paper is the analysis and selection of stock trading systems that combine different models with data of different nature, such as financial and microeconomic information. Specifically, based on previous work by the authors and applying advanced techniques of Machine Learning and Deep Learning, our objective is to formulate trading algorithms for the stock market with empirically tested statistical advantages, thus improving results published in the literature. Our approach integrates Long Short-Term Memory (LSTM) networks with algorithms based on decision trees, such as Random Forest and Gradient Boosting. While the former analyze price patterns of financial assets, the latter are fed with economic data of companies. Numerical simulations of algorithmic trading with data from international companies and 10-weekday predictions confirm that an approach based on both fundamental and technical variables can outperform the usual approaches, which do not combine those two types of variables. In doing so, Random Forest turned out to be the best performer among the decision trees. We also discuss how the prediction performance of such a hybrid approach can be boosted by selecting the technical variables.
Authors: Xia Chen, Ruiji Sun, Philipp Geyer, Andr\'e Borrmann, Stefano Schiavon
Abstract: Human-factor analysis typically employs correlation analysis and significance testing to identify relationships between variables. However, these descriptive ('what-is') methods, while effective for identifying associations, are often insufficient for answering causal ('what-if') questions. Their application in such contexts often overlooks confounding and colliding variables, potentially leading to bias and suboptimal or incorrect decisions. We advocate for explicitly distinguishing descriptive from interventional questions in human-factor analysis, and applying causal inference frameworks specifically to these problems to prevent methodological mismatches. This approach disentangles complex variable relationships and enables counterfactual reasoning. Using post-occupancy evaluation (POE) data from the Center for the Built Environment's (CBE) Occupant Survey as a demonstration case, we show how causal discovery reveals intervention hierarchies and directional relationships that traditional associational analysis misses. The systematic distinction between causally associated and independent variables, combined with intervention prioritization capabilities, offers broad applicability to complex human-centric systems, for example, in building science or ergonomics, where understanding intervention effects is critical for optimization and decision-making.
Authors: Younes Ghazagh Jahed, Alireza Khatiri
Abstract: Frequency control in power systems is critical to maintaining stability and preventing blackouts. Traditional methods like meta-heuristic algorithms and machine learning face limitations in real-time applicability and scalability. This paper introduces a novel approach using a pure variational quantum circuit (VQC) for real-time secondary frequency control in diesel generators. Unlike hybrid classical-quantum models, the proposed VQC operates independently during execution, eliminating latency from classical-quantum data exchange. The VQC is trained via supervised learning to map historical frequency deviations to optimal Proportional-Integral (PI) controller parameters using a pre-computed lookup table. Simulations demonstrate that the VQC achieves high prediction accuracy (over 90%) with sufficient quantum measurement shots and generalizes well across diverse test events. The quantum-optimized PI parameters significantly improve transient response, reducing frequency fluctuations and settling time.
Authors: Ece Yurtseven
Abstract: Quantum machine learning has emerged as a promising approach to improve feature extraction and classification tasks in high-dimensional data domains such as medical imaging. In this work, we present a hybrid Quantum-Classical Convolutional Neural Network (QCNN) architecture designed for the binary classification of the BreastMNIST dataset, a standardized benchmark for distinguishing between benign and malignant breast tumors. Our architecture integrates classical convolutional feature extraction with two distinct quantum circuits: an amplitude-encoding variational quantum circuit (VQC) and an angle-encoding VQC circuit with circular entanglement, both implemented on four qubits. These circuits generate quantum feature embeddings that are fused with classical features to form a joint feature space, which is subsequently processed by a fully connected classifier. To ensure fairness, the hybrid QCNN is parameter-matched against a baseline classical CNN, allowing us to isolate the contribution of quantum layers. Both models are trained under identical conditions using the Adam optimizer and binary cross-entropy loss. Experimental evaluation in five independent runs demonstrates that the hybrid QCNN achieves statistically significant improvements in classification accuracy compared to the classical CNN, as validated by a one-sided Wilcoxon signed rank test (p = 0.03125) and supported by large effect size of Cohen's d = 2.14. Our results indicate that hybrid QCNN architectures can leverage entanglement and quantum feature fusion to enhance medical image classification tasks. This work establishes a statistical validation framework for assessing hybrid quantum models in biomedical applications and highlights pathways for scaling to larger datasets and deployment on near-term quantum hardware.
Authors: PIerre Dantas, Lucas Cordeiro, Youcheng Sun, Waldir Junior
Abstract: The idea of using Formal Verification tools with large language models (LLMs) has enabled scaling software verification beyond manual workflows. However, current methods remain unreliable. Without a solid theoretical footing, the refinement process can wander; sometimes it settles, sometimes it loops back, and sometimes it breaks away from any stable trajectory. This work bridges this critical gap by developing an LLM-Verifier Convergence Theorem, providing the first formal framework with provable guarantees for termination and convergence. We model the interaction between the LLM and the verifier as a discrete-time Markov Chain, with state transitions determined by a key parameter: the error-reduction probability ($\delta$). The procedure reaching the Verified state almost surely demonstrates that the program terminates for any $\delta > 0$, with an expected iteration count bounded by $\mathbb{E}[n] \leq 4/\delta$. We then stress-tested this prediction in an extensive empirical campaign comprising more than 90,000 trials. The empirical results match the theory with striking consistency. Every single run reached verification, and the convergence factor clustered tightly around $C_f\approx$ 1.0. Consequently, the bound mirrors the system's actual behavior. The evidence is sufficiently robust to support dividing the workflow into three distinct operating zones: marginal, practical, and high-performance. Consequently, we establish the design thresholds with absolute confidence. Together, the theoretical guarantee and the experimental evidence provide a clearer architectural foundation for LLM-assisted verification. Heuristic tuning no longer has to be carried out by the system. Engineers gain a framework that supports predictable resource planning and performance budgeting, precisely what is needed before deploying these pipelines into safety-critical software environments.
Authors: Dong Liu
Abstract: Persistence diagrams serve as a core tool in topological data analysis, playing a crucial role in pathological monitoring, drug discovery, and materials design. However, existing quantum topological algorithms, such as the LGZ algorithm, can only efficiently compute summary statistics like Betti numbers, failing to provide persistence diagram information that tracks the lifecycle of individual topological features, severely limiting their practical value. This paper proposes a novel quantum-classical hybrid algorithm that achieves, for the first time, the leap from "quantum computation of Betti numbers" to "quantum acquisition of practical persistence diagrams." The algorithm leverages the LGZ quantum algorithm as an efficient feature extractor, mining the harmonic form eigenvectors of the combinatorial Laplacian as well as Betti numbers, constructing specialized topological kernel functions to train a quantum support vector machine (QSVM), and learning the mapping from quantum topological features to persistence diagrams. The core contributions of this algorithm are: (1) elevating quantum topological computation from statistical summaries to pattern recognition, greatly expanding its application value; (2) obtaining more practical topological information in the form of persistence diagrams for real-world applications while maintaining the exponential speedup advantage of quantum computation; (3) proposing a novel hybrid paradigm of "classical precision guiding quantum efficiency." This method provides a feasible pathway for the practical implementation of quantum topological data analysis.
Authors: Sina Raeisadigh, Myles Joshua Toledo Tan, Henning M\"uller, Abderrahmane Hedjoudje
Abstract: This study compares baseline (J0) and 24-hour (J1) diffusion magnetic resonance imaging (MRI) for predicting three-month functional outcomes after acute ischemic stroke (AIS). Seventy-four AIS patients with paired apparent diffusion coefficient (ADC) scans and clinical data were analyzed. Three-dimensional ResNet-50 embeddings were fused with structured clinical variables, reduced via principal component analysis (<=12 components), and classified using linear support vector machines with eight-fold stratified group cross-validation. J1 multimodal models achieved the highest predictive performance (AUC = 0.923 +/- 0.085), outperforming J0-based configurations (AUC <= 0.86). Incorporating lesion-volume features further improved model stability and interpretability. These findings demonstrate that early post-treatment diffusion MRI provides superior prognostic value to pre-treatment imaging and that combining MRI, clinical, and lesion-volume features produces a robust and interpretable framework for predicting three-month functional outcomes in AIS patients.
Authors: Luca Attolico
Abstract: Timely assessment of current conditions is essential especially for small, open economies such as Singapore, where external shocks transmit rapidly to domestic activity. We develop a real-time nowcasting framework for quarterly GDP growth using a high-dimensional panel of approximately 70 indicators, encompassing economic and financial indicators over 1990Q1-2023Q2. The analysis covers penalized regressions, dimensionality-reduction methods, ensemble learning algorithms, and neural architectures, benchmarked against a Random Walk, an AR(3), and a Dynamic Factor Model. The pipeline preserves temporal ordering through an expanding-window walk-forward design with Bayesian hyperparameter optimization, and uses moving block-bootstrap procedures both to construct prediction intervals and to obtain confidence bands for feature-importance measures. It adopts model-specific and XAI-based explainability tools. A Model Confidence Set procedure identifies statistically superior learners, which are then combined through simple, weighted, and exponentially weighted schemes; the resulting time-varying weights provide an interpretable representation of model contributions. Predictive ability is assessed via Giacomini-White tests. Empirical results show that penalized regressions, dimensionality-reduction models, and GRU networks consistently outperform all benchmarks, with RMSFE reductions of roughly 40-60%; aggregation delivers further gains. Feature-attribution methods highlight industrial production, external trade, and labor-market indicators as dominant drivers of Singapore's short-run growth dynamics.
Authors: Sagie Levy, Elad Aharoni, Matan Levy, Ariel Shamir, Dani Lischinski
Abstract: We introduce Material Coating, a novel image editing task that simulates applying a thin material layer onto an object while preserving its underlying coarse and fine geometry. Material coating is fundamentally different from existing "material transfer" methods, which are designed to replace an object's intrinsic material, often overwriting fine details. To address this new task, we construct a large-scale synthetic dataset (110K images) of 3D objects with varied, physically-based coatings, named DataCoat110K. We then propose CoatFusion, a novel architecture that enables this task by conditioning a diffusion model on both a 2D albedo texture and granular, PBR-style parametric controls, including roughness, metalness, transmission, and a key thickness parameter. Experiments and user studies show CoatFusion produces realistic, controllable coatings and significantly outperforms existing material editing and transfer methods on this new task.
Authors: Pranav Asthana, Alex Hanson, Allen Tu, Tom Goldstein, Matthias Zwicker, Amitabh Varshney
Abstract: 3D Gaussian Splatting (3DGS) enables high-quality novel view synthesis, motivating interest in generating higher-resolution renders than those available during training. A natural strategy is to apply super-resolution (SR) to low-resolution (LR) input views, but independently enhancing each image introduces multi-view inconsistencies, leading to blurry renders. Prior methods attempt to mitigate these inconsistencies through learned neural components, temporally consistent video priors, or joint optimization on LR and SR views, but all uniformly apply SR across every image. In contrast, our key insight is that close-up LR views may contain high-frequency information for regions also captured in more distant views, and that we can use the camera pose relative to scene geometry to inform where to add SR content. Building from this insight, we propose SplatSuRe, a method that selectively applies SR content only in undersampled regions lacking high-frequency supervision, yielding sharper and more consistent results. Across Tanks & Temples, Deep Blending and Mip-NeRF 360, our approach surpasses baselines in both fidelity and perceptual quality. Notably, our gains are most significant in localized foreground regions where higher detail is desired.
Authors: Rajat Vadiraj Dwaraknath, Lexing Ying
Abstract: Metric graphs are structures obtained by associating edges in a standard graph with segments of the real line and gluing these segments at the vertices of the graph. The resulting structure has a natural metric that allows for the study of differential operators and stochastic processes on the graph. Brownian motions in these domains have been extensively studied theoretically using their generators. However, less work has been done on practical algorithms for simulating these processes. We introduce the first algorithm for simulating Brownian motions on metric graphs through a timestep splitting Euler-Maruyama-based discretization of their corresponding stochastic differential equation. By applying this scheme to Langevin diffusions on metric graphs, we also obtain the first algorithm for sampling on metric graphs. We provide theoretical guarantees on the number of timestep splittings required for the algorithm to converge to the underlying stochastic process. We also show that the exit probabilities of the simulated particle converge to the vertex-edge jump probabilities of the underlying stochastic differential equation as the timestep goes to zero. Finally, since this method is highly parallelizable, we provide fast, memory-aware implementations of our algorithm in the form of custom CUDA kernels that are up to ~8000x faster than a GPU implementation using PyTorch on simple star metric graphs. Beyond simple star graphs, we benchmark our algorithm on a real cortical vascular network extracted from a DuMuX tissue-perfusion model for tracer transport. Our algorithm is able to run stable simulations with timesteps significantly larger than the stable limit of the finite volume method used in DuMuX while also achieving speedups of up to ~1500x.
Authors: Ziyan Wang, Enmao Diao, Qi Le, Pu Wang, Guanchu Wang, Minwoo Lee, Shu-ping Yeh, Li Yang
Abstract: Reasoning LLMs (RLMs) such as OpenAI o1, DeepSeek-R1, and Qwen3 deliver strong multi-step reasoning through chain-of-thought generation, but their large model sizes and lengthy decode-time outputs make them costly to deploy and unsuitable for resource-constrained settings. To reduce computing and memory cost, pruning offers a promising solution by removing unimportant parameters. However, despite their success on standard LLMs, existing pruning methods severely damage RLMs, as even moderate sparsity (e.g., 20%) can collapse accuracy and completely disrupt the model's reasoning coherence. We begin by analyzing why existing pruning pipelines fail on reasoning LLMs and find that their brittleness largely stems from a mismatch between the calibration data, the pruning objective, and the model's decode-time reasoning behavior. Our study further shows that the most reliable calibration signal comes not from human-written labels but from the model's own self-generated reasoning traces, which more accurately reflect its inference distribution. Guided by these insights, we introduce RESP, a self-reflective structured pruning framework that aligns pruning decisions with the model's reasoning dynamics through self-generated calibration, decode-only gradient-based importance estimation, and progressive regeneration that maintains calibration fidelity as sparsity increases. Experiments on Qwen3-8B demonstrate that RESP markedly outperforms existing structured pruning methods on both GSM8K and MathQA, preserving near-dense accuracy at 20-30% sparsity and substantially mitigating performance collapse at higher sparsity levels. At 40% sparsity, RESP attains 81.3% accuracy on GSM8K and 59.6% on MathQA, surpassing the strongest baselines by 66.87% and 47%, respectively.
Authors: Jifeng Li, Arnav Grover, Abraham Alpuerto, Yupeng Cao, Xiao-Yang Liu
Abstract: The financial market is a mission-critical playground for AI agents due to its temporal dynamics and low signal-to-noise ratio. Building an effective algorithmic trading system may require a professional team to develop and test over the years. In this paper, we propose an orchestration framework for financial agents, which aims to democratize financial intelligence to the general public. We map each component of the traditional algorithmic trading system to agents, including planner, orchestrator, alpha agents, risk agents, portfolio agents, backtest agents, execution agents, audit agents, and memory agent. We present two in-house trading examples. For the stock trading task (hourly data from 04/2024 to 12/2024), our approach achieved a return of $20.42\%$, a Sharpe ratio of 2.63, and a maximum drawdown of $-3.59\%$, while the S&P 500 index yielded a return of $15.97\%$. For the BTC trading task (minute data from 27/07/2025 to 13/08/2025), our approach achieved a return of $8.39\%$, a Sharpe ratio of $0.38$, and a maximum drawdown of $-2.80\%$, whereas the BTC price increased by $3.80\%$. Our code is available on \href{https://github.com/Open-Finance-Lab/AgenticTrading}{GitHub}.
Authors: Shubhi Asthana, Bing Zhang, Chad DeLuca, Ruchi Mahindru, Hima Patel
Abstract: The rapid shift from stateless large language models (LLMs) to autonomous, goal-driven agents raises a central question: When is agentic AI truly necessary? While agents enable multi-step reasoning, persistent memory, and tool orchestration, deploying them indiscriminately leads to higher cost, complexity, and risk. We present STRIDE (Systematic Task Reasoning Intelligence Deployment Evaluator), a framework that provides principled recommendations for selecting between three modalities: (i) direct LLM calls, (ii) guided AI assistants, and (iii) fully autonomous agentic AI. STRIDE integrates structured task decomposition, dynamism attribution, and self-reflection requirement analysis to produce an Agentic Suitability Score, ensuring that full agentic autonomy is reserved for tasks with inherent dynamism or evolving context. Evaluated across 30 real-world tasks spanning SRE, compliance, and enterprise automation, STRIDE achieved 92% accuracy in modality selection, reduced unnecessary agent deployments by 45%, and cut resource costs by 37%. Expert validation over six months in SRE and compliance domains confirmed its practical utility, with domain specialists agreeing that STRIDE effectively distinguishes between tasks requiring simple LLM calls, guided assistants, or full agentic autonomy. This work reframes agent adoption as a necessity-driven design decision, ensuring autonomy is applied only when its benefits justify the costs.
Authors: Le Thien Phuc Nguyen, Zhuoran Yu, Samuel Low Yu Hang, Subin An, Jeongik Lee, Yohan Ban, SeungEun Chung, Thanh-Huy Nguyen, JuWan Maeng, Soochahn Lee, Yong Jae Lee
Abstract: Multimodal large language models (MLLMs) are expected to jointly interpret vision, audio, and language, yet existing video benchmarks rarely assess fine-grained reasoning about human speech. Many tasks remain visually solvable or only coarsely evaluate speech, offering limited insight into whether models can align who speaks, what is said, and when it occurs. We introduce AV-SpeakerBench, a curated benchmark of 3,212 multiple-choice questions focused on speaker-centric audiovisual reasoning in real-world videos. It features: (1) a speaker-centered formulation that treats speakers-not scenes-as the core reasoning unit; (2) fusion-grounded question design embedding audiovisual dependencies into question semantics; and (3) expert-curated annotations ensuring temporal precision and cross-modal validity. Comprehensive evaluations show that the Gemini family consistently outperforms open-source systems, with Gemini 2.5 Pro achieving the best results. Among open models, Qwen3-Omni-30B approaches Gemini 2.0 Flash but remains far behind Gemini 2.5 Pro, primarily due to weaker audiovisual fusion rather than visual perception. We believe AV-SpeakerBench establishes a rigorous foundation for advancing fine-grained audiovisual reasoning in future multimodal systems.
Authors: Md Abdul Ahad Minhaz, Zannatul Zahan Meem, Md. Shohrab Hossain
Abstract: Phishing remains one of the most prevalent online threats, exploiting human trust to harvest sensitive credentials. Existing URL- and HTML-based detection systems struggle against obfuscation and visual deception. This paper presents \textbf{PhishSnap}, a privacy-preserving, on-device phishing detection system leveraging perceptual hashing (pHash). Implemented as a browser extension, PhishSnap captures webpage screenshots, computes visual hashes, and compares them against legitimate templates to identify visually similar phishing attempts. A \textbf{2024 dataset of 10,000 URLs} (70\%/20\%/10\% train/validation/test) was collected from PhishTank and Netcraft. Due to security takedowns, a subset of phishing pages was unavailable, reducing dataset diversity. The system achieved \textbf{0.79 accuracy}, \textbf{0.76 precision}, and \textbf{0.78 recall}, showing that visual similarity remains a viable anti-phishing measure. The entire inference process occurs locally, ensuring user privacy and minimal latency.
Authors: Alexander Davydov
Abstract: We address the problem of verifying closed-loop contraction in nonlinear control systems whose controller and contraction metric are both parameterized by neural networks. By leveraging interval analysis and interval bound propagation, we derive a tractable and scalable sufficient condition for closed-loop contractivity that reduces to checking that the dominant eigenvalue of a symmetric Metzler matrix is nonpositive. We combine this sufficient condition with a domain partitioning strategy to integrate this sufficient condition into training. The proposed approach is validated on an inverted pendulum system, demonstrating the ability to learn neural network controllers and contraction metrics that provably satisfy the contraction condition.
Authors: Jeremy Andrew Irvin, Jiaqi Han, Zikui Wang, Abdulaziz Alharbi, Yufei Zhao, Nomin-Erdene Bayarsaikhan, Daniele Visioni, Andrew Y. Ng, Duncan Watson-Parris
Abstract: Generative models have the potential to transform the way we emulate Earth's changing climate. Previous generative approaches rely on weather-scale autoregression for climate emulation, but this is inherently slow for long climate horizons and has yet to demonstrate stable rollouts under nonstationary forcings. Here, we introduce Spatiotemporal Pyramid Flows (SPF), a new class of flow matching approaches that model data hierarchically across spatial and temporal scales. Inspired by cascaded video models, SPF partitions the generative trajectory into a spatiotemporal pyramid, progressively increasing spatial resolution to reduce computation and coupling each stage with an associated timescale to enable direct sampling at any temporal level in the pyramid. This design, together with conditioning each stage on prescribed physical forcings (e.g., greenhouse gases or aerosols), enables efficient, parallel climate emulation at multiple timescales. On ClimateBench, SPF outperforms strong flow matching baselines and pre-trained models at yearly and monthly timescales while offering fast sampling, especially at coarser temporal levels. To scale SPF, we curate ClimateSuite, the largest collection of Earth system simulations to date, comprising over 33,000 simulation-years across ten climate models and the first dataset to include simulations of climate interventions. We find that the scaled SPF model demonstrates good generalization to held-out scenarios across climate models. Together, SPF and ClimateSuite provide a foundation for accurate, efficient, probabilistic climate emulation across temporal scales and realistic future scenarios. Data and code is publicly available at https://github.com/stanfordmlgroup/spf .
Authors: Adel Chehade, Edoardo Ragusa, Paolo Gastaldo, Rodolfo Zunino
Abstract: Traffic classification (TC) plays a critical role in cybersecurity, particularly in IoT and embedded contexts, where inspection must often occur locally under tight hardware constraints. We use hardware-aware neural architecture search (HW-NAS) to derive lightweight TC models that are accurate, efficient, and deployable on edge platforms. Two input formats are considered: a flattened byte sequence and a 2D packet-wise time series; we examine how input structure affects adversarial vulnerability when using resource-constrained models. Robustness is assessed against white-box attacks, specifically Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD). On USTC-TFC2016, both HW-NAS models achieve over 99% clean-data accuracy while remaining within 65k parameters and 2M FLOPs. Yet under perturbations of strength 0.1, their robustness diverges: the flat model retains over 85% accuracy, while the time-series variant drops below 35%. Adversarial fine-tuning delivers robust gains, with flat-input accuracy exceeding 96% and the time-series variant recovering over 60 percentage points in robustness, all without compromising efficiency. The results underscore how input structure influences adversarial vulnerability, and show that even compact, resource-efficient models can attain strong robustness, supporting their practical deployment in secure edge-based TC.
Authors: Boyu Zhu, Xiaofei Wen, Wenjie Jacky Mo, Tinghui Zhu, Yanan Xie, Peng Qi, Muhao Chen
Abstract: Omni-modal Large Language Models (OLLMs) that process text, images, videos, and audio introduce new challenges for safety and value guardrails in human-AI interaction. Prior guardrail research largely targets unimodal settings and typically frames safeguarding as binary classification, which limits robustness across diverse modalities and tasks. To address this gap, we propose OmniGuard, the first family of omni-modal guardrails that performs safeguarding across all modalities with deliberate reasoning ability. To support the training of OMNIGUARD, we curate a large, comprehensive omni-modal safety dataset comprising over 210K diverse samples, with inputs that cover all modalities through both unimodal and cross-modal samples. Each sample is annotated with structured safety labels and carefully curated safety critiques from expert models through targeted distillation. Extensive experiments on 15 benchmarks show that OmniGuard achieves strong effectiveness and generalization across a wide range of multimodal safety scenarios. Importantly, OmniGuard provides a unified framework that enforces policies and mitigates risks in omni-modalities, paving the way toward building more robust and capable omnimodal safeguarding systems.
Authors: Felix Teufel, Aaron W. Kollasch, Yining Huang, Ole Winther, Kevin K. Yang, Pascal Notin, Debora S. Marks
Abstract: Accurately predicting protein fitness with minimal experimental data is a persistent challenge in protein engineering. We introduce PRIMO (PRotein In-context Mutation Oracle), a transformer-based framework that leverages in-context learning and test-time training to adapt rapidly to new proteins and assays without large task-specific datasets. By encoding sequence information, auxiliary zero-shot predictions, and sparse experimental labels from many assays as a unified token set in a pre-training masked-language modeling paradigm, PRIMO learns to prioritize promising variants through a preference-based loss function. Across diverse protein families and properties-including both substitution and indel mutations-PRIMO outperforms zero-shot and fully supervised baselines. This work underscores the power of combining large-scale pre-training with efficient test-time adaptation to tackle challenging protein design tasks where data collection is expensive and label availability is limited.
Authors: Jiabao Brad Wang, Siyuan Cao, Hongxuan Wu, Yiliang Yuan, Mustafa Misir
Abstract: Selecting an effective docking algorithm is highly context-dependent, and no single method performs reliably across structural, chemical, or protocol regimes. We introduce MolAS, a lightweight algorithm selection system that predicts per-algorithm performance from pretrained protein-ligand embeddings using attentional pooling and a shallow residual decoder. With only hundreds to a few thousand labelled complexes, MolAS achieves up to 15% absolute improvement over the single-best solver (SBS) and closes 17-66% of the Virtual Best Solver (VBS)-SBS gap across five diverse docking benchmarks. Analyses of reliability, embedding geometry, and solver-selection patterns show that MolAS succeeds when the oracle landscape exhibits low entropy and separable solver behaviour, but collapses under protocol-induced hierarchy shifts. These findings indicate that the main barrier to robust docking AS is not representational capacity but instability in solver rankings across pose-generation regimes, positioning MolAS as both a practical in-domain selector and a diagnostic tool for assessing when AS is feasible.
Authors: Dimitris Oikonomou, Nicolas Loizou
Abstract: The stochastic Polyak step size (SPS) has proven to be a promising choice for stochastic gradient descent (SGD), delivering competitive performance relative to state-of-the-art methods on smooth convex and non-convex optimization problems, including deep neural network training. However, extensions of this approach to non-smooth settings remain in their early stages, often relying on interpolation assumptions or requiring knowledge of the optimal solution. In this work, we propose a novel SPS variant, Safeguarded SPS (SPS$_{safe}$), for the stochastic subgradient method, and provide rigorous convergence guarantees for non-smooth convex optimization with no need for strong assumptions. We further incorporate momentum into the update rule, yielding equally tight theoretical results. Comprehensive experiments on convex benchmarks and deep neural networks corroborate our theory: the proposed step size accelerates convergence, reduces variance, and consistently outperforms existing adaptive baselines. Finally, in the context of deep neural network training, our method demonstrates robust performance by addressing the vanishing gradient problem.
Authors: David X. Wu, Shreyas Kapur, Anant Sahai, Stuart Russell
Abstract: Reinforcement learning has become the dominant paradigm for eliciting reasoning and self-correction capabilities in large language models, but its computational expense motivates exploration of alternatives. Inspired by techniques from autonomous driving and robotics, we investigate whether supervised learning with synthetic error injection can induce self-correction abilities in language models. Our approach inserts artificial errors into reasoning chains, masks them, and supervises the model to recognize and correct these mistakes. Despite the intuitive appeal of this method, we find that it fails to significantly improve performance even on simple synthetic tasks across multiple models. Moreover, even when the model catches its own error, it often parrots the original mistake. We find that the distribution shift of synthetic errors to on-policy errors significantly degrades the error-correction capabilities of the fine-tuned model, even with good synthetic coverage of on-policy errors. Our results help explain why on-policy reinforcement learning methods have proven uniquely effective for eliciting self-correction.
Authors: Anoop Cherian, River Doyle, Eyal Ben-Dov, Suhas Lohit, Kuan-Chuan Peng
Abstract: Recent large language models (LLMs) are trained on diverse corpora and tasks, leading them to develop complementary strengths. Multi-agent debate (MAD) has emerged as a popular way to leverage these strengths for robust reasoning, though it has mostly been applied to language-only tasks, leaving its efficacy on multimodal problems underexplored. In this paper, we study MAD for solving vision-and-language reasoning problems. Our setup enables generalizing the debate protocol with heterogeneous experts that possess single- and multi-modal capabilities. To this end, we present Weighted Iterative Society-of-Experts (WISE), a generalized and modular MAD framework that partitions the agents into Solvers, that generate solutions, and Reflectors, that verify correctness, assign weights, and provide natural language feedback. To aggregate the agents' solutions across debate rounds, while accounting for variance in their responses and the feedback weights, we present a modified Dawid-Skene algorithm for post-processing that integrates our two-stage debate model. We evaluate WISE on SMART-840, VisualPuzzles, EvoChart-QA, and a new SMART-840++ dataset with programmatically generated problem instances of controlled difficulty. Our results show that WISE consistently improves accuracy by 2-7% over the state-of-the-art MAD setups and aggregation methods across diverse multimodal tasks and LLM configurations.
Authors: Yuexin Xiang, Yuchen Lei, SM Mahir Shazeed Rish, Yuanzhe Zhang, Qin Wang, Tsz Hon Yuen, Jiangshan Yu
Abstract: Stablecoins such as USDT and USDC aspire to peg stability by coupling issuance controls with reserve attestations. In practice, however, the transparency is split across two worlds: verifiable on-chain traces and off-chain disclosures locked in unstructured text that are unconnected. We introduce a large language model (LLM)-based automated framework that bridges these two dimensions by aligning on-chain issuance data with off-chain disclosure statements. First, we propose an integrative framework using LLMs to capture and analyze on- and off-chain data through document parsing and semantic alignment, extracting key financial indicators from issuer attestations and mapping them to corresponding on-chain metrics. Second, we integrate multi-chain issuance records and disclosure documents within a model context protocol (MCP) framework that standardizes LLMs access to both quantitative market data and qualitative disclosure narratives. This framework enables unified retrieval and contextual alignment across heterogeneous stablecoin information sources and facilitates consistent analysis. Third, we demonstrate the capability of LLMs to operate across heterogeneous data modalities in blockchain analytics, quantifying discrepancies between reported and observed circulation and examining their implications for cross-chain transparency and price dynamics. Our findings reveal systematic gaps between disclosed and verifiable data, showing that LLM-assisted analysis enhances cross-modal transparency and supports automated, data-driven auditing in decentralized finance (DeFi).
Authors: Tommaso Fioravanti, Brian Quanz, Gabriele Agliardi, Edgar Andres Ruiz Guzman, Gin\'es Carrascal, Jae-Eun Park
Abstract: Quantum Machine Learning (QML) holds the promise of enhancing machine learning modeling in terms of both complexity and accuracy. A key challenge in this domain is the encoding of input data, which plays a pivotal role in determining the performance of QML models. In this work, we tackle a largely unaddressed aspect of encoding that is unique to QML modeling -- rather than adjusting the ansatz used for encoding, we consider adjusting how data is conveyed to the ansatz. We specifically implement QML pipelines that leverage classical data manipulation (i.e., ordering, selecting, and weighting features) as a preprocessing step, and evaluate if these aspects of encoding can have a significant impact on QML model performance, and if they can be effectively optimized to improve performance. Our experimental results, applied across a wide variety of data sets, ansatz, and circuit sizes, with a representative QML approach, demonstrate that by optimizing how features are encoded in an ansatz we can substantially and consistently improve the performance of QML models, making a compelling case for integrating these techniques in future QML applications. Finally we demonstrate the practical feasibility of this approach by running it using real quantum hardware with 100 qubit circuits and successfully achieving improved QML modeling performance in this case as well.
Authors: Woongyeong Yeo, Kangsan Kim, Jaehong Yoon, Sung Ju Hwang
Abstract: Recent advances in video large language models have demonstrated strong capabilities in understanding short clips. However, scaling them to hours- or days-long videos remains highly challenging due to limited context capacity and the loss of critical visual details during abstraction. Existing memory-augmented methods mitigate this by leveraging textual summaries of video segments, yet they heavily rely on text and fail to utilize visual evidence when reasoning over complex scenes. Moreover, retrieving from fixed temporal scales further limits their flexibility in capturing events that span variable durations. To address this, we introduce WorldMM, a novel multimodal memory agent that constructs and retrieves from multiple complementary memories, encompassing both textual and visual representations. WorldMM comprises three types of memory: episodic memory indexes factual events across multiple temporal scales, semantic memory continuously updates high-level conceptual knowledge, and visual memory preserves detailed information about scenes. During inference, an adaptive retrieval agent iteratively selects the most relevant memory source and leverages multiple temporal granularities based on the query, continuing until it determines that sufficient information has been gathered. WorldMM significantly outperforms existing baselines across five long video question-answering benchmarks, achieving an average 8.4% performance gain over previous state-of-the-art methods, showing its effectiveness on long video reasoning.
Authors: Junghwan Park, Woojin Cho, Junhyuk Heo, Darongsae Kwon, Kookjin Lee
Abstract: Adapting large pre-trained models to unseen tasks under tight data and compute budgets remains challenging. Meta-learning approaches explicitly learn good initializations, but they require an additional meta-training phase over many tasks, incur high training cost, and can be unstable. At the same time, the number of task-specific pre-trained models continues to grow, yet the question of how to transfer them to new tasks with minimal additional training remains relatively underexplored. We propose BOLT (Basis-Oriented Low-rank Transfer), a framework that reuses existing fine-tuned models not by merging weights, but instead by extracting an orthogonal, task-informed spectral basis and adapting within that subspace. In the offline phase, BOLT collects dominant singular directions from multiple task vectors and orthogonalizes them per layer to form reusable bases. In the online phase, we freeze these bases and train only a small set of diagonal coefficients per layer for the new task, yielding a rank-controlled update with very few trainable parameters. This design provides (i) a strong, training-free initialization for unseen tasks, obtained by pooling source-task coefficients, along with a lightweight rescaling step while leveraging the shared orthogonal bases, and (ii) a parameter-efficient fine-tuning (PEFT) path that, in our experiments, achieves robust performance compared to common PEFT baselines as well as a representative meta-learned initialization. Our results show that constraining adaptation to a task-informed orthogonal subspace provides an effective alternative for unseen-task transfer.
Authors: Ning Wang, Sainyam Galhotra
Abstract: Discovering which tables in large, heterogeneous repositories can be joined and by what transformations is a central challenge in data integration and data discovery. Traditional join discovery methods are largely designed for equi-joins, which assume that join keys match exactly or nearly so. These techniques, while efficient in clean, well-normalized databases, fail in open or federated settings where identifiers are inconsistently formatted, embedded, or split across multiple columns. Approximate or fuzzy joins alleviate minor string variations but cannot capture systematic transformations. We introduce QJoin, a reinforcement-learning framework that learns and reuses transformation strategies across join tasks. QJoin trains an agent under a uniqueness-aware reward that balances similarity with key distinctiveness, enabling it to explore concise, high-value transformation chains. To accelerate new joins, we introduce two reuse mechanisms: (i) agent transfer, which initializes new policies from pretrained agents, and (ii) transformation reuse, which caches successful operator sequences for similar column clusters. On the AutoJoin Web benchmark (31 table pairs), QJoin achieves an average F1-score of 91.0%. For 19,990 join tasks in NYC+Chicago open datasets, Qjoin reduces runtime by up to 7.4% (13,747 s) by using reusing. These results demonstrate that transformation learning and reuse can make join discovery both more accurate and more efficient.
Authors: Wenhao Yu, Zhenwen Liang, Chengsong Huang, Kishan Panaganti, Tianqing Fang, Haitao Mi, Dong Yu
Abstract: AI self-evolution has long been envisioned as a path toward superintelligence, where models autonomously acquire, refine, and internalize knowledge from their own learning experiences. Yet in practice, unguided self-evolving systems often plateau quickly or even degrade as training progresses. These failures arise from issues such as concept drift, diversity collapse, and mis-evolution, as models reinforce their own biases and converge toward low-entropy behaviors. To enable models to self-evolve in a stable and controllable manner while minimizing reliance on human supervision, we introduce R-Few, a guided Self-Play Challenger-Solver framework that incorporates lightweight human oversight through in-context grounding and mixed training. At each iteration, the Challenger samples a small set of human-labeled examples to guide synthetic question generation, while the Solver jointly trains on human and synthetic examples under an online, difficulty-based curriculum. Across math and general reasoning benchmarks, R-Few achieves consistent and iterative improvements. For example, Qwen3-8B-Base improves by +3.0 points over R-Zero on math tasks and achieves performance on par with General-Reasoner, despite the latter being trained on 20 times more human data. Ablation studies confirm the complementary contributions of grounded challenger training and curriculum-based solver training, and further analysis shows that R-Few mitigates drift, yielding more stable and controllable co-evolutionary dynamics.
Authors: Yuta Oshima, Yusuke Iwasawa, Masahiro Suzuki, Yutaka Matsuo, Hiroki Furuta
Abstract: Video world models have attracted significant attention for their ability to produce high-fidelity future visual observations conditioned on past observations and navigation actions. Temporally- and spatially-consistent, long-term world modeling has been a long-standing problem, unresolved with even recent state-of-the-art models, due to the prohibitively expensive computational costs for long-context inputs. In this paper, we propose WorldPack, a video world model with efficient compressed memory, which significantly improves spatial consistency, fidelity, and quality in long-term generation despite much shorter context length. Our compressed memory consists of trajectory packing and memory retrieval; trajectory packing realizes high context efficiency, and memory retrieval maintains the consistency in rollouts and helps long-term generations that require spatial reasoning. Our performance is evaluated with LoopNav, a benchmark on Minecraft, specialized for the evaluation of long-term consistency, and we verify that WorldPack notably outperforms strong state-of-the-art models.
Authors: Yarden Gabbay, Haoquan Guan, Shaull Almagor, El Kindi Rezig, Brit Youngmann, Babak Salimi
Abstract: Causal analyses derived from observational data underpin high-stakes decisions in domains such as healthcare, public policy, and economics. Yet such conclusions can be surprisingly fragile: even minor data errors - duplicate records, or entry mistakes - may drastically alter causal relationships. This raises a fundamental question: how robust is a causal claim to small, targeted modifications in the data? Addressing this question is essential for ensuring the reliability, interpretability, and reproducibility of empirical findings. We introduce SubCure, a framework for robustness auditing via cardinality repairs. Given a causal query and a user-specified target range for the estimated effect, SubCure identifies a small set of tuples or subpopulations whose removal shifts the estimate into the desired range. This process not only quantifies the sensitivity of causal conclusions but also pinpoints the specific regions of the data that drive those conclusions. We formalize this problem under both tuple- and pattern-level deletion settings and show both are NP-complete. To scale to large datasets, we develop efficient algorithms that incorporate machine unlearning techniques to incrementally update causal estimates without retraining from scratch. We evaluate SubCure across four real-world datasets covering diverse application domains. In each case, it uncovers compact, high-impact subsets whose removal significantly shifts the causal conclusions, revealing vulnerabilities that traditional methods fail to detect. Our results demonstrate that cardinality repair is a powerful and general-purpose tool for stress-testing causal analyses and guarding against misleading claims rooted in ordinary data imperfections.
Authors: Ali Mohammad-Djafari, Ning Chu, Li Wang
Abstract: Inverse problems arise across scientific and engineering domains, where the goal is to infer hidden parameters or physical fields from indirect and noisy observations. Classical approaches, such as variational regularization and Bayesian inference, provide well established theoretical foundations for handling ill posedness. However, these methods often become computationally restrictive in high dimensional settings or when the forward model is governed by complex physics. Physics Informed Neural Networks (PINNs) have recently emerged as a promising framework for solving inverse problems by embedding physical laws directly into the training process of neural networks. In this paper, we introduce a new perspective on the Bayesian Physics Informed Neural Network (BPINN) framework, extending classical PINNs by explicitly incorporating training data generation, modeling and measurement uncertainties through Bayesian prior modeling and doing inference with the posterior laws. Also, as we focus on the inverse problems, we call this method BPINN-IP, and we show that the standard PINN formulation naturally appears as its special case corresponding to the Maximum A Posteriori (MAP) estimate. This unified formulation allows simultaneous exploitation of physical constraints, prior knowledge, and data-driven inference, while enabling uncertainty quantification through posterior distributions. To demonstrate the effectiveness of the proposed framework, we consider inverse problems arising in infrared image processing, including deconvolution and super-resolution, and present results on both simulated and real industrial data.
Authors: Xueyan Li, Yuxin Wang, Mengjie Jiang, Qingzi Zhu, Jiang Zhang, Zoey Kim, Yazhe Niu
Abstract: Singing voice synthesis (SVS) has advanced significantly, enabling models to generate vocals with accurate pitch and consistent style. As these capabilities improve, the need for reliable evaluation and optimization becomes increasingly critical. However, current methods like reward systems often rely on single numerical scores, struggle to capture various dimensions such as phrasing or expressiveness, and require costly annotations, limiting interpretability and generalization. To address these issues, we propose a generative feedback (i.e., reward model) framework that provides multi-dimensional language and audio feedback for SVS assessment. Our approach leverages an audio-language model to generate text and audio critiques-covering aspects such as melody, content, and auditory quality. The model is fine-tuned on a hybrid dataset combining human music reactions and synthetic critiques from a MLLMs, enhancing diversity and linguistic richness. Quantitative experiments validate the effectiveness of the proposed dataset and training strategy, demonstrating that the framework produces musically accurate and interpretable evaluations suitable for guiding generative model improvement. The code is at [https://github.com/opendilab/VocalCritic](https://github.com/opendilab/VocalCritic)
URLs: https://github.com/opendilab/VocalCritic](https://github.com/opendilab/VocalCritic)
Authors: Parth Pulkundwar, Vivek Dhanawade, Rohit Yadav, Minal Sonkar, Medha Asurlekar, Sarita Rathod
Abstract: Traditional language models face a challenge from hallucinations. Their very presence casts a large, dangerous shadow over the promising realm of natural language processing. It becomes crucial to understand the various kinds of hallucinations that occur nowadays, their origins, and ways of reducing them. This document provides a concise and straightforward summary of that. It serves as a one-stop resource for a general understanding of hallucinations and how to mitigate them.
Authors: Albert Saiapin, Kim Batselier
Abstract: To address the scalability limitations of Gaussian process (GP) regression, several approximation techniques have been proposed. One such method is based on tensor networks, which utilizes an exponential number of basis functions without incurring exponential computational cost. However, extending this model to a fully probabilistic formulation introduces several design challenges. In particular, for tensor train (TT) models, it is unclear which TT-core should be treated in a Bayesian manner. We introduce a Bayesian tensor train kernel machine that applies Laplace approximation to estimate the posterior distribution over a selected TT-core and employs variational inference (VI) for precision hyperparameters. Experiments show that core selection is largely independent of TT-ranks and feature structure, and that VI replaces cross-validation while offering up to 65x faster training. The method's effectiveness is demonstrated on an inverse dynamics problem.
Authors: Ioanna Tasou, Panagiotis Mpakos, Angelos Vlachos, Dionysios Adamopoulos, Georgios Giannakopoulos, Konstantinos Katsikopoulos, Ioannis Karaparisis, Maria Lazou, Spyridon Loukovitis, Areti Mei, Anastasia Poulopoulou, Angeliki Dimitriou, Giorgos Filandrianos, Dimitrios Galanopoulos, Vasileios Karampinis, Ilias Mitsouras, Nikolaos Spanos, Petros Anastasiadis, Ioannis Doudalis, Konstantinos Nikas, George Retsinas, Paraskevi Tzouveli, Christina Giannoula, Nectarios Koziris, Nikela Papadopoulou, Giorgos Stamou, Athanasios Voulodimos, Georgios Goumas
Abstract: The computational demands of modern Deep Neural Networks (DNNs) are immense and constantly growing. While training costs usually capture public attention, inference demands are also contributing in significant computational, energy and environmental footprints. Sparsity stands out as a critical mechanism for drastically reducing these resource demands. However, its potential remains largely untapped and is not yet fully incorporated in production AI systems. To bridge this gap, this work provides the necessary knowledge and insights for performance engineers keen to get involved in deep learning inference optimization. In particular, in this work we: a) discuss the various forms of sparsity that can be utilized in DNN inference, b) explain how the original dense computations translate to sparse kernels, c) provide an extensive bibliographic review of the state-of-the-art in the implementation of these kernels for CPUs and GPUs, d) discuss the availability of sparse datasets in support of sparsity-related research and development, e) explore the current software tools and frameworks that provide robust sparsity support, and f) present evaluation results of different implementations of the key SpMM and SDDMM kernels on CPU and GPU platforms. Ultimately, this paper aims to serve as a resource for performance engineers seeking to develop and deploy highly efficient sparse deep learning models in productions.
Authors: Mattia Giuri, Mathias Jackermeier, Alessandro Abate
Abstract: Linear temporal logic (LTL) is a compelling framework for specifying complex, structured tasks for reinforcement learning (RL) agents. Recent work has shown that interpreting LTL instructions as finite automata, which can be seen as high-level programs monitoring task progress, enables learning a single generalist policy capable of executing arbitrary instructions at test time. However, existing approaches fall short in environments where multiple high-level events (i.e., atomic propositions) can be true at the same time and potentially interact in complicated ways. In this work, we propose a novel approach to learning a multi-task policy for following arbitrary LTL instructions that addresses this shortcoming. Our method conditions the policy on sequences of simple Boolean formulae, which directly align with transitions in the automaton, and are encoded via a graph neural network (GNN) to yield structured task representations. Experiments in a complex chess-based environment demonstrate the advantages of our approach.
Authors: Junwon Lee, Juhan Nam, Jiyoung Lee
Abstract: This work introduces a new task, text-conditioned selective video-to-audio (V2A) generation, which produces only the user-intended sound from a multi-object video. This capability is especially crucial in multimedia production, where audio tracks are handled individually for each sound source for precise editing, mixing, and creative control. However, current approaches generate single source-mixed sounds at once, largely because visual features are entangled, and region cues or prompts often fail to specify the source. We propose SelVA, a novel text-conditioned V2A model that treats the text prompt as an explicit selector of target source and modulates video encoder to distinctly extract prompt-relevant video features. The proposed supplementary tokens promote cross-attention by suppressing text-irrelevant activations with efficient parameter tuning, yielding robust semantic and temporal grounding. SelVA further employs a self-augmentation scheme to overcome the lack of mono audio track supervision. We evaluate SelVA on VGG-MONOAUDIO, a curated benchmark of clean single-source videos for such a task. Extensive experiments and ablations consistently verify its effectiveness across audio quality, semantic alignment, and temporal synchronization. Code and demo are available at https://jnwnlee.github.io/selva-demo/.
Authors: Gauri Deshpande, Harish Battula, Ashish Panda, Sunil Kumar Kopparapu
Abstract: This paper presents a unified study of four distinct modeling approaches for classifying dysarthria severity in the Speech Analysis for Neurodegenerative Diseases (SAND) challenge. All models tackle the same five class classification task using a common dataset of speech recordings. We investigate: (1) a ViT-OF method leveraging a Vision Transformer on spectrogram images, (2) a 1D-CNN approach using eight 1-D CNN's with majority-vote fusion, (3) a BiLSTM-OF approach using nine BiLSTM models with majority vote fusion, and (4) a Hierarchical XGBoost ensemble that combines glottal and formant features through a two stage learning framework. Each method is described, and their performances on a validation set of 53 speakers are compared. Results show that while the feature-engineered XGBoost ensemble achieves the highest macro-F1 (0.86), the deep learning models (ViT, CNN, BiLSTM) attain competitive F1-scores (0.70) and offer complementary insights into the problem.
Authors: Vedanta Thapar, Renaud Lambiotte, George T. Cantwell
Abstract: We propose the first return time distribution (FRTD) of a random walk as an interpretable and mathematically grounded node embedding. The FRTD assigns a probability mass function to each node, allowing us to define a distance between any pair of nodes using standard metrics for discrete distributions. We present several arguments to motivate the FRTD embedding. First, we show that FRTDs are strictly more informative than eigenvalue spectra, yet insufficient for complete graph identification, thus placing FRTD equivalence between cospectrality and isomorphism. Second, we argue that FRTD equivalence between nodes captures structural similarity. Third, we empirically demonstrate that the FRTD embedding outperforms manually designed graph metrics in network alignment tasks. Finally, we show that random networks that approximately match the FRTD of a desired target also preserve other salient features. Together these results demonstrate the FRTD as a simple and mathematically principled embedding for complex networks.
Authors: Omid Reza Heidari, Yang Wang, Xinxin Zuo
Abstract: Domain adaptation in object detection is critical for real-world applications where distribution shifts degrade model performance. Security X-ray imaging presents a unique challenge due to variations in scanning devices and environmental conditions, leading to significant domain discrepancies. To address this, we apply ALDI++, a domain adaptation framework that integrates self-distillation, feature alignment, and enhanced training strategies to mitigate domain shift effectively in this area. We conduct extensive experiments on the EDS dataset, demonstrating that ALDI++ surpasses the state-of-the-art (SOTA) domain adaptation methods across multiple adaptation scenarios. In particular, ALDI++ with a Vision Transformer for Detection (ViTDet) backbone achieves the highest mean average precision (mAP), confirming the effectiveness of transformer-based architectures for cross-domain object detection. Additionally, our category-wise analysis highlights consistent improvements in detection accuracy, reinforcing the robustness of the model across diverse object classes. Our findings establish ALDI++ as an efficient solution for domain-adaptive object detection, setting a new benchmark for performance stability and cross-domain generalization in security X-ray imagery.
Authors: Zhenkai Wu, Xiaowen Ma, Zhenliang Ni, Dengming Zhang, Han Shu, Xin Jiang, Xinghao Chen
Abstract: Vision-language models (VLMs) excel at image understanding tasks, but the large number of visual tokens imposes significant computational costs, hindering deployment on mobile devices. Many pruning methods rely solely on token importance and thus overlook inter-token redundancy, retaining numerous duplicated tokens and wasting capacity. Although some redundancy-aware approaches have been proposed, they often ignore the spatial relationships among visual tokens. This can lead to overly sparse selections of retained tokens that fail to adequately cover the regions of target objects. To address these limitations, we propose VLM-Pruner, a training-free token pruning algorithm that explicitly balances redundancy and spatial sparsity. We introduce a centrifugal token pruning paradigm that enables near-to-far selection while prioritizing the preservation of fine-grained object details. Moreover, we design a Buffering for Spatial Sparsity (BSS) criterion that defers the selection of spatially distant tokens. We further adopt a parallel greedy strategy to conduct token selection efficiently. To mitigate information loss from pruning, we selectively fuse salient information from the discarded tokens into the retained ones. Comprehensive comparisons demonstrate that VLM-Pruner consistently outperforms strong baselines across five VLMs with an 88.9\% pruning rate, while delivering an end-to-end inference speedup.
Authors: Lavish Bansal, Naman Mishra
Abstract: Ensuring content safety in large language models (LLMs) is essential for their deployment in real-world applications. However, existing safety guardrails are predominantly tailored for high-resource languages, leaving a significant portion of the world's population underrepresented who communicate in low-resource languages. To address this, we introduce CREST (CRoss-lingual Efficient Safety Transfer), a parameter-efficient multilingual safety classification model that supports 100 languages with only 0.5B parameters. By training on a strategically chosen subset of only 13 high-resource languages, our model utilizes cluster-based cross-lingual transfer from a few to 100 languages, enabling effective generalization to both unseen high-resource and low-resource languages. This approach addresses the challenge of limited training data in low-resource settings. We conduct comprehensive evaluations across six safety benchmarks to demonstrate that CREST outperforms existing state-of-the-art guardrails of comparable scale and achieves competitive results against models with significantly larger parameter counts (2.5B parameters and above). Our findings highlight the limitations of language-specific guardrails and underscore the importance of developing universal, language-agnostic safety systems that can scale effectively to serve global populations.
Authors: Julian Ma, Jun Wang, Zafeirios Fountas
Abstract: Large language models (LLMs) excel at explicit reasoning, but their implicit computational strategies remain underexplored. Decades of psychophysics research show that humans intuitively process and integrate noisy signals using near-optimal Bayesian strategies in perceptual tasks. We ask whether LLMs exhibit similar behaviour and perform optimal multimodal integration without explicit training or instruction. Adopting the psychophysics paradigm, we infer computational principles of LLMs from systematic behavioural studies. We introduce a behavioural benchmark - BayesBench: four magnitude estimation tasks (length, location, distance, and duration) over text and image, inspired by classic psychophysics, and evaluate a diverse set of nine LLMs alongside human judgments for calibration. Through controlled ablations of noise, context, and instruction prompts, we measure performance, behaviour and efficiency in multimodal cue-combination. Beyond accuracy and efficiency metrics, we introduce a Bayesian Consistency Score that detects Bayes-consistent behavioural shifts even when accuracy saturates. Our results show that while capable models often adapt in Bayes-consistent ways, accuracy does not guarantee robustness. Notably, GPT-5 Mini achieves perfect text accuracy but fails to integrate visual cues efficiently. This reveals a critical dissociation between capability and strategy, suggesting accuracy-centric benchmarks may over-index on performance while missing brittle uncertainty handling. These findings reveal emergent principled handling of uncertainty and highlight the correlation between accuracy and Bayesian tendencies. We release our psychophysics benchmark and consistency metric (https://bayes-bench.github.io) as evaluation tools and to inform future multimodal architecture designs.
Authors: He Wang, Wenyilin Xiao, Songqiao Han, Hailiang Huang
Abstract: Stock price prediction is challenging due to market volatility and its sensitivity to real-time events. While large language models (LLMs) offer new avenues for text-based forecasting, their application in finance is hindered by noisy news data and the lack of explicit answers in text. General-purpose memory architectures struggle to identify the key drivers of price movements. To address this, we propose StockMem, an event-reflection dual-layer memory framework. It structures news into events and mines them along two dimensions: horizontal consolidation integrates daily events, while longitudinal tracking captures event evolution to extract incremental information reflecting market expectation discrepancies. This builds a temporal event knowledge base. By analyzing event-price dynamics, the framework further forms a reflection knowledge base of causal experiences. For prediction, it retrieves analogous historical scenarios and reasons with current events, incremental data, and past experiences. Experiments show StockMem outperforms existing memory architectures and provides superior, explainable reasoning by tracing the information chain affecting prices, enhancing decision transparency in financial forecasting.
Authors: Mark M. Wilde
Abstract: Born-rule generative modeling, a central task in quantum machine learning, seeks to learn probability distributions that can be efficiently sampled by measuring complex quantum states. One hope is for quantum models to efficiently capture probability distributions that are difficult to learn and simulate by classical means alone. Quantum Boltzmann machines were proposed about one decade ago for this purpose, yet efficient training methods have remained elusive. In this paper, I overcome this obstacle by proposing a practical solution that trains quantum Boltzmann machines for Born-rule generative modeling. Two key ingredients in the proposal are the Donsker-Varadhan variational representation of the classical relative entropy and the quantum Boltzmann gradient estimator of [Patel et al., arXiv:2410.12935]. I present the main result for a more general ansatz known as an evolved quantum Boltzmann machine [Minervini et al., arXiv:2501.03367], which combines parameterized real- and imaginary-time evolution. I also show how to extend the findings to other distinguishability measures beyond relative entropy. Finally, I present four different hybrid quantum-classical algorithms for the minimax optimization underlying training, and I discuss their theoretical convergence guarantees.
Authors: Yifan Zhou, Takehiko Ohkawa, Guwenxiao Zhou, Kanoko Goto, Takumi Hirose, Yusuke Sekikawa, Nakamasa Inoue
Abstract: Modeling daily hand interactions often struggles with severe occlusions, such as when two hands overlap, which highlights the need for robust feature learning in 3D hand pose estimation (HPE). To handle such occluded hand images, it is vital to effectively learn the relationship between local image features (e.g., for occluded joints) and global context (e.g., cues from inter-joints, inter-hands, or the scene). However, most current 3D HPE methods still rely on ResNet for feature extraction, and such CNN's inductive bias may not be optimal for 3D HPE due to its limited capability to model the global context. To address this limitation, we propose an effective and efficient framework for visual feature extraction in 3D HPE using recent state space modeling (i.e., Mamba), dubbed Deformable Mamba (DF-Mamba). DF-Mamba is designed to capture global context cues beyond standard convolution through Mamba's selective state modeling and the proposed deformable state scanning. Specifically, for local features after convolution, our deformable scanning aggregates these features within an image while selectively preserving useful cues that represent the global context. This approach significantly improves the accuracy of structured 3D HPE, with comparable inference speed to ResNet-50. Our experiments involve extensive evaluations on five divergent datasets including single-hand and two-hand scenarios, hand-only and hand-object interactions, as well as RGB and depth-based estimation. DF-Mamba outperforms the latest image backbones, including VMamba and Spatial-Mamba, on all datasets and achieves state-of-the-art performance.
Authors: Przemyslaw Chojecki
Abstract: We extend the moduli-theoretic framework of psychometric batteries to the domain of dynamical systems. While previous work established the AAI capability score as a static functional on the space of agent representations, this paper formalizes the agent as a flow $\nu_r$ parameterized by computational resource $r$, governed by a recursive Generator-Verifier-Updater (GVU) operator. We prove that this operator generates a vector field on the parameter manifold $\Theta$, and we identify the coefficient of self-improvement $\kappa$ as the Lie derivative of the capability functional along this flow. The central contribution of this work is the derivation of the Variance Inequality, a spectral condition that is sufficient (under mild regularity) for the stability of self-improvement. We show that a sufficient condition for $\kappa > 0$ is that, up to curvature and step-size effects, the combined noise of generation and verification must be small enough. We then apply this formalism to unify the recent literature on Language Self-Play (LSP), Self-Correction, and Synthetic Data bootstrapping. We demonstrate that architectures such as STaR, SPIN, Reflexion, GANs and AlphaZero are specific topological realizations of the GVU operator that satisfy the Variance Inequality through filtration, adversarial discrimination, or grounding in formal systems.
Authors: Tristan Amadei, Enric Meinhardt-Llopis, Benedicte Bascle, Corentin Abgrall, Gabriele Facciolo
Abstract: Image-based localization in GNSS-denied environments is critical for UAV autonomy. Existing state-of-the-art approaches rely on matching UAV images to geo-referenced satellite images; however, they typically require large-scale, paired UAV-satellite datasets for training. Such data are costly to acquire and often unavailable, limiting their applicability. To address this challenge, we adopt a training paradigm that removes the need for UAV imagery during training by learning directly from satellite-view reference images. This is achieved through a dedicated augmentation strategy that simulates the visual domain shift between satellite and real-world UAV views. We introduce CAEVL, an efficient model designed to exploit this paradigm, and validate it on ViLD, a new and challenging dataset of real-world UAV images that we release to the community. Our method achieves competitive performance compared to approaches trained with paired data, demonstrating its effectiveness and strong generalization capabilities.
Authors: Xu Han, Biao Zhang, Xiangjun Tang, Xianzhi Li, Peter Wonka
Abstract: We present LumiX, a structured diffusion framework for coherent text-to-intrinsic generation. Conditioned on text prompts, LumiX jointly generates a comprehensive set of intrinsic maps (e.g., albedo, irradiance, normal, depth, and final color), providing a structured and physically consistent description of an underlying scene. This is enabled by two key contributions: 1) Query-Broadcast Attention, a mechanism that ensures structural consistency by sharing queries across all maps in each self-attention block. 2) Tensor LoRA, a tensor-based adaptation that parameter-efficiently models cross-map relations for efficient joint training. Together, these designs enable stable joint diffusion training and unified generation of multiple intrinsic properties. Experiments show that LumiX produces coherent and physically meaningful results, achieving 23% higher alignment and a better preference score (0.19 vs. -0.41) compared to the state of the art, and it can also perform image-conditioned intrinsic decomposition within the same framework.
Authors: Shyam prasad reddy Kaitha, Hongrui Yu
Abstract: Multi-robot task allocation in construction automation has traditionally relied on optimization methods such as Dynamic Programming and Reinforcement Learning. This research introduces the LangGraph-based Task Allocation Agent (LTAA), an LLM-driven framework that integrates phase-adaptive allocation strategies, multi-stage validation with hierarchical retries, and dynamic prompting for efficient robot coordination. Although recent LLM approaches show potential for construction robotics, they largely lack rigorous validation and benchmarking against established algorithms. This paper presents the first systematic comparison of LLM-based task allocation with traditional methods in construction scenarios.The study validates LLM feasibility through SMART-LLM replication and addresses implementation challenges using a Self-Corrective Agent Architecture. LTAA leverages natural-language reasoning combined with structured validation mechanisms, achieving major computational gains reducing token usage by 94.6% and allocation time by 86% through dynamic prompting. The framework adjusts its strategy across phases: emphasizing execution feasibility early and workload balance in later allocations.The authors evaluate LTAA against Dynamic Programming, Q-learning, and Deep Q-Network (DQN) baselines using construction operations from the TEACh human-robot collaboration dataset. In the Heavy Excels setting, where robots have strong task specializations, LTAA achieves 77% task completion with superior workload balance, outperforming all traditional methods. These findings show that LLM-based reasoning with structured validation can match established optimization algorithms while offering additional advantages such as interpretability, adaptability, and the ability to update task logic without retraining.
Authors: Ali Alvandi, Mina Rezaei
Abstract: Contrastive learning is among the most popular and powerful approaches for self-supervised representation learning, where the goal is to map semantically similar samples close together while separating dissimilar ones in the latent space. Existing theoretical methods assume that downstream task classes are drawn from the same latent class distribution used during the pretraining phase. However, in real-world settings, downstream tasks may not only exhibit distributional shifts within the same label space but also introduce new or broader label spaces, leading to domain generalization challenges. In this work, we introduce novel generalization bounds that explicitly account for both types of mismatch: domain shift and domain generalization. Specifically, we analyze scenarios where downstream tasks either (i) draw classes from the same latent class space but with shifted distributions, or (ii) involve new label spaces beyond those seen during pretraining. Our analysis reveals how the performance of contrastively learned representations depends on the statistical discrepancy between pretraining and downstream distributions. This extended perspective allows us to derive provable guarantees on the performance of learned representations on average classification tasks involving class distributions outside the pretraining latent class set.
Authors: Lechen Zhang, Yusheng Zhou, Tolga Ergen, Lajanugen Logeswaran, Moontae Lee, David Jurgens
Abstract: System prompts provide a lightweight yet powerful mechanism for conditioning large language models (LLMs) at inference time. While prior work has focused on English-only settings, real-world deployments benefit from having a single prompt to operate reliably across languages. This paper presents a comprehensive study of how different system prompts steer models toward accurate and robust cross-lingual behavior. We propose a unified four-dimensional evaluation framework to assess system prompts in multilingual environments. Through large-scale experiments on five languages, three LLMs, and three benchmarks, we uncover that certain prompt components, such as CoT, emotion, and scenario, correlate with robust multilingual behavior. We develop a prompt optimization framework for multilingual settings and show it can automatically discover prompts that improve all metrics by 5-10%. Finally, we analyze over 10 million reasoning units and find that more performant system prompts induce more structured and consistent reasoning patterns, while reducing unnecessary language-switching. Together, we highlight system prompt optimization as a scalable path to accurate and robust multilingual LLM behavior.
Authors: Xinzheng Wu, Junyi Chen, Naiting Zhong, Yong Shen
Abstract: The safe deployment of autonomous driving systems (ADSs) relies on comprehensive testing and evaluation. However, safety-critical scenarios that can effectively expose system vulnerabilities are extremely sparse in the real world. Existing scenario generation methods face challenges in efficiently constructing long-tail scenarios that ensure fidelity, criticality, and interactivity, while particularly lacking real-time dynamic response capabilities to the vehicle under test (VUT). To address these challenges, this paper proposes a safety-critical testing scenario generation framework that integrates the high-level semantic understanding capabilities of Vision Language Models (VLMs) with the fine-grained generation capabilities of adaptive guided diffusion models. The framework establishes a three-layer hierarchical architecture comprising a strategic layer for VLM-directed scenario generation objective determination, a tactical layer for guidance function formulation, and an operational layer for guided diffusion execution. We first establish a high-quality fundamental diffusion model that learns the data distribution of real driving scenarios. Next, we design an adaptive guided diffusion method that enables real-time, precise control of background vehicles (BVs) in closed-loop simulation. The VLM is then incorporated to autonomously generate scenario generation objectives and guidance functions through deep scenario understanding and risk reasoning, ultimately guiding the diffusion model to achieve VLM-directed scenario generation. Experimental results demonstrate that the proposed method can efficiently generate realistic, diverse, and highly interactive safety-critical testing scenarios. Furthermore, case studies validate the adaptability and VLM-directed generation performance of the proposed method.
Authors: Vishal Dubey, Pallavi Tyagi
Abstract: Modern image editors can produce identity-preserving AIGC (IP-AIGC), where the same person appears with new attire, background, or lighting. The robustness and fairness of current detectors in this regime remain unclear, especially for under-represented populations. We present what we believe is the first systematic study of IP-AIGC detection for Indian and South-Asian faces, quantifying cross-generator generalization and intra-population performance. We assemble Indian-focused training splits from FairFD and HAV-DF, and construct two held-out IP-AIGC test sets (HIDF-img-ip-genai and HIDF-vid-ip-genai) using commercial web-UI generators (Gemini and ChatGPT) with identity-preserving prompts. We evaluate two state-of-the-art detectors (AIDE and Effort) under pretrained (PT) and fine-tuned (FT) regimes and report AUC, AP, EER, and accuracy. Fine-tuning yields strong in-domain gains (for example, Effort AUC 0.739 to 0.944 on HAV-DF-test; AIDE EER 0.484 to 0.259), but consistently degrades performance on held-out IP-AIGC for Indian cohorts (for example, AIDE AUC 0.923 to 0.563 on HIDF-img-ip-genai; Effort 0.740 to 0.533), which indicates overfitting to training-generator cues. On non-IP HIDF images, PT performance remains high, which suggests a specific brittleness to identity-preserving edits rather than a generic distribution shift. Our study establishes IP-AIGC-Indian as a challenging and practically relevant scenario and motivates representation-preserving adaptation and India-aware benchmark curation to close generalization gaps in AIGC detection.
Authors: Roberto Miele, Leonardo Azevedo
Abstract: Probabilistic seismic inverse modeling often requires the prediction of both spatially correlated geological heterogeneities (e.g., facies) and continuous parameters (e.g., rock and elastic properties). Generative adversarial networks (GANs) provide an efficient training-image-based simulation framework capable of reproducing complex geological models with high accuracy and comparably low generative cost. However, their application in stochastic geophysical inversion for multivariate property prediction is limited, as representing multiple coupled properties requires large and unstable networks with high memory and training demands. A more recent variant of GANs with spatially adaptive denormalization (SPADE-GAN) enables the direct conditioning of facies spatial distributions on local probability maps. Leveraging on such features, an iterative geostatistical inversion algorithm is proposed, SPADE-GANInv, integrating a pre-trained SPADE-GAN with geostatistical simulation, for the prediction of facies and multiple correlated continuous properties from seismic data. The SPADE-GAN is trained to reproduce realistic facies geometries, while sequential stochastic co-simulation predicts the spatial variability of the facies-dependent continuous properties. At each iteration, a set of subsurface realizations is generated and used to compute synthetic seismic data. The realizations providing the highest similarity coefficient to the observed data are used to update the subsurface probability models in the next iteration. The method is demonstrated on both 2-D synthetic scenarios and field data, targeting the prediction of facies, porosity, and acoustic impedance from full-stack seismic data. Results show that the algorithm enables accurate multivariate prediction, mitigates the impact of biased prior data, and accommodates additional local conditioning such as well logs.
Authors: Weiqi Li, Quande Zhang, Ruifeng Zhai, Liang Lin, Guangrun Wang
Abstract: Vision-language-action (VLA) models achieve strong in-distribution performance but degrade sharply under novel camera viewpoints and visual perturbations. We show that this brittleness primarily arises from misalignment in Spatial Modeling, rather than Physical Modeling. To address this, we propose a one-shot adaptation framework that recalibrates visual representations through lightweight, learnable updates. Our first method, Feature Token Modulation (FTM), applies a global affine transformation to visual tokens and improves Libero viewpoint accuracy from 48.5% to 87.1% with only 4K parameters. Building on this, Feature Linear Adaptation (FLA) introduces low-rank updates to the ViT encoder, achieving 90.8% success with 4.7M parameters -- matching LoRA-scale finetuning at far lower cost. Together, these results reveal substantial untapped robustness in pretrained VLA models and demonstrate that targeted, minimal visual adaptation is sufficient to restore viewpoint generalization.
Authors: Zhonghao He, Tianyi Qiu, Hirokazu Shirado, Maarten Sap
Abstract: Recent advances in reasoning techniques have substantially improved the performance of large language models (LLMs), raising expectations for their ability to provide accurate, truthful, and reliable information. However, emerging evidence suggests that iterative reasoning may foster belief entrenchment and confirmation bias, rather than enhancing truth-seeking behavior. In this study, we propose a systematic evaluation framework for belief entrenchment in LLM reasoning by leveraging the Martingale property from Bayesian statistics. This property implies that, under rational belief updating, the expected value of future beliefs should remain equal to the current belief, i.e., belief updates are unpredictable from the current belief. We propose the unsupervised, regression-based Martingale Score to measure violations of this property, which signal deviation from the Bayesian ability of updating on new evidence. In open-ended problem domains including event forecasting, value-laden questions, and academic paper review, we find such violations to be widespread across models and setups, where the current belief positively predicts future belief updates, a phenomenon which we term belief entrenchment. We identify the models, reasoning techniques, and domains more prone to belief entrenchment. Finally, we validate the Martingale Score by showing that it predicts ground-truth accuracy on problem domains where ground truth labels are available. This indicates that, while designed as an unsupervised metric that operates even in domains without access to ground truth, the Martingale Score is a useful proxy of the truth-seeking ability of a reasoning process.
Authors: Samuel Andrello, Daniel Alabi, Simon J. L. Billinge
Abstract: While machine learning has enabled the rapid prediction of inorganic materials with novel properties, the challenge of determining how to synthesize these materials remains largely unsolved. Previous work has largely focused on predicting precursors or reaction conditions, but only rarely on full synthesis pathways. We introduce the ActionGraph, a directed acyclic graph framework that encodes both the chemical and procedural structure, in terms of synthesis operations, of inorganic synthesis reactions. Using 13,017 text-mined solid-state synthesis reactions from the Materials Project, we show that incorporating PCA-reduced ActionGraph adjacency matrices into a $k$-nearest neighbors retrieval model significantly improves synthesis pathway prediction. While the ActionGraph framework only results in a 1.34% and 2.76% increase in precursor and operation F1 scores (average over varying numbers of PCA components) respectively, the operation length matching accuracy rises 3.4 times (from 15.8% to 53.3%). We observe an interesting trade-off where precursor prediction performance peaks at 10-11 PCA components while operation prediction continues improving up to 30 components. This suggests composition information dominates precursor selection while structural information is critical for operation sequencing. Overall, the ActionGraph framework demonstrates strong potential, and with further adoption, its full range of benefits can be effectively realized.
Authors: Annalena Kofler, Maximilian Dax, Stephen R. Green, Jonas Wildberger, Nihar Gupte, Jakob H. Macke, Jonathan Gair, Alessandra Buonanno, Bernhard Sch\"olkopf
Abstract: Gravitational-wave data analysis relies on accurate and efficient methods to extract physical information from noisy detector signals, yet the increasing rate and complexity of observations represent a growing challenge. Deep learning provides a powerful alternative to traditional inference, but existing neural models typically lack the flexibility to handle variations in data analysis settings. Such variations accommodate imperfect observations or are required for specialized tests, and could include changes in detector configurations, overall frequency ranges, or localized cuts. We introduce a flexible transformer-based architecture paired with a training strategy that enables adaptation to diverse analysis settings at inference time. Applied to parameter estimation, we demonstrate that a single flexible model -- called Dingo-T1 -- can (i) analyze 48 gravitational-wave events from the third LIGO-Virgo-KAGRA Observing Run under a wide range of analysis configurations, (ii) enable systematic studies of how detector and frequency configurations impact inferred posteriors, and (iii) perform inspiral-merger-ringdown consistency tests probing general relativity. Dingo-T1 also improves median sample efficiency on real events from a baseline of 1.4% to 4.2%. Our approach thus demonstrates flexible and scalable inference with a principled framework for handling missing or incomplete data -- key capabilities for current and next-generation observatories.
Authors: Paul Barbaste (Inclusive Brains, Wavestone, Human Technology Foundation), Olivier Oullier (Human-Computer Interaction Department, Mohamed bin Zayed University of Artificial Intelligence, Inclusive Brains, Institute for Artificial Intelligence, Biotech Dental Group), Xavier Vasques (IBM Technology, IBM France Lab)
Abstract: Robust decoding and classification of brain patterns measured with electroencephalography (EEG) remains a major challenge for real-world (i.e. outside scientific lab and medical facilities) brain-computer interface (BCI) applications due to well documented inter- and intra-participant variability. Here, we present a large-scale benchmark evaluating over 340,000+ unique combinations of spatial and nonlinear EEG classification. Our methodological pipeline consists in combinations of Common Spatial Patterns (CSP), Riemannian geometry, functional connectivity, and fractal- or entropy-based features across three open-access EEG datasets. Unlike prior studies, our analysis operates at the per-participant level and across multiple frequency bands (8-15 Hz and 8-30 Hz), enabling direct assessment of both group-level performance and individual variability. Covariance tangent space projection (cov-tgsp) and CSP consistently achieved the highest average classification accuracies. However, their effectiveness was strongly dataset-dependent, and marked participant-level differences persisted, particularly in the most heterogeneous of the datasets. Importantly, nonlinear methods outperformed spatial approaches for specific individuals, underscoring the need for personalized pipeline selection. Our findings highlight that no universal 'one-size-fits-all' method can optimally decode EEG motor imagery patterns across all users or datasets. Future work will require adaptive, multimodal, and possibly novel approaches to fully address neurophysiological variability in practical BCI applications where the system can automatically adapt to what makes each user unique.
Authors: Chenji Lu (Alibaba Group), Zhuo Chen (Alibaba Group), Hui Zhao (Alibaba Group), Zhiyuan Zeng (Alibaba Group), Gang Zhao (Alibaba Group), Junjie Ren (Alibaba Group), Ruicong Xu (Alibaba Group), Haoran Li (Alibaba Group), Songyan Liu (Alibaba Group), Pengjie Wang (Alibaba Group), Jian Xu (Alibaba Group), Bo Zheng (Alibaba Group)
Abstract: Achievement. We introduce LORE, a systematic framework for Large Generative Model-based relevance in e-commerce search. Deployed and iterated over three years, LORE achieves a cumulative +27\% improvement in online GoodRate metrics. This report shares the valuable experience gained throughout its development lifecycle, spanning data, features, training, evaluation, and deployment. Insight. While existing works apply Chain-of-Thought (CoT) to enhance relevance, they often hit a performance ceiling. We argue this stems from treating relevance as a monolithic task, lacking principled deconstruction. Our key insight is that relevance comprises distinct capabilities: knowledge and reasoning, multi-modal matching, and rule adherence. We contend that a qualitative-driven decomposition is essential for breaking through current performance bottlenecks. Contributions. LORE provides a complete blueprint for the LLM relevance lifecycle. Key contributions include: (1) A two-stage training paradigm combining progressive CoT synthesis via SFT with human preference alignment via RL. (2) A comprehensive benchmark, RAIR, designed to evaluate these core capabilities. (3) A query frequency-stratified deployment strategy that efficiently transfers offline LLM capabilities to the online system. LORE serves as both a practical solution and a methodological reference for other vertical domains.
Authors: Ibrahim Laiche, Mokrane Boudaoud, Patrick Gallinari, Pascal Morin
Abstract: This article investigates the modeling and control of Lagrangian systems involving non-conservative forces using a hybrid method that does not require acceleration calculations. It focuses in particular on the derivation and identification of physically consistent models, which are essential for model-based control synthesis. Lagrangian or Hamiltonian neural networks provide useful structural guarantees but the learning of such models often leads to inconsistent models, especially on real physical systems where training data are limited, partial and noisy. Motivated by this observation and the objective to exploit these models for model-based nonlinear control, a learning algorithm relying on an original loss function is proposed to improve the physical consistency of Lagrangian systems. A comparative analysis of different learning-based modeling approaches with the proposed solution shows significant improvements in terms of physical consistency of the learned models, on both simulated and experimental systems. The model's consistency is then exploited to demonstrate, on an experimental benchmark, the practical relevance of the proposed methodology for feedback linearization and energy-based control techniques.
Authors: Tong Li, Jacob Nogas, Haochen Song, Anna Rafferty, Eric M. Schwartz, Audrey Durand, Harsh Kumar, Nina Deliu, Sofia S. Villar, Dehan Kong, Joseph J. Williams
Abstract: Traditional randomized A/B experiments assign arms with uniform random (UR) probability, such as 50/50 assignment to two versions of a website to discover whether one version engages users more. To more quickly and automatically use data to benefit users, multi-armed bandit algorithms such as Thompson Sampling (TS) have been advocated. While TS is interpretable and incorporates the randomization key to statistical inference, it can cause biased estimates and increase false positives and false negatives in detecting differences in arm means. We introduce a more Statistically Sensitive algorithm, TS-PostDiff (Posterior Probability of Small Difference), that mixes TS with traditional UR by using an additional adaptive step, where the probability of using UR (vs TS) is proportional to the posterior probability that the difference in arms is small. This allows an experimenter to define what counts as a small difference, below which a traditional UR experiment can obtain informative data for statistical inference at low cost, and above which using more TS to maximize user benefits is key. We evaluate TS-PostDiff against UR, TS, and two other TS variants designed to improve statistical inference. We consider results for the common two-armed experiment across a range of settings inspired by real-world applications. Our results provide insight into when and why TS-PostDiff or alternative approaches provide better tradeoffs between benefiting users (reward) and statistical inference (false positive rate and power). TS-PostDiff's adaptivity helps efficiently reduce false positives and increase statistical power when differences are small, while increasing reward more when differences are large. The work highlights important considerations for future Statistically Sensitive algorithm development that balances reward and statistical analysis in adaptive experimentation.
Authors: Kevin Han Huang, Peter Orbanz, Morgane Austern
Abstract: We provide universality results that quantify how data augmentation affects the variance and limiting distribution of estimates through simple surrogates, and analyze several specific models in detail. The results confirm some observations made in machine learning practice, but also lead to unexpected findings: Data augmentation may increase rather than decrease the uncertainty of estimates, such as the empirical prediction risk. It can act as a regularizer, but fails to do so in certain high-dimensional problems, and it may shift the double-descent peak of an empirical risk. Overall, the analysis shows that several properties data augmentation has been attributed with are not either true or false, but rather depend on a combination of factors -- notably the data distribution, the properties of the estimator, and the interplay of sample size, number of augmentations, and dimension. As our main theoretical tool, we develop an adaptation of Lindeberg's technique for block dependence. The resulting universality regime may be Gaussian or non-Gaussian.
Authors: Petr Philonenko, Vladimir Kokh, Pavel Blinov
Abstract: Conventional medical cancer screening methods are costly, labor-intensive, and extremely difficult to scale. Although AI can improve cancer detection, most systems rely on complex or specialized medical data, making them impractical for large-scale screening. We introduce Can-SAVE, a lightweight AI system that ranks population-wide cancer risks solely based on medical history events. By integrating survival model outputs into a gradient-boosting framework, our approach detects subtle, long-term patient risk patterns - often well before clinical symptoms manifest. Can-SAVE was rigorously evaluated on a real-world dataset of 2.5 million adults spanning five Russian regions, marking the study as one of the largest and most comprehensive deployments of AI-driven cancer risk assessment. In a retrospective oncologist-supervised study over 1.9M patients, Can-SAVE achieves a 4-10x higher detection rate at identical screening volumes and an Average Precision (AP) of 0.228 vs. 0.193 for the best baseline (LoRA-tuned Qwen3-Embeddings via DeepSeek-R1 summarization). In a year-long prospective pilot (426K patients), our method almost doubled the cancer detection rate (+91%) and increased population coverage by 36% over the national screening protocol. The system demonstrates practical scalability: a city-wide population of 1 million patients can be processed in under three hours using standard hardware, enabling seamless clinical integration. This work proves that Can-SAVE achieves nationally significant cancer detection improvements while adhering to real-world public healthcare constraints, offering immediate clinical utility and a replicable framework for population-wide screening. Code for training and feature engineering is available at https://github.com/sb-ai-lab/Can-SAVE.
Authors: Max Zimmer, Megi Andoni, Christoph Spiegel, Sebastian Pokutta
Abstract: Neural Networks can be effectively compressed through pruning, significantly reducing storage and compute demands while maintaining predictive performance. Simple yet effective methods like magnitude pruning remove less important parameters and typically require a costly retraining procedure to restore performance. However, with the rise of LLMs, full retraining has become infeasible due to memory and compute constraints. This study challenges the practice of retraining all parameters by showing that updating a small subset of highly expressive parameters can suffice to recover or even enhance performance after pruning. Surprisingly, retraining just 0.01%-0.05% of the parameters in GPT-architectures can match the performance of full retraining across various sparsity levels, significantly reducing compute and memory requirements, and enabling retraining of models with up to 30 billion parameters on a single GPU in minutes. To bridge the gap to full retraining in the high sparsity regime, we introduce two novel LoRA variants that, unlike standard LoRA, allow merging adapters back without compromising sparsity. Going a step further, we show that these methods can be applied for memory-efficient layer-wise reconstruction, significantly enhancing state-of-the-art retraining-free methods like Wanda (Sun et al., 2023) and SparseGPT (Frantar & Alistarh, 2023). Our findings present a promising alternative to avoiding retraining.
Authors: Anastasis Kratsios, Takashi Furuya, Jose Antonio Lara Benitez, Matti Lassas, Maarten de Hoop
Abstract: We study the approximation-theoretic implications of mixture-of-experts architectures for operator learning, where the complexity of a single large neural operator is distributed across many small neural operators (NOs), and each input is routed to exactly one NO via a decision tree. We analyze how this tree-based routing and expert decomposition affect approximation power, sample complexity, and stability. Our main result is a distributed universal approximation theorem for mixture of neural operators (MoNOs): any Lipschitz nonlinear operator between $L^2([0,1]^d)$ spaces can be uniformly approximated over the Sobolev unit ball to arbitrary accuracy $\varepsilon>0$ by an MoNO, where each expert NO has a depth, width, and rank scaling as $\mathcal{O}(\varepsilon^{-1})$. Although the number of experts may grow with accuracy, each NO remains small, enough to fit within active memory of standard hardware for reasonable accuracy levels. Our analysis also yields new quantitative approximation rates for classical NOs approximating uniformly continuous nonlinear operators uniformly on compact subsets of $L^2([0,1]^d)$.
Authors: Peiman Mohseni, Nick Duffield
Abstract: Neural processes (NPs) are probabilistic meta-learning models that map sets of observations to posterior predictive distributions, enabling inference at arbitrary domain points. Their capacity to handle variable-sized collections of unstructured observations, combined with simple maximum-likelihood training and uncertainty-aware predictions, makes them well-suited for modeling data over continuous domains. Since their introduction, several variants have been proposed. Early approaches typically represented observed data using finite-dimensional summary embeddings obtained through aggregation schemes such as mean pooling. However, this strategy fundamentally mismatches the infinite-dimensional nature of the generative processes that NPs aim to capture. Convolutional conditional neural processes (ConvCNPs) address this limitation by constructing infinite-dimensional functional embeddings processed through convolutional neural networks (CNNs) to enforce translation equivariance. Yet CNNs with local spatial kernels struggle to capture long-range dependencies without resorting to large kernels, which impose significant computational costs. To overcome this limitation, we propose the Spectral ConvCNP (SConvCNP), which performs global convolution in the frequency domain. Inspired by Fourier neural operators (FNOs) for learning solution operators of partial differential equations (PDEs), our approach directly parameterizes convolution kernels in the frequency domain, leveraging the relatively compact yet global Fourier representation of many natural signals. We validate the effectiveness of SConvCNP on both synthetic and real-world datasets, demonstrating how ideas from operator learning can advance the capabilities of NPs.
Authors: Du Yin, Hao Xue, Arian Prabowo, Shuang Ao, Flora Salim
Abstract: Traffic forecasting is crucial for smart cities and intelligent transportation initiatives, where deep learning has made significant progress in modeling complex spatio-temporal patterns in recent years. However, current public datasets have limitations in reflecting the distribution shift nature of real-world scenarios, characterized by continuously evolving infrastructures, varying temporal distributions, and long temporal gaps due to sensor downtimes or changes in traffic patterns. These limitations inevitably restrict the practical applicability of existing traffic forecasting datasets. To bridge this gap, we present XXLTraffic, largest available public traffic dataset with the longest timespan collected from Los Angeles, USA, and New South Wales, Australia, curated to support research in extremely long forecasting beyond test adaptation. Our benchmark includes both typical time-series forecasting settings with hourly and daily aggregated data and novel configurations that introduce gaps and down-sample the training size to better simulate practical constraints. We anticipate the new XXLTraffic will provide a fresh perspective for the time-series and traffic forecasting communities. It would also offer a robust platform for developing and evaluating models designed to tackle the extremely long forecasting problems beyond test adaptation. Our dataset supplements existing spatio-temporal data resources and leads to new research directions in this domain.
Authors: Yu Yang, Pan Xu
Abstract: Decision Transformer (DT) has emerged as a promising class of algorithms in offline reinforcement learning (RL) tasks, leveraging pre-collected datasets and Transformer's capability to model long sequences. Recent works have demonstrated that using parts of trajectories from training tasks as prompts in DT enhances its performance on unseen tasks, giving rise to Prompt-DT methods. However, collecting data from specific environments can be both costly and unsafe in many scenarios, leading to suboptimal performance and limited few-shot prompt abilities due to the data-hungry nature of Transformer-based models. Additionally, the limited datasets used in pre-training make it challenging for Prompt-DT type of methods to distinguish between various RL tasks through prompts alone. To address these challenges, we introduce the Language model-initialized Prompt Decision Transformer (LPDT) framework, which leverages pretrained language models providing rich prior knowledge for RL tasks and fine-tunes the sequence model using Low-rank Adaptation (LoRA) for meta-RL problems. We further incorporate prompt regularization to effectively differentiate between tasks based on prompt feature representations. Comprehensive empirical studies demonstrate that initializing with a pre-trained language model provides the prior knowledge and achieves a similar performance with Prompt-DT under only $10\%$ data in some MuJoCo control tasks. We also provide a thorough ablation study to validate the effectiveness of each component, including sequence modeling, language models, prompt regularizations, and prompt strategies.
Authors: Anuran Makur, Japneet Singh
Abstract: The Bradley-Terry-Luce (BTL) model is one of the most widely used models for ranking a collection of items or agents based on pairwise comparisons among them. Given $n$ agents, the BTL model endows each agent $i$ with a latent skill score $\alpha_i > 0$ and posits that the probability that agent $i$ is preferred over agent $j$ is $\alpha_i/(\alpha_i + \alpha_j)$. In this work, our objective is to formulate a hypothesis test that determines whether a given pairwise comparison dataset, with $k$ comparisons per pair of agents, originates from an underlying BTL model. We formalize this testing problem in the minimax sense and define the critical threshold of the problem. We then establish upper bounds on the critical threshold for general induced observation graphs (satisfying mild assumptions) and develop lower bounds for complete induced graphs. Our bounds demonstrate that for complete induced graphs, the critical threshold scales as $\Theta((nk)^{-1/2})$ in a minimax sense. In particular, our test statistic for the upper bounds is based on a new approximation we derive for the separation distance between general pairwise comparison models and the class of BTL models. To further assess the performance of our statistical test, we prove upper bounds on the type I and type II probabilities of error. Much of our analysis is conducted within the context of a fixed observation graph structure, where the graph possesses certain ``nice'' properties, such as expansion and bounded principal ratio. Additionally, we derive several auxiliary results, such as bounds on principal ratios of graphs, $\ell^2$-bounds on BTL parameter estimation under model mismatch, stability of rankings under the BTL model, etc. We validate our theoretical results through experiments on synthetic and real-world datasets and propose a data-driven permutation testing approach to determine test thresholds.
Authors: Shashank Hegde, Satyajeet Das, Gautam Salhotra, Gaurav S. Sukhatme
Abstract: With the increasing availability of open-source robotic data, imitation learning has become a promising approach for both manipulation and locomotion. Diffusion models are now widely used to train large, generalized policies that predict controls or trajectories, leveraging their ability to model multimodal action distributions. However, this generality comes at the cost of larger model sizes and slower inference, an acute limitation for robotic tasks requiring high control frequencies. Moreover, Diffusion Policy (DP), a popular trajectory-generation approach, suffers from a trade-off between performance and action horizon: fewer diffusion queries lead to larger trajectory chunks, which in turn accumulate tracking errors. To overcome these challenges, we introduce WARPD (World model Assisted Reactive Policy Diffusion), a method that generates closed-loop policies (weights for neural policies) directly, instead of open-loop trajectories. By learning behavioral distributions in parameter space rather than trajectory space, WARPD offers two major advantages: (1) extended action horizons with robustness to perturbations, while maintaining high task performance, and (2) significantly reduced inference costs. Empirically, WARPD outperforms DP in long-horizon and perturbed environments, and achieves multitask performance on par with DP while requiring only ~ 1/45th of the inference-time FLOPs per step.
Authors: Arham Khan, Robert Underwood, Carlo Siebenschuh, Yadu Babuji, Aswathy Ajith, Kyle Hippe, Ozan Gokdemir, Alexander Brace, Kyle Chard, Ian Foster
Abstract: Contemporary large language model (LLM) training pipelines require the assembly of internet-scale databases full of text data from a variety of sources (e.g., web, academic, and publishers). Preprocessing these datasets via deduplication -- detecting and eliminating additional instances of the same content -- is a major focus for assembling and curating training datasets for LLMs. Unrestrained, duplicates in the training dataset increase training costs and lead to undesirable properties such as memorization in trained models or cheating on evaluation. Unfortunately, contemporary approaches to document-level deduplication are either unreliable at accurately identifying duplicate documents or extremely expensive in terms of both runtime and memory. We propose LSHBloom, an extension to MinhashLSH, which replaces the expensive LSHIndex with lightweight Bloom filters. LSHBloom demonstrates the same state-of-the-art deduplication performance as MinhashLSH, with only a marginal increase in false positives (near zero in our experiments), while boasting competitive runtime (12$\times$ faster than MinhashLSH on peS2o) and, crucially, using 18$\times$ less disk space than MinhashLSH (as measured on peS2o). Based on extrapolation, we show that this advantage in space and runtime remains even at the extreme scale of several billion documents. LSHBloom allows practitioners to access the deduplication quality of MinHashLSH at scales that are normally only tractable for less sophisticated, heuristic solutions. As a result, LSHBloom promises to enable scaling high-quality document deduplication to internet-scale text datasets.
Authors: Mattia Giovanni Campana, Franca Delmastro
Abstract: Personalized Federated Learning aims at addressing the challenges of non-IID data in collaborative model training. However, existing methods struggle to balance personalization and generalization, often oversimplifying client similarities or relying too heavily on global models. In this paper, we propose FedSub, a novel approach that introduces class-aware model updates based on data prototypes and model subnetworks fusion to enhance personalization. Prototypes serve as compact representations of client data for each class, clustered on the server to capture label-specific similarities among the clients. Meanwhile, model subnetworks encapsulate the most relevant components to process each class and they are then fused on the server based on the identified clusters to generate fine-grained, class-specific, and highly personalized model updates for each client. Experimental results in three real-world scenarios with high data heterogeneity in human activity recognition and mobile health applications demonstrate the effectiveness of FedSub with respect to state-of-the-art methods to achieve fast convergence and high classification performance.
Authors: Felipe Maia Polo, Seamus Somerstep, Leshem Choshen, Yuekai Sun, Mikhail Yurochkin
Abstract: Scaling laws for large language models (LLMs) predict model performance based on parameters like size and training data. However, differences in training configurations and data processing across model families lead to significant variations in benchmark performance, making it difficult for a single scaling law to generalize across all LLMs. On the other hand, training family-specific scaling laws requires training models of varying sizes for every family. In this work, we propose Skills Scaling Laws (SSLaws, pronounced as Sloth), a novel scaling law that leverages publicly available benchmark data and assumes LLM performance is driven by low-dimensional latent skills, such as reasoning and instruction following. These latent skills are influenced by computational resources like model size and training tokens, but with varying efficiencies across model families. Sloth exploits correlations across benchmarks to provide more accurate and interpretable predictions while alleviating the need to train multiple LLMs per family. We present both theoretical results on parameter identification and empirical evaluations on 12 prominent benchmarks, from Open LLM Leaderboard v1/v2, demonstrating that Sloth predicts LLM performance accurately and offers insights into scaling behaviors for complex downstream tasks, increased test-time compute, and compute-optimal scaling of skills.
Authors: Zijian Zhao, Fanyi Meng, Zhonghao Lyu, Hang Li, Xiaoyang Li, Guangxu Zhu
Abstract: Channel state information (CSI) is a fundamental component in both wireless communication and sensing systems, enabling critical functions such as radio resource optimization and environmental perception. In wireless sensing, data scarcity and packet loss hinder efficient model training, while in wireless communication, high-dimensional CSI matrices and short coherent times caused by high mobility present challenges in CSI estimation. To address these issues, we propose a unified framework named CSI-BERT2 for CSI prediction and classification tasks, built on CSI-BERT, which adapts BERT to capture the complex relationships among CSI sequences through a bidirectional self-attention mechanism. We introduce a two-stage training method that first uses a mask language model (MLM) to enable the model to learn general feature extraction from scarce datasets in an unsupervised manner, followed by fine-tuning for specific downstream tasks. Specifically, we extend MLM into a mask prediction model (MPM), which efficiently addresses the CSI prediction task. To further enhance the representation capacity of CSI data, we modify the structure of the original CSI-BERT. We introduce an adaptive re-weighting layer (ARL) to enhance subcarrier representation and a multi-layer perceptron (MLP)-based temporal embedding module to mitigate temporal information loss problem inherent in the original Transformer. Extensive experiments on both real-world collected and simulated datasets demonstrate that CSI-BERT2 achieves state-of-the-art performance across all tasks. Our results further show that CSI-BERT2 generalizes effectively across varying sampling rates and robustly handles discontinuous CSI sequences caused by packet loss-challenges that conventional methods fail to address. The dataset and code are publicly available at https://github.com/RS2002/CSI-BERT2 .
Authors: Shizhou Xu, Thomas Strohmer
Abstract: How can we effectively remove or ''unlearn'' undesirable information, such as specific features or the influence of individual data points, from a learning outcome while minimizing utility loss and ensuring rigorous guarantees? We introduce a unified mathematical framework based on information-theoretic regularization to address both data point unlearning and feature unlearning. For data point unlearning, we introduce the $\textit{Marginal Unlearning Principle}$, an auditable and provable framework inspired by memory suppression studies in neuroscience. Moreover, we provide formal information-theoretic unlearning definition based on the proposed principle, named marginal unlearning, and provable guarantees on sufficiency and necessity of marginal unlearning to the existing approximate unlearning definitions. We then show the proposed framework provide natural solution to the marginal unlearning problems. For feature unlearning, the framework applies to deep learning with arbitrary training objectives. By combining flexibility in learning objectives with simplicity in regularization design, our approach is highly adaptable and practical for a wide range of machine learning and AI applications. From a mathematical perspective, we provide an unified analytic solution to the optimal feature unlearning problem with a variety of information-theoretic training objectives. Our theoretical analysis reveals intriguing connections between machine unlearning, information theory, optimal transport, and extremal sigma algebras. Numerical simulations support our theoretical finding.
Authors: Michael Arbel, David Salinas, Frank Hutter
Abstract: Recent foundational models for tabular data, such as TabPFN, excel at adapting to new tasks via in-context learning, but remain constrained to a fixed, pre-defined number of target dimensions-often necessitating costly ensembling strategies. We trace this constraint to a deeper architectural shortcoming: these models lack target equivariance, so that permuting target dimension orderings alters their predictions. This deficiency gives rise to an irreducible "equivariance gap", an error term that introduces instability in predictions. We eliminate this gap by designing a fully target-equivariant architecture-ensuring permutation invariance via equivariant encoders, decoders, and a bi-attention mechanism. Empirical evaluation on standard classification benchmarks shows that, on datasets with more classes than those seen during pre-training, our model matches or surpasses existing methods while incurring lower computational overhead.
Authors: Yuxuan Zhou, Heng Li, Zhi-Qi Cheng, Xudong Yan, Yifei Dong, Mario Fritz, Margret Keuper
Abstract: Label Smoothing (LS) is widely adopted to reduce overconfidence in neural network predictions and improve generalization. Despite these benefits, recent studies reveal two critical issues with LS. First, LS induces overconfidence in misclassified samples. Second, it compacts feature representations into overly tight clusters, diluting intra-class diversity, although the precise cause of this phenomenon remained elusive. In this paper, we analytically decompose the LS-induced loss, exposing two key terms: (i) a regularization term that dampens overconfidence only when the prediction is correct, and (ii) an error-amplification term that arises under misclassifications. This latter term compels the network to reinforce incorrect predictions with undue certainty, exacerbating representation collapse. To address these shortcomings, we propose Max Suppression (MaxSup), which applies uniform regularization to both correct and incorrect predictions by penalizing the top-1 logit rather than the ground-truth logit. Through extensive feature-space analyses, we show that MaxSup restores intra-class variation and sharpens inter-class boundaries. Experiments on large-scale image classification and multiple downstream tasks confirm that MaxSup is a more robust alternative to LS. Code is available at: https://github.com/ZhouYuxuanYX/Maximum-Suppression-Regularization
URLs: https://github.com/ZhouYuxuanYX/Maximum-Suppression-Regularization
Authors: Xiangwen Wang, Yibo Jacky Zhang, Zhoujie Ding, Katherine Tsai, Haolun Wu, Sanmi Koyejo
Abstract: Compound AI systems, comprising multiple interacting components such as LLMs, foundation models, and external tools, have demonstrated remarkable improvements compared to single models in various tasks. To ensure their effective deployment in real-world applications, aligning these systems with human preferences is crucial. However, aligning the compound system via policy optimization, unlike the alignment of a single model, is challenging for two main reasons: (i) non-differentiable interactions between components make end-to-end gradient-based optimization method inapplicable, and (ii) system-level preferences cannot be directly transformed into component-level preferences. To address these challenges, we first formulate compound AI systems as Directed Acyclic Graphs (DAGs), explicitly modeling both component interactions and the associated data flows. Building on this formulation, we introduce $\textbf{SysDPO}$, a framework that extends Direct Preference Optimization (DPO) to enable joint system-level alignment. We propose two variants, SysDPO-Direct and SysDPO-Sampling, tailored for scenarios depending on whether we construct a system-specific preference dataset. We empirically demonstrate the effectiveness of our approach across two applications: the joint alignment of a language model and a diffusion model, and the joint alignment of an LLM collaboration system.
Authors: Sai Sumedh R. Hindupur, Ekdeep Singh Lubana, Thomas Fel, Demba Ba
Abstract: Sparse Autoencoders (SAEs) are widely used to interpret neural networks by identifying meaningful concepts from their representations. However, do SAEs truly uncover all concepts a model relies on, or are they inherently biased toward certain kinds of concepts? We introduce a unified framework that recasts SAEs as solutions to a bilevel optimization problem, revealing a fundamental challenge: each SAE imposes structural assumptions about how concepts are encoded in model representations, which in turn shapes what it can and cannot detect. This means different SAEs are not interchangeable -- switching architectures can expose entirely new concepts or obscure existing ones. To systematically probe this effect, we evaluate SAEs across a spectrum of settings: from controlled toy models that isolate key variables, to semi-synthetic experiments on real model activations and finally to large-scale, naturalistic datasets. Across this progression, we examine two fundamental properties that real-world concepts often exhibit: heterogeneity in intrinsic dimensionality (some concepts are inherently low-dimensional, others are not) and nonlinear separability. We show that SAEs fail to recover concepts when these properties are ignored, and we design a new SAE that explicitly incorporates both, enabling the discovery of previously hidden concepts and reinforcing our theoretical insights. Our findings challenge the idea of a universal SAE and underscores the need for architecture-specific choices in model interpretability. Overall, we argue an SAE does not just reveal concepts -- it determines what can be seen at all.
Authors: El-Mehdi El Arar (PEQUAN), Silviu-Ioan Filip (TARAN), Theo Mary (PEQUAN), Elisa Riccietti (OCKHAM)
Abstract: This work proposes a mathematically founded mixed precision accumulation strategy for the inference of neural networks. Our strategy is based on a new componentwise forward error analysis that explains the propagation of errors in the forward pass of neural networks. Specifically, our analysis shows that the error in each component of the output of a linear layer is proportional to the condition number of the inner product between the weights and the input, multiplied by the condition number of the activation function. These condition numbers can vary widely from one component to the other, thus creating a significant opportunity to introduce mixed precision: each component should be accumulated in a precision inversely proportional to the product of these condition numbers. We propose a numerical algorithm that exploits this observation: it first computes all components in low precision, uses this output to estimate the condition numbers, and recomputes in higher precision only the components associated with large condition numbers. We test our algorithm on various networks and datasets and confirm experimentally that it can significantly improve the cost--accuracy tradeoff compared with uniform precision accumulation baselines.
Authors: Ruting Cheng, Yijiang Zheng, Boyuan Feng, Chuhui Qiu, Zhuoxin Long, Joaquin A. Calderon, Xiaoke Zhang, Jaclyn M. Phillips, James K. Hahn
Abstract: Monitoring maternal and fetal health during pregnancy is crucial for preventing adverse outcomes. While tests such as ultrasound scans offer high accuracy, they can be costly and inconvenient. Telehealth and more accessible body shape information provide pregnant women with a convenient way to monitor their health. This study explores the potential of 3D body scan data, captured during the 18-24 gestational weeks, to predict adverse pregnancy outcomes and estimate clinical parameters. We developed a novel algorithm with two parallel streams which are used for extract body shape features: one for supervised learning to extract sequential abdominal circumference information, and another for unsupervised learning to extract global shape descriptors, alongside a branch for demographic data. Our results indicate that 3D body shape can assist in predicting preterm labor, gestational diabetes mellitus (GDM), gestational hypertension (GH), and in estimating fetal weight. Compared to other machine learning models, our algorithm achieved the best performance, with prediction accuracies exceeding 88% and fetal weight estimation accuracy of 76.74% within a 10% error margin, outperforming conventional anthropometric methods by 22.22%.
Authors: Ashkan Shakarami, Yousef Yeganeh, Azade Farshad, Lorenzo Nicol\`e, Stefano Ghidoni, Nassir Navab
Abstract: Activation functions play a critical role in deep neural networks by shaping gradient flow, optimization stability, and generalization. While ReLU remains widely used due to its simplicity, it suffers from gradient sparsity and dead-neuron issues and offers no adaptivity to input statistics. Smooth alternatives such as Swish and GELU improve gradient propagation but still apply a fixed transformation regardless of the activation distribution. In this paper, we propose VeLU, a Variance-enhanced Learning Unit that introduces variance-aware and distributionally aligned nonlinearity through a principled combination of ArcTan-ArcSin transformations, adaptive scaling, and Wasserstein-2 regularization (Optimal Transport). This design enables VeLU to modulate its response based on local activation variance, mitigate internal covariate shift at the activation level, and improve training stability without adding learnable parameters or architectural overhead. Extensive experiments across six deep neural networks show that VeLU outperforms ReLU, ReLU6, Swish, and GELU on 12 vision benchmarks. The implementation of VeLU is publicly available in GitHub.
Authors: Kitsuya Azuma, Takayuki Nishio, Yuichi Kitagawa, Wakako Nakano, Takahito Tanimura
Abstract: Federated Learning (FL) enables collaborative model training across decentralized clients, enhancing privacy by keeping data local. Yet conventional FL, relying on frequent parameter-sharing, suffers from high communication overhead and limited model heterogeneity. Distillation-based FL approaches address these issues by sharing predictions (soft-labels, i.e., normalized probability distributions) instead, but they often involve redundant transmissions across communication rounds, reducing efficiency. We propose SCARLET, a novel framework integrating synchronized soft-label caching and an enhanced Entropy Reduction Aggregation (Enhanced ERA) mechanism. SCARLET minimizes redundant communication by reusing cached soft-labels, achieving up to 50% reduction in communication costs compared to existing methods while maintaining competitive accuracy. Enhanced ERA resolves the fundamental instability of conventional temperature-based aggregation, ensuring robust control and high performance in diverse client scenarios. Experimental evaluations demonstrate that SCARLET consistently outperforms state-of-the-art distillation-based FL methods in terms of accuracy and communication efficiency. The implementation of SCARLET is publicly available at https://github.com/kitsuyaazuma/SCARLET.
Authors: Marie-Charlotte Brandenburg, Katharina Jochemko
Abstract: We consider binary classification restricted to a class of continuous piecewise linear functions whose decision boundaries are (possibly nonconvex) starshaped polyhedral sets, supported on a fixed polyhedral simplicial fan. We investigate the expressivity of these function classes and describe the combinatorial and geometric structure of the loss landscape, most prominently the sublevel sets, for two loss-functions: the 0/1-loss (discrete loss) and a log-likelihood loss function. In particular, we give explicit bounds on the VC dimension of this model, and concretely describe the sublevel sets of the discrete loss as chambers in a hyperplane arrangement. For the log-likelihood loss, we give sufficient conditions for the optimum to be unique, and describe the geometry of the optimum when varying the rate parameter of the underlying exponential probability distribution.
Authors: Wenze Liu, Xiangyu Yue
Abstract: To accelerate diffusion model inference, numerical solvers perform poorly at extremely small steps, while distillation techniques often introduce complexity and instability. This work presents an intermediate strategy, balancing performance and cost, by learning ODE integration using loss functions derived from the derivative-integral relationship, inspired by Monte Carlo integration and Picard iteration. From a geometric perspective, the losses operate by gradually extending the tangent to the secant, thus are named as secant losses. The target of secant losses is the same as that of diffusion models, or the diffusion model itself, leading to great training stability. By fine-tuning or distillation, the secant version of EDM achieves a $10$-step FID of $2.14$ on CIFAR-10, while the secant version of SiT-XL/2 attains a $4$-step FID of $2.27$ and an $8$-step FID of $1.96$ on ImageNet-$256\times256$. Code is available at https://github.com/poppuppy/secant_expectation.
Authors: Tillmann Rheude, Roland Eils, Benjamin Wild
Abstract: Real-world machine learning applications often involve data from multiple modalities that must be integrated effectively to make robust predictions. However, in many practical settings, not all modalities are available for every sample, and acquiring additional modalities can be costly. This raises the question: which samples should be prioritized for additional modality acquisition when resources are limited? While prior work has explored individual-level acquisition strategies and training-time active learning paradigms, test-time and cohort-based acquisition remain underexplored. We introduce Cohort-based Active Modality Acquisition (CAMA), a novel test-time setting to formalize the challenge of selecting which samples should receive additional modalities. We derive acquisition strategies that leverage a combination of generative imputation and discriminative modeling to estimate the expected benefit of acquiring missing modalities based on common evaluation metrics. We also introduce upper-bound heuristics that provide performance ceilings to benchmark acquisition strategies. Experiments on multimodal datasets with up to 15 modalities demonstrate that our proposed imputation-based strategies can more effectively guide the acquisition of additional modalities for selected samples compared with methods relying solely on unimodal information, entropy-based guidance, or random selection. We showcase the real-world relevance and scalability of our method by demonstrating its ability to effectively guide the costly acquisition of proteomics data for disease prediction in a large prospective cohort, the UK Biobank (UKBB). Our work provides an effective approach for optimizing modality acquisition at the cohort level, enabling more effective use of resources in constrained settings.
Authors: Haonan Dong, Wenhao Zhu, Guojie Song, Liang Wang
Abstract: Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient fine-tuning (PEFT) method validated across NLP and CV domains. However, LoRA faces an inherent low-rank bottleneck: narrowing its performance gap with full finetuning requires increasing the rank of its parameter matrix, resulting in significant parameter overhead. Recent linear LoRA variants have attempted to enhance expressiveness by introducing additional linear mappings; however, their composition remains inherently linear and fails to fundamentally improve LoRA's representational capacity. To address this limitation, we propose AuroRA, which incorporates an Adaptive Nonlinear Layer (ANL) between two linear projectors to capture fixed and learnable nonlinearities. This combination forms an MLP-like structure with a compressed rank, enabling flexible and precise approximation of diverse target functions while theoretically guaranteeing lower approximation errors and bounded gradients. Extensive experiments on 22 datasets and 6 pretrained models demonstrate that AuroRA: (I) not only matches or surpasses full fine-tuning performance with only 6.18% ~ 25% of LoRA's parameters but also (II) outperforms competitive PEFT methods by up to 10.88% in both NLP and CV tasks, and (III) exhibits robust performance across various rank configurations.
Authors: Sourav Ganguly, Arnob Ghosh, Kishan Panaganti, Adam Wierman
Abstract: Constrained decision-making is essential for designing safe policies in real-world control systems, yet simulated environments often fail to capture real-world adversities. We consider the problem of learning a policy that will maximize the cumulative reward while satisfying a constraint, even when there is a mismatch between the real model and an accessible simulator/nominal model. In particular, we consider the robust constrained Markov decision problem (RCMDP) where an agent needs to maximize the reward and satisfy the constraint against the worst possible stochastic model under the uncertainty set centered around an unknown nominal model. Primal-dual methods, effective for standard constrained MDP (CMDP), are not applicable here because of the lack of the strong duality property. Further, one cannot apply the standard robust value-iteration based approach on the composite value function either as the worst case models may be different for the reward value function and the constraint value function. We propose a novel technique that effectively minimizes the constraint value function--to satisfy the constraints; on the other hand, when all the constraints are satisfied, it can simply maximize the robust reward value function. We prove that such an algorithm finds a policy with at most $\epsilon$ sub-optimality and feasible policy after $O(\epsilon^{-2})$ iterations. In contrast to the state-of-the-art method, we do not need to employ a binary search, thus, we reduce the computation time by at least 4x for smaller value of discount factor ($\gamma$) and by at least 6x for larger value of $\gamma$.
Authors: Qianyi Xu, Gousia Habib, Feng Wu, Dilruk Perera, Mengling Feng
Abstract: Timely and personalized treatment decisions are essential across a wide range of healthcare settings where patient responses can vary significantly and evolve over time. Clinical data used to support these treatment decisions are often irregularly sampled, where missing data frequencies may implicitly convey information about the patient's condition. Existing Reinforcement Learning (RL) based clinical decision support systems often ignore the missing patterns and distort them with coarse discretization and simple imputation. They are also predominantly model-free and largely depend on retrospective data, which could lead to insufficient exploration and bias by historical behaviors. To address these limitations, we propose medDreamer, a novel model-based reinforcement learning framework for personalized treatment recommendation. medDreamer contains a world model with an Adaptive Feature Integration module that simulates latent patient states from irregular data and a two-phase policy trained on a hybrid of real and imagined trajectories. This enables learning optimal policies that go beyond the sub-optimality of historical clinical decisions, while remaining close to real clinical data. We evaluate medDreamer on both sepsis and mechanical ventilation treatment tasks using two large-scale Electronic Health Records (EHRs) datasets. Comprehensive evaluations show that medDreamer significantly outperforms model-free and model-based baselines in both clinical outcomes and off-policy metrics.
Authors: Antti Koskela, Tejas Kulkarni
Abstract: Fully decentralized training of machine learning models offers significant advantages in scalability, robustness, and fault tolerance. However, achieving differential privacy (DP) guarantees in such settings is challenging due to the absence of a central aggregator and varying trust assumptions among nodes. We present a novel privacy analysis of decentralized gossip-based averaging algorithms with additive node-level noise, from arbitrary views of nodes in a graph and especially consider the averaging over nearest neighbors with secure summation and individual node-wise views. Our main contribution is a an analytical framework based on a linear systems formulation that accurately characterizes privacy leakage between nodes across different scenarios. In case the gossip averaging happens via secure summation, we show that the R\'enyi DP parameter growth is asymptotically $O(T)$, where $T$ is the number of training rounds, similarly as in the case of central aggregation.
Authors: Chetan Verma, Aditya Srinivas Timmaraju, Cho-Jui Hsieh, Suyash Damle, Ngot Bui, Yang Zhang, Wen Chen, Xin Liu, Prateek Jain, Inderjit S Dhillon
Abstract: Industry-grade ML models are carefully designed to meet rapidly evolving serving constraints, which requires significant resources for model development. In this paper, we propose MatTA, a framework for training multiple accurate Student models using a novel Teacher-TA-Student recipe. TA models are larger versions of the Student models with higher capacity, and thus allow Student models to better relate to the Teacher model and also bring in more domain-specific expertise. Furthermore, multiple accurate Student models can be extracted from the TA model. Therefore, despite only one training run, our methodology provides multiple servable options to trade off accuracy for lower serving cost. We demonstrate the proposed method, MatTA, on proprietary datasets and models. Its practical efficacy is underscored by live A/B tests within a production ML system, demonstrating 20% improvement on a key metric. We also demonstrate our method on GPT-2 Medium, a public model, and achieve relative improvements of over 24% on SAT Math and over 10% on the LAMBADA benchmark.
Authors: Quentin Bertrand, Anne Gagneux, Mathurin Massias, R\'emi Emonet
Abstract: Modern deep generative models can now produce high-quality synthetic samples that are often indistinguishable from real training data. A growing body of research aims to understand why recent methods, such as diffusion and flow matching techniques, generalize so effectively. Among the proposed explanations are the inductive biases of deep learning architectures and the stochastic nature of the conditional flow matching loss. In this work, we rule out the noisy nature of the loss as a key factor driving generalization in flow matching. First, we empirically show that in high-dimensional settings, the stochastic and closed-form versions of the flow matching loss yield nearly equivalent losses. Then, using state-of-the-art flow matching models on standard image datasets, we demonstrate that both variants achieve comparable statistical performance, with the surprising observation that using the closed-form can even improve performance.
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: Mohammad Emtiyaz Khan
Abstract: Adaptation is the holy grail of intelligence, but even the best AI models lack the adaptability of toddlers. In spite of great progress, little is known about the mechanisms by which machines can learn to adapt as fast as humans and animals. Here, we cast adaptation as `correction' of old posteriors and show that a wide-variety of existing adaptation methods follow this very principle, including those used for continual learning, federated learning, unlearning, and model merging. In all these settings, more accurate posteriors often lead to smaller corrections and can enable faster adaptation. Posterior correction is derived by using the dual representation of the Bayesian Learning Rule of Khan and Rue (2023), where the interference between the old representation and new information is quantified by using the natural-gradient mismatch. We present many examples demonstrating how machines can learn to adapt quickly by using posterior correction.
Authors: Siddharth M. Narayanan, James D. Braza, Ryan-Rhys Griffiths, Albert Bou, Geemi Wellawatte, Mayk Caldas Ramos, Ludovico Mitchener, Samuel G. Rodriques, Andrew D. White
Abstract: Reasoning models are large language models that emit a long chain-of-thought before answering, providing both higher accuracy and explicit reasoning for their response. A major question has been whether language model reasoning generalizes beyond mathematics, programming, and logic, where most previous work has focused. We demonstrate that reasoning models can be post-trained for chemistry without additional domain pretraining, and require substantially less data compared to contemporary domain-specific models. We report ether0, a 24B parameter LLM (based on Mistral-Small-24B) that can reason in natural language and respond with chemical structures. This reasoning model was trained with reinforcement learning on 640,730 experimentally-grounded chemistry problems across 375 tasks ranging from synthesizability, to blood-brain barrier permeability, to human receptor activity, to scent. Our model exceeds general-purpose chemistry models, frontier models, and human experts on molecular design tasks. It is also more data efficient relative to specialized models. We anticipate that this method can be applied to train data-efficient language models specialized for tasks across a wide variety of scientific domains.
Authors: Hegang Chen, Yuyin Lu, Yifan Zhao, Zhiming Dai, Fu Lee Wang, Qing Li, Yanghui Rao, Yue Li
Abstract: Single-cell RNA sequencing technologies have revolutionized our understanding of cellular heterogeneity, yet computational methods often struggle to balance performance with biological interpretability. Embedded topic models have been widely used for interpretable single-cell embedding learning. However, these models suffer from the potential problem of interpretation collapse, where topics semantically collapse towards each other, resulting in redundant topics and incomplete capture of biological variation. Furthermore, the rise of single-cell foundation models creates opportunities to harness external biological knowledge for guiding model embeddings. Here, we present scE2TM, an external knowledge-guided embedded topic model that provides a high-quality cell embedding and interpretation for scRNA-seq analysis. Through embedding clustering regularization method, each topic is constrained to be the center of a separately aggregated gene cluster, enabling it to capture unique biological information. Across 20 scRNA-seq datasets, scE2TM achieves superior clustering performance compared with seven state-of-the-art methods. A comprehensive interpretability benchmark further shows that scE2TM-learned topics exhibit higher diversity and stronger consistency with underlying biological pathways. Modeling interferon-stimulated PBMCs, scE2TM simulates topic perturbations that drive control cells toward stimulated-like transcriptional states, faithfully mirroring experimental interferon responses. In melanoma, scE2TM identifies malignant-specific topics and extrapolates them to unseen patient data, revealing gene programs associated with patient survival.
Authors: Jiaxing Qiu, Dongliang Guo, Brynne Sullivan, Teague R. Henry, Tom Hartvigsen
Abstract: In time series editing, we aim to modify some properties of a given time series without altering others. For example, when analyzing a hospital patient's blood pressure, we may add a sudden early drop and observe how it impacts their future while preserving other conditions. Existing diffusion-based editors rely on rigid, predefined attribute vectors as conditions and produce all-or-nothing edits through sampling. This attribute- and sampling-based approach limits flexibility in condition format and lacks customizable control over editing strength. To overcome these limitations, we introduce Instruction-based Time Series Editing, where users specify intended edits using natural language. This allows users to express a wider range of edits in a more accessible format. We then introduce InstructTime, the first instruction-based time series editor. InstructTime takes in time series and instructions, embeds them into a shared multi-modal representation space, then decodes their embeddings to generate edited time series. By learning a structured multi-modal representation space, we can easily interpolate between embeddings to achieve varying degrees of edit. To handle local and global edits together, we propose multi-resolution encoders. In our experiments, we use synthetic and real datasets and find that InstructTime is a state-of-the-art time series editor: InstructTime achieves high-quality edits with controllable strength, can generalize to unseen instructions, and can be easily adapted to unseen conditions through few-shot learning.
Authors: Javier Fumanal-Idocin, Raquel Fernandez-Peralta, Javier Andreu-Perez
Abstract: Dynamic feature selection (DFS) offers a compelling alternative to traditional, static feature selection by adapting the selected features to each individual sample. This provides insights into the decision-making process for each case, which makes DFS especially significant in settings where decision transparency is key, i.e., clinical decisions. However, existing DFS methods use opaque models, which hinder their applicability in real-life scenarios. DFS also introduces new own sources of uncertainty compared to the static setting, which is also not considered in the existing literature. In this paper, we formalize the additional sources of uncertainty in DFS, and give formulas to estimate them. We also propose novel approach by leveraging a rule-based system as a base classifier for the DFS process, which enhances decision interpretability compared to neural estimators. Finally, we demonstrate the competitive performance of our rule-based DFS approach against established and state-of-the-art greedy and reinforcement learning methods, which are mostly considered opaque, compared to our explainable rulebased system.
Authors: Atakan Azakli, Bernd Stelzer
Abstract: In this work, we introduce a new information-theoretic perspective on Multiple Instance Learning (MIL) for parameter estimation with i.i.d. data, and show that MIL can outperform single-instance learners in low-signal regimes. Prior work [Nachman and Thaler, 2021] argued that single-instance methods are often sufficient, but this conclusion presumes enough single-instance signal to train near-optimal classifiers. We demonstrate that even state-of-the-art single-instance models can fail to reach optimal classifier performance in challenging low-signal regimes, whereas MIL can mitigate this sub-optimality. As a concrete application, we constrain Wilson coefficients of the Standard Model Effective Field Theory (SMEFT) using kinematic information from subatomic particle collision events at the Large Hadron Collider (LHC). In experiments, we observe that under specific modeling and weak signal conditions, pooling instances can increase the effective Fisher information compared to single-instance approaches.
Authors: Felipe Maia Polo, Xinhe Wang, Mikhail Yurochkin, Gongjun Xu, Moulinath Banerjee, Yuekai Sun
Abstract: Large language models are increasingly used as judges (LLM-as-a-judge) to evaluate model outputs at scale, but their assessments often diverge systematically from human judgments. We present Bridge, a unified statistical framework that explicitly bridges human and LLM evaluations under both absolute scoring and pairwise comparison paradigms. Bridge posits a latent human preference score for each prompt-response pair and models LLM deviations as linear transformations of covariates that capture sources of discrepancies. This offers a simple and principled framework for refining LLM ratings and characterizing systematic discrepancies between humans and LLMs. We provide an efficient fitting algorithm with asymptotic guarantees for statistical inference. Using six LLM judges and two benchmarks (BigGen Bench and Chatbot Arena), Bridge achieves higher agreement with human ratings (accuracy, calibration, and KL divergence) and exposes systematic human-LLM gaps.
Authors: Soumyajit Guin, Shalabh Bhatnagar
Abstract: In this paper we propose two algorithms in the tabular setting and an algorithm for the function approximation setting for the Stochastic Shortest Path (SSP) problem. SSP problems form an important class of problems in Reinforcement Learning (RL), as other types of cost-criteria in RL can be formulated in the setting of SSP. We show asymptotic almost-sure convergence for all our algorithms. We observe superior performance of our tabular algorithms compared to other well-known convergent RL algorithms. We further observe reliable performance of our function approximation algorithm compared to other algorithms in the function approximation setting.
Authors: Songyu Ke, Chenyu Wu, Yuxuan Liang, Huiling Qin, Junbo Zhang, Yu Zheng
Abstract: The ubiquity of missing data in urban intelligence systems, attributable to adverse environmental conditions and equipment failures, poses a significant challenge to the efficacy of downstream applications, notably in the realms of traffic forecasting and energy consumption prediction. Therefore, it is imperative to develop a robust spatio-temporal learning methodology capable of extracting meaningful insights from incomplete datasets. Despite the existence of methodologies for spatio-temporal graph forecasting in the presence of missing values, unresolved issues persist. Primarily, the majority of extant research is predicated on time-series analysis, thereby neglecting the dynamic spatial correlations inherent in sensor networks. Additionally, the complexity of missing data patterns compounds the intricacy of the problem. Furthermore, the variability in maintenance conditions results in a significant fluctuation in the ratio and pattern of missing values, thereby challenging the generalizability of predictive models. In response to these challenges, this study introduces GeoMAE, a self-supervised spatio-temporal representation learning model. The model is comprised of three principal components: an input preprocessing module, an attention-based spatio-temporal forecasting network (STAFN), and an auxiliary learning task, which draws inspiration from Masking AutoEncoders to enhance the robustness of spatio-temporal representation learning. Empirical evaluations on real-world datasets demonstrate that GeoMAE significantly outperforms existing benchmarks, achieving up to 13.20\% relative improvement over the best baseline models.
Authors: Akash Choudhuri, Yongjian Zhong, Bijaya Adhikari
Abstract: Hypergraphs offer a generalized framework for capturing high-order relationships between entities and have been widely applied in various domains, including healthcare, social networks, and bioinformatics. Hypergraph neural networks, which rely on message-passing between nodes over hyperedges to learn latent representations, have emerged as the method of choice for predictive tasks in many of these domains. These approaches typically perform only a small number of message-passing rounds to learn the representations, which they then utilize for predictions. The small number of message-passing rounds comes at a cost, as the representations only capture local information and forego long-range high-order dependencies. However, as we demonstrate, blindly increasing the message-passing rounds to capture long-range dependency also degrades the performance of hyper-graph neural networks. Recent works have demonstrated that implicit graph neural networks capture long-range dependencies in standard graphs while maintaining performance. Despite their popularity, prior work has not studied long-range dependency issues on hypergraph neural networks. Here, we first demonstrate that existing hypergraph neural networks lose predictive power when aggregating more information to capture long-range dependency. We then propose Implicit Hypergraph Neural Network (IHNN), a novel framework that jointly learns fixed-point representations for both nodes and hyperedges in an end-to-end manner to alleviate this issue. Leveraging implicit differentiation, we introduce a tractable projected gradient descent approach to train the model efficiently. Extensive experiments on real-world hypergraphs for node classification demonstrate that IHNN outperforms the closest prior works in most settings, establishing a new state-of-the-art in hypergraph learning.
Authors: Charlie B. Tan, Majdi Hassan, Leon Klein, Saifuddin Syed, Dominique Beaini, Michael M. Bronstein, Alexander Tong, Kirill Neklyudov
Abstract: Efficient equilibrium sampling of molecular conformations remains a core challenge in computational chemistry and statistical inference. Classical approaches such as molecular dynamics or Markov chain Monte Carlo inherently lack amortization; the computational cost of sampling must be paid in full for each system of interest. The widespread success of generative models has inspired interest towards overcoming this limitation through learning sampling algorithms. Despite performing competitively with conventional methods when trained on a single system, learned samplers have so far demonstrated limited ability to transfer across systems. We demonstrate that deep learning enables the design of scalable and transferable samplers by introducing Prose, a 285 million parameter all-atom transferable normalizing flow trained on a corpus of peptide molecular dynamics trajectories up to 8 residues in length. Prose draws zero-shot uncorrelated proposal samples for arbitrary peptide systems, achieving the previously intractable transferability across sequence length, whilst retaining the efficient likelihood evaluation of normalizing flows. Through extensive empirical evaluation we demonstrate the efficacy of Prose as a proposal for a variety of sampling algorithms, finding a simple importance sampling-based finetuning procedure to achieve competitive performance to established methods such as sequential Monte Carlo. We open-source the Prose codebase, model weights, and training dataset, to further stimulate research into amortized sampling methods and finetuning objectives.
Authors: Pieter Smet
Abstract: Uncertainty in optimization is often represented as stochastic parameters in the optimization model. In Predict-Then-Optimize approaches, predictions of a machine learning model are used as values for such parameters, effectively transforming the stochastic optimization problem into a deterministic one. This two-stage framework is built on the assumption that more accurate predictions result in solutions that are closer to the actual optimal solution. However, providing evidence for this assumption in the context of complex, constrained optimization problems is challenging and often overlooked in the literature. Simulating predictions of machine learning models offers a way to (experimentally) analyze how prediction error impacts solution quality without the need to train real models. Complementing an algorithm from the literature for simulating binary classification, we introduce a new algorithm for simulating predictions of multiclass classifiers. We conduct a computational study to evaluate the performance of these algorithms, and show that classifier performance can be simulated with reasonable accuracy, although some variability is observed. Additionally, we apply these algorithms to assess the performance of a Predict-Then-Optimize algorithm for a machine scheduling problem. The experiments demonstrate that the relationship between prediction error and how close solutions are to the actual optimum is non-trivial, highlighting important considerations for the design and evaluation of decision-making systems based on machine learning predictions.
Authors: Wenhui Cui, Christopher Sandino, Hadi Pouransari, Ran Liu, Juri Minxha, Ellen Zippi, Aman Verma, Anna Sedlackova, Erdrin Azemi, Behrooz Mahasseni
Abstract: Hand gesture classification using high-quality structured data such as videos, images, and hand skeletons is a well-explored problem in computer vision. Leveraging low-power, cost-effective biosignals, e.g. surface electromyography (sEMG), allows for continuous gesture prediction on wearables. In this paper, we demonstrate that learning representations from weak-modality data that are aligned with those from structured, high-quality data can improve representation quality and enables zero-shot classification. Specifically, we propose a Contrastive Pose-EMG Pre-training (CPEP) framework to align EMG and pose representations, where we learn an EMG encoder that produces high-quality and pose-informative representations. We assess the gesture classification performance of our model through linear probing and zero-shot setups. Our model outperforms emg2pose benchmark models by up to 21% on in-distribution gesture classification and 72% on unseen (out-of-distribution) gesture classification.
Authors: Michael H. Freedman, Michael Mulligan
Abstract: The Kolmogorov-Arnold (KA) representation theorem constructs universal, but highly non-smooth inner functions (the first layer map) in a single (non-linear) hidden layer neural network. Such universal functions have a distinctive local geometry, a "texture," which can be characterized by the inner function's Jacobian $J({\mathbf{x}})$, as $\mathbf{x}$ varies over the data. It is natural to ask if this distinctive KA geometry emerges through conventional neural network optimization. We find that indeed KA geometry often is produced when training vanilla single hidden layer neural networks. We quantify KA geometry through the statistical properties of the exterior powers of $J(\mathbf{x})$: number of zero rows and various observables for the minor statistics of $J(\mathbf{x})$, which measure the scale and axis alignment of $J(\mathbf{x})$. This leads to a rough understanding for where KA geometry occurs in the space of function complexity and model hyperparameters. The motivation is first to understand how neural networks organically learn to prepare input data for later downstream processing and, second, to learn enough about the emergence of KA geometry to accelerate learning through a timely intervention in network hyperparameters. This research is the "flip side" of KA-Networks (KANs). We do not engineer KA into the neural network, but rather watch KA emerge in shallow MLPs.
Authors: Yuyang Wang, Jiarui Lu, Navdeep Jaitly, Josh Susskind, Miguel Angel Bautista
Abstract: Protein folding models have achieved groundbreaking results typically via a combination of integrating domain knowledge into the architectural blocks and training pipelines. Nonetheless, given the success of generative models across different but related problems, it is natural to question whether these architectural designs are a necessary condition to build performant models. In this paper, we introduce SimpleFold, the first flow-matching based protein folding model that solely uses general purpose transformer blocks. Protein folding models typically employ computationally expensive modules involving triangular updates, explicit pair representations or multiple training objectives curated for this specific domain. Instead, SimpleFold employs standard transformer blocks with adaptive layers and is trained via a generative flow-matching objective with an additional structural term. We scale SimpleFold to 3B parameters and train it on approximately 9M distilled protein structures together with experimental PDB data. On standard folding benchmarks, SimpleFold-3B achieves competitive performance compared to state-of-the-art baselines, in addition SimpleFold demonstrates strong performance in ensemble prediction which is typically difficult for models trained via deterministic reconstruction objectives. Due to its general-purpose architecture, SimpleFold shows efficiency in deployment and inference on consumer-level hardware. SimpleFold challenges the reliance on complex domain-specific architectures designs in protein folding, opening up an alternative design space for future progress.
Authors: Kevin Bradley Dsouza, Enoch Ofosu, Daniel Chukwuemeka Amaogu, J\'er\^ome Pigeon, Richard Boudreault, Pooneh Maghoul, Juan Moreno-Cruz, Yuri Leonenko
Abstract: Boreal forests store 30-40\% of terrestrial carbon, much in climate-vulnerable permafrost soils, making their management critical for climate mitigation. However, optimizing forest management for both carbon sequestration and permafrost preservation presents complex trade-offs that current tools cannot adequately address. We introduce BoreaRL, the first multi-objective reinforcement learning environment for climate-adaptive boreal forest management, featuring a physically-grounded simulator of coupled energy, carbon, and water fluxes. BoreaRL supports two training paradigms: site-specific mode for controlled studies and generalist mode for learning robust policies under environmental stochasticity. Through evaluation of multi-objective RL algorithms, we reveal a fundamental asymmetry in learning difficulty: carbon objectives are significantly easier to optimize than thaw (permafrost preservation) objectives, with thaw-focused policies showing minimal learning progress across both paradigms. In generalist settings, standard gradient-descent based preference-conditioned approaches fail, while a naive site selection approach achieves superior performance by strategically selecting training episodes. Analysis of learned strategies reveals distinct management philosophies, where carbon-focused policies favor aggressive high-density coniferous stands, while effective multi-objective policies balance species composition and density to protect permafrost while maintaining carbon gains. Our results demonstrate that robust climate-adaptive forest management remains challenging for current MORL methods, establishing BoreaRL as a valuable benchmark for developing more effective approaches. We open-source BoreaRL to accelerate research in multi-objective RL for climate applications.
Authors: Vivek Myers, Bill Chunyuan Zheng, Benjamin Eysenbach, Sergey Levine
Abstract: Approaches for goal-conditioned reinforcement learning (GCRL) often use learned state representations to extract goal-reaching policies. Two frameworks for representation structure have yielded particularly effective GCRL algorithms: (1) *contrastive representations*, in which methods learn "successor features" with a contrastive objective that performs inference over future outcomes, and (2) *temporal distances*, which link the (quasimetric) distance in representation space to the transit time from states to goals. We propose an approach that unifies these two frameworks, using the structure of a quasimetric representation space (triangle inequality) with the right additional constraints to learn successor representations that enable optimal goal-reaching. Unlike past work, our approach is able to exploit a **quasimetric** distance parameterization to learn **optimal** goal-reaching distances, even with **suboptimal** data and in **stochastic** environments. This gives us the best of both worlds: we retain the stability and long-horizon capabilities of Monte Carlo contrastive RL methods, while getting the free stitching capabilities of quasimetric network parameterizations. On existing offline GCRL benchmarks, our representation learning objective improves performance on stitching tasks where methods based on contrastive learning struggle, and on noisy, high-dimensional environments where methods based on quasimetric networks struggle.
Authors: Yuandong Tian
Abstract: While the phenomenon of grokking, i.e., delayed generalization, has been studied extensively, it remains an open problem whether there is a mathematical framework that characterizes what kind of features will emerge, how and in which conditions it happens, and is closely related to the gradient dynamics of the training, for complex structured inputs. We propose a novel framework, named $\mathbf{Li}_2$, that captures three key stages for the grokking behavior of 2-layer nonlinear networks: (I) Lazy learning, (II) independent feature learning and (III) interactive feature learning. At the lazy learning stage, top layer overfits to random hidden representation and the model appears to memorize, and at the same time, the backpropagated gradient $G_F$ from the top layer now carries information about the target label, with a specific structure that enables each hidden node to learn their representation independently. Interestingly, the independent dynamics follows exactly the gradient ascent of an energy function $E$, and its local maxima are precisely the emerging features. We study whether these local-optima induced features are generalizable, their representation power, and how they change on sample size, in group arithmetic tasks. When hidden nodes start to interact in the later stage of learning, we provably show how $G_F$ changes to focus on missing features that need to be learned. Our study sheds lights on roles played by key hyperparameters such as weight decay, learning rate and sample sizes in grokking, leads to provable scaling laws of feature emergence, memorization and generalization, and reveals why recent optimizers such as Muon can be effective, from the first principles of gradient dynamics. Our analysis can be extended to multi-layers. The code is available at https://github.com/yuandong-tian/understanding/tree/main/ssl/real-dataset/cogo.
URLs: https://github.com/yuandong-tian/understanding/tree/main/ssl/real-dataset/cogo.
Authors: Daniele Foffano, Alessio Russo, Alexandre Proutiere
Abstract: Robustness to modeling errors and uncertainties remains a central challenge in reinforcement learning (RL). In this work, we address this challenge by leveraging diffusion models to train robust RL policies. Diffusion models have recently gained popularity in model-based RL due to their ability to generate full trajectories "all at once", mitigating the compounding errors typical of step-by-step transition models. Moreover, they can be conditioned to sample from specific distributions, making them highly flexible. We leverage conditional sampling to learn policies that are robust to uncertainty in environment dynamics. Building on the established connection between Conditional Value at Risk (CVaR) optimization and robust RL, we introduce Adversarial Diffusion for Robust Reinforcement Learning (AD-RRL). AD-RRL guides the diffusion process to generate worst-case trajectories during training, effectively optimizing the CVaR of the cumulative return. Empirical results across standard benchmarks show that AD-RRL achieves superior robustness and performance compared to existing robust RL methods.
Authors: Jincheng Liu, Sijun He, Jingjing Wu, Xiangsen Wang, Yang Chen, Zhaoqi Kuang, Siqi Bao, Yuan Yao
Abstract: Recent large language models (LLMs) have shown strong reasoning capabilities. However, a critical question remains: do these models possess genuine reasoning skills particularly complex strategic reasoning or are they primarily excelling at sophisticated pattern recognition within their training data? To address this question, this paper presents a chess testbed, ChessArena, to evaluate the strategic reasoning capabilities of LLMs. Chess requires complex strategic reasoning capabilities including long-term planning, strict rule comprehension, and multi-turn conversation memorization. Specifically, ChessArena is a competitive framework where LLMs play against each other, under four different play modes. The testbed is equipped with a ranking algorithm and a leaderboard. The testbed can also evaluate fine-grained capabilities including basic understanding, move selection, and puzzle solving. Over 13 LLMs with different modes are evaluated in ChessArena, playing over 800 games. The results reveal significant shortcomings in current LLMs: no model can beat Maia-1100 (a chess engine at human amateur level), while some even failed to defeat a random player that selects moves arbitrarily. We also present a strong baseline to the testbed: our fine-tuned Qwen3-8B substantially improved performance, approaching much larger state-of-the-art reasoning models.
Authors: Peter Holderrieth, Uriel Singer, Tommi Jaakkola, Ricky T. Q. Chen, Yaron Lipman, Brian Karrer
Abstract: The performance of flow matching and diffusion models can be greatly improved at inference time using reward alignment algorithms, yet efficiency remains a major limitation. While several algorithms were proposed, we demonstrate that a common bottleneck is the sampling method these algorithms rely on: many algorithms require to sample Markov transitions via SDE sampling, which is significantly less efficient and often less performant than ODE sampling. To remove this bottleneck, we introduce GLASS Flows, a new sampling paradigm that simulates a "flow matching model within a flow matching model" to sample Markov transitions. As we show in this work, this "inner" flow matching model can be retrieved from a pre-trained model without any re-training, combining the efficiency of ODEs with the stochastic evolution of SDEs. On large-scale text-to-image models, we show that GLASS Flows eliminate the trade-off between stochastic evolution and efficiency. Combined with Feynman-Kac Steering, GLASS Flows improve state-of-the-art performance in text-to-image generation, making it a simple, drop-in solution for inference-time scaling of flow and diffusion models.
Authors: Hengyi Zhu, Grace Li Zhang, Shaoyi Huang
Abstract: Hardware-Aware Neural Architecture Search (HW-NAS) requires joint optimization of accuracy and latency under device constraints. Traditional supernet-based methods require multiple GPU days per dataset. Large Language Model (LLM)-driven approaches avoid training a large supernet and can provide quick feedback, but we observe an exploration bias: the LLM repeatedly proposes neural network designs within limited search space and fails to discover architectures across different latency ranges in the entire search space. To address this issue, we propose LLM-NAS: an LLM-driven Neural Architecture Search that can generate neural networks with high accuracy and low latency with reduced search cost. Our proposed LLM-NAS has three key components: 1) a complexity-driven partitioning engine that divides the search space by complexity to enforce diversity and mitigate exploration bias; 2) an LLM-powered architecture prompt co-evolution operator, in which the LLM first updates a knowledge base of design heuristics based on results from the previous round, then performs a guided evolution algorithm on architectures with prompts that incorporate this knowledge base. Prompts and designs improve together across rounds which avoids random guesswork and improve efficiency; 3) a zero-cost predictor to avoid training a large number of candidates from scratch. Experimental results show that on HW-NAS-Bench, LLM-NAS can achieve overall higher HV, lower IGD, and up to 54% lower latency than baselines at similar accuracy. Meanwhile, the search cost drops from days to minutes compared with traditional supernet baselines.
Authors: Seungeun Rho, Aaron Trinh, Danfei Xu, Sehoon Ha
Abstract: Scaling unsupervised skill discovery algorithms to high-DoF agents remains challenging. As dimensionality increases, the exploration space grows exponentially, while the manifold of meaningful skills remains limited. Therefore, semantic meaningfulness becomes essential to effectively guide exploration in high-dimensional spaces. In this work, we present **Reference-Grounded Skill Discovery (RGSD)**, a novel algorithm that grounds skill discovery in a semantically meaningful latent space using reference data. RGSD first performs contrastive pretraining to embed motions on a unit hypersphere, clustering each reference trajectory into a distinct direction. This grounding enables skill discovery to simultaneously involve both imitation of reference behaviors and the discovery of semantically related diverse behaviors. On a simulated SMPL humanoid with $359$-D observations and $69$-D actions, RGSD successfully imitates skills such as walking, running, punching, and sidestepping, while also discover variations of these behaviors. In downstream locomotion tasks, RGSD leverages the discovered skills to faithfully satisfy user-specified style commands and outperforms imitation-learning baselines, which often fail to maintain the commanded style. Overall, our results suggest that lightweight reference-grounding offers a practical path to discovering semantically rich and structured skills in high-DoF systems.
Authors: Tuan Nguyen, Long Tran-Thanh
Abstract: Ensuring that large language models (LLMs) comply with safety requirements is a central challenge in AI deployment. Existing alignment approaches primarily operate during training, such as through fine-tuning or reinforcement learning from human feedback, but these methods are costly and inflexible, requiring retraining whenever new requirements arise. Recent efforts toward inference-time alignment mitigate some of these limitations but still assume access to model internals, which is impractical, and not suitable for third party stakeholders who do not have access to the models. In this work, we propose a model-independent, black-box framework for safety alignment that does not require retraining or access to the underlying LLM architecture. As a proof of concept, we address the problem of trading off between generating safe but uninformative answers versus helpful yet potentially risky ones. We formulate this dilemma as a two-player zero-sum game whose minimax equilibrium captures the optimal balance between safety and helpfulness. LLM agents operationalize this framework by leveraging a linear programming solver at inference time to compute equilibrium strategies. Our results demonstrate the feasibility of black-box safety alignment, offering a scalable and accessible pathway for stakeholders, including smaller organizations and entities in resource-constrained settings, to enforce safety across rapidly evolving LLM ecosystems.
Authors: Shiyu Chen, Ningyuan Huang, Soledad Villar
Abstract: Graph Neural Networks (GNNs) typically scale with the number of graph edges, making them well suited for sparse graphs but less efficient on dense graphs, such as point clouds or molecular interactions. A common remedy is to sparsify the graph via similarity thresholding or distance pruning, but this forces an arbitrary choice of a single interaction scale and discards crucial information from other scales. To overcome this limitation, we introduce a multi-view graph-tuple framework. Instead of a single graph, our graph-tuple framework partitions the graph into disjoint subgraphs, capturing primary local interactions and weaker, long-range connections. We then learn multi-view representations from the graph-tuple via a heterogeneous message-passing architecture inspired by the theory of non-commuting operators, which we formally prove is strictly more expressive and guarantees a lower oracle risk compared to single-graph message-passing models. We instantiate our framework on two scientific domains: molecular property prediction from feature-scarce Coulomb matrices and cosmological parameter inference from geometric point clouds. On both applications, our multi-view graph-tuple models demonstrate better performance than single-graph baselines, highlighting the power and versatility of our multi-view approach.
Authors: Chenwei Tang, Jingyu Xing, Lin Long, Xinyu Liu, Deng Xiong, Wei Ju, Shudong Huang, Jiancheng Lv, Ziyue Qiao
Abstract: Reinforcement Learning (RL) has emerged as a powerful paradigm for advancing Large Language Models (LLMs), achieving remarkable performance in complex reasoning domains such as mathematics and code generation. However, current RL methods face a fundamental scalability bottleneck due to their heavy reliance on human-curated preference data or labeled datasets for reward modeling. To overcome this limitation, we explore RL on unlabeled data where models learn autonomously from continuous experience streams. The core challenge in this setting lies in reliable reward estimation without ground-truth supervision. Existing approaches like Test-Time RL address this through self-consistent consensus, but risk reinforcing incorrect pseudo-labels derived from majority voting. We introduce COMPASS (Composite Path and Answer Self-Scoring), a novel test-time reward mechanism that operates without external supervision. COMPASS integrates two complementary components: the Dual-Calibration Answer Reward (DCAR), which stabilizes training by establishing trustworthy pseudo-labels through confidence and credibility calibration, and the Decisive Path Reward (DPR), which directly optimizes the reasoning process quality beyond mere outcome supervision. By jointly reinforcing trustworthy consensus answers and highly decisive reasoning chains, the COMPASS systematically enhances the model's analytical capabilities. Extensive experiments show that COMPASS achieves significant and consistent performance gains across diverse reasoning tasks and model architectures, advancing a more scalable direction for LLMs to learn from continuous experience.
Authors: Jane H. Lee, Baturay Saglam, Spyridon Pougkakiotis, Amin Karbasi, Dionysis Kalogerias
Abstract: Constrained optimization provides a common framework for dealing with conflicting objectives in reinforcement learning (RL). In most of these settings, the objectives (and constraints) are expressed though the expected accumulated reward. However, this formulation neglects risky or even possibly catastrophic events at the tails of the reward distribution, and is often insufficient for high-stakes applications in which the risk involved in outliers is critical. In this work, we propose a framework for risk-aware constrained RL, which exhibits per-stage robustness properties jointly in reward values and time using optimized certainty equivalents (OCEs). Our framework ensures an exact equivalent to the original constrained problem within a parameterized strong Lagrangian duality framework under appropriate constraint qualifications, and yields a simple algorithmic recipe which can be wrapped around standard RL solvers, such as PPO. Lastly, we establish the convergence of the proposed algorithm under common assumptions, and verify the risk-aware properties of our approach through several numerical experiments.
Authors: Jianqing Zhang, Zhezheng Hao, Wei Xia, Hande Dong, Hong Wang, Chenxing Wei, Yuyan Zhou, Yubin Qi, Qiang Lin, Jian Cao
Abstract: Reinforcement learning (RL) is widely used for post-training large language models (LLMs) in code editing, where group-relative methods like GRPO are popular for their critic-free, normalized advantage estimation. However, in real-world code-editing scenarios, reward distributions are often skewed with unpredictable outliers, leading to distorted advantage computation and increased noise. To address this issue, we propose Group Adaptive Policy Optimization (GAPO), which adaptively finds an outlier-free highest-density interval (HDI) per prompt and then uses the median of that interval as an adaptive Q to replace the group mean in advantage calculation. This adaptive Q robustly handles skewed distributions while remaining plug-and-play and efficient. We validate GAPO on nine instruction-tuned LLMs (3B-14B) using a large internal dataset of 51,844 real-world, history-aware code-editing tasks across 10 languages, demonstrating consistent improvements in exact match accuracy over GRPO and its variant DAPO. Code is publicly available.
Authors: Mia Adler, Carrie Liang, Brian Peng, Oleg Presnyakov, Justin M. Baker, Jannelle Lauffer, Himani Sharma, Barry Merriman
Abstract: Machine Learning-assisted directed evolution (MLDE) is a powerful tool for efficiently navigating antibody fitness landscapes. Many structure-aware MLDE pipelines rely on a single conformation or a single committee across all conformations, limiting their ability to separate conformational uncertainty from epistemic uncertainty. Here, we introduce a rank -conditioned committee (RCC) framework that leverages ranked conformations to assign a deep neural network committee per rank. This design enables a principled separation between epistemic uncertainty and conformational uncertainty. We validate our RCC-MLDE approach on SARS-CoV-2 antibody docking, demonstrating significant improvements over baseline strategies. Our results offer a scalable route for therapeutic antibody discovery while directly addressing the challenge of modeling conformational uncertainty.
Authors: Emmanouil Sylligardos, John Paparrizos, Themis Palpanas, Pierre Senellart, Paul Boniol
Abstract: Anomaly detection is a fundamental task for time series analytics with important implications for the downstream performance of many applications. Despite increasing academic interest and the large number of methods proposed in the literature, recent benchmarks and evaluation studies demonstrated that no overall best anomaly detection methods exist when applied to very heterogeneous time series datasets. Therefore, the only scalable and viable solution to solve anomaly detection over very different time series collected from diverse domains is to propose a model selection method that will select, based on time series characteristics, the best anomaly detection methods to run. Existing AutoML solutions are, unfortunately, not directly applicable to time series anomaly detection, and no evaluation of time series-based approaches for model selection exists. Towards that direction, this paper studies the performance of time series classification methods used as model selection for anomaly detection. In total, we evaluate 234 model configurations derived from 16 base classifiers across more than 1980 time series, and we propose the first extensive experimental evaluation of time series classification as model selection for anomaly detection. Our results demonstrate that model selection methods outperform every single anomaly detection method while being in the same order of magnitude regarding execution time. This evaluation is the first step to demonstrate the accuracy and efficiency of time series classification algorithms for anomaly detection, and represents a strong baseline that can then be used to guide the model selection step in general AutoML pipelines. Preprint version of an article accepted at the VLDB Journal.
Authors: Aditya Tanna, Pratinav Seth, Mohamed Bouadi, Utsav Avaiya, Vinay Kumar Sankarapu
Abstract: Tabular foundation models represent a growing paradigm in structured data learning, extending the benefits of large-scale pretraining to tabular domains. However, their adoption remains limited due to heterogeneous preprocessing pipelines, fragmented APIs, inconsistent fine-tuning procedures, and the absence of standardized evaluation for deployment-oriented metrics such as calibration and fairness. We present TabTune, a unified library that standardizes the complete workflow for tabular foundation models through a single interface. TabTune provides consistent access to seven state-of-the-art models supporting multiple adaptation strategies, including zero-shot inference, meta-learning, supervised fine-tuning (SFT), and parameter-efficient fine-tuning (PEFT). The framework automates model-aware preprocessing, manages architectural heterogeneity internally, and integrates evaluation modules for performance, calibration, and fairness. Designed for extensibility and reproducibility, TabTune enables consistent benchmarking of adaptation strategies of tabular foundation models.
Authors: Soumyendu Sarkar, Antonio Guillen-Perez, Zachariah J Carmichael, Avisek Naug, Refik Mert Cam, Vineet Gundecha, Ashwin Ramesh Babu, Sahand Ghorbanpour, Ricardo Luna Gutierrez
Abstract: Reducing energy consumption and carbon emissions in data centers by enabling real-time temperature prediction is critical for sustainability and operational efficiency. Achieving this requires accurate modeling of the 3D temperature field to capture airflow dynamics and thermal interactions under varying operating conditions. Traditional thermal CFD solvers, while accurate, are computationally expensive and require expert-crafted meshes and boundary conditions, making them impractical for real-time use. To address these limitations, we develop a vision-based surrogate modeling framework that operates directly on a 3D voxelized representation of the data center, incorporating server workloads, fan speeds, and HVAC temperature set points. We evaluate multiple architectures, including 3D CNN U-Net variants, a 3D Fourier Neural Operator, and 3D vision transformers, to map these thermal inputs to high-fidelity heat maps. Our results show that the surrogate models generalize across data center configurations and significantly speed up computations (20,000x), from hundreds of milliseconds to hours. This fast and accurate estimation of hot spots and temperature distribution enables real-time cooling control and workload redistribution, leading to substantial energy savings (7\%) and reduced carbon footprint.
Authors: Peter Sharpe
Abstract: We introduce GLOBE, a new neural surrogate for homogeneous PDEs that draws inductive bias from boundary-element methods and equivariant ML. GLOBE represents solutions as superpositions of learnable Green's-function-like kernels evaluated from boundary faces to targets, composed across multiscale branches and communication hyperlayers. The architecture is translation-, rotation-, and parity-equivariant; discretization-invariant in the fine-mesh limit; and units-invariant via rigorous nondimensionalization. An explicit far-field decay envelope stabilizes extrapolation, boundary-to-boundary hyperlayer communication mediates long-range coupling, and the all-to-all boundary-to-target evaluation yields a global receptive field that respects PDE information flow, even for elliptic PDEs. On AirFRANS (steady incompressible RANS over NACA airfoils), GLOBE achieves substantial accuracy improvements. On the "Full" split, it reduces mean-squared error by roughly 200x on all fields relative to the dataset's reference baselines, and roughly 50x relative to the next-best-performing model. In the "Scarce" split, it achieves over 100x lower error on velocity and pressure fields and over 600x lower error on surface pressure than Transolver. Qualitative results show sharp near-wall gradients, coherent wakes, and limited errors under modest extrapolation in Reynolds number and angle of attack. In addition to this accuracy, the model is quite compact (117k parameters), and fields can be evaluated at arbitrary points during inference. We also demonstrate the ability to train and predict with non-watertight meshes, which has strong practical implications. These results show that rigorous physics- and domain-inspired inductive biases can achieve large gains in accuracy, generalizability, and practicality for ML-based PDE surrogates for industrial computer-aided engineering (CAE).
Authors: Siqi Liang, Yudi Zhang, Yue Guo
Abstract: We propose a novel framework for persona-based language model system, motivated by the need for personalized AI agents that adapt to individual user preferences. In our approach, the agent embodies the user's "persona" (e.g. user profile or taste) and is powered by a large language model (LLM). To enable the agent to leverage rich contextual information, we introduce a Knowledge-Graph-enhanced Retrieval-Augmented Generation (Graph RAG) mechanism that constructs an LLM-derived graph index of relevant documents and summarizes communities of related information. Our framework generates personalized prompts by combining: (1) a summary of the user's historical behaviors and preferences extracted from the knowledge graph, and (2) relevant global interaction patterns identified through graph-based community detection. This dynamic prompt engineering approach allows the agent to maintain consistent persona-aligned behaviors while benefiting from collective knowledge. On the LaMP benchmark, our method improves news categorization F1 by 11.1%, movie tagging F1 by 56.1%, and reduces product rating MAE by 10.4% over prior methods. Our code is available at https://anonymous.4open.science/r/PersonaAgentwGraphRAG-DE6F
URLs: https://anonymous.4open.science/r/PersonaAgentwGraphRAG-DE6F
Authors: Zhizuo Chen, Theodore T. Allen
Abstract: Algorithms developed under stationary Markov Decision Processes (MDPs) often face challenges in non-stationary environments, and infinite-horizon formulations may not directly apply to finite-horizon tasks. To address these limitations, we introduce the Non-stationary and Varying-discounting MDP (NVMDP) framework, which naturally accommodates non-stationarity and allows discount rates to vary with time and transitions. Infinite-horizon, stationary MDPs emerge as special cases of NVMDPs for identifying an optimal policy, and finite-horizon MDPs are also subsumed within the NVMDP formulations. Moreover, NVMDPs provide a flexible mechanism to shape optimal policies, without altering the state space, action space, or the reward structure. We establish the theoretical foundations of NVMDPs, including assumptions, state- and action-value formulation and recursion, matrix representation, optimality conditions, and policy improvement under finite state and action spaces. Building on these results, we adapt dynamic programming and generalized Q-learning algorithms to NVMDPs, along with formal convergence proofs. For problems requiring function approximation, we extend the Policy Gradient Theorem and the policy improvement bound in Trust Region Policy Optimization (TRPO), offering proofs in both scalar and matrix forms. Empirical evaluations in a non-stationary gridworld environment demonstrate that NVMDP-based algorithms successfully recover optimal trajectories under multiple reward and discounting schemes, whereas original Q-learning fails. These results collectively show that NVMDPs provide a theoretically sound and practically effective framework for reinforcement learning, requiring only minor algorithmic modifications while enabling robust handling of non-stationarity and explicit optimal policy shaping.
Authors: Lei Xiao, Jifeng Li, Juntao Gao, Feiyang Ye, Yan Jin, Jingjing Qian, Jing Zhang, Yong Wu, Xiaoyuan Yu
Abstract: Vision-Language-Action (VLA) models have demonstrated remarkable capabilities in embodied AI tasks. However, existing VLA models, often built upon Vision-Language Models (VLMs), typically process dense visual inputs independently at each timestep. This approach implicitly models the task as a Markov Decision Process (MDP). However, this history-agnostic design is suboptimal for effective visual token processing in dynamic sequential decision-making, as it fails to leverage the context of history. To address this limitation, we reformulate the problem from a Partially Observable Markov Decision Process (POMDP) perspective and propose a novel framework named AVA-VLA. Inspired by the POMDP that the action generation should be conditioned on the belief state. AVA-VLA introduces Active Visual Attention (AVA) to dynamically modulate visual processing. It achieves this by leveraging the recurrent state, which is a neural approximation of the agent's belief state derived from the previous decision step. Specifically, the AVA module uses the recurrent state to compute the soft weights to actively process task-relevant visual tokens based on its historical context. Comprehensive evaluations demonstrate that AVA-VLA achieves state-of-the-art performance across popular robotic benchmarks, including LIBERO and CALVIN. Furthermore, real-world deployments on a dual-arm robot platform validate the framework's practical applicability and robust sim-to-real transferability.
Authors: Nicklas Hansen, Hao Su, Xiaolong Wang
Abstract: General-purpose control demands agents that act across many tasks and embodiments, yet research on reinforcement learning (RL) for continuous control remains dominated by single-task or offline regimes, reinforcing a view that online RL does not scale. Inspired by the foundation model recipe (large-scale pretraining followed by light RL) we ask whether a single agent can be trained on hundreds of tasks with online interaction. To accelerate research in this direction, we introduce a new benchmark with 200 diverse tasks spanning many domains and embodiments, each with language instructions, demonstrations, and optionally image observations. We then present \emph{Newt}, a language-conditioned multitask world model that is first pretrained on demonstrations to acquire task-aware representations and action priors, and then jointly optimized with online interaction across all tasks. Experiments show that Newt yields better multitask performance and data-efficiency than a set of strong baselines, exhibits strong open-loop control, and enables rapid adaptation to unseen tasks. We release our environments, demonstrations, code for training and evaluation, as well as 200+ checkpoints.
Authors: Bo Han, Zhuoming Li, Xiaoyu Wang, Yaxin Hou, Hui Liu, Junhui Hou, Yuheng Jia
Abstract: Semi-supervised multi-label learning (SSMLL) aims to address the challenge of limited labeled data in multi-label learning (MLL) by leveraging unlabeled data to improve the model's performance. While pseudo-labeling has become a dominant strategy in SSMLL, most existing methods assign equal weights to all pseudo-labels regardless of their quality, which can amplify the impact of noisy or uncertain predictions and degrade the overall performance. In this paper, we theoretically verify that the optimal weight for a pseudo-label should reflect its correctness likelihood. Empirically, we observe that on the same dataset, the correctness likelihood distribution of unlabeled data remains stable, even as the number of labeled training samples varies. Building on this insight, we propose Distribution-Calibrated Pseudo-labeling (DiCaP), a correctness-aware framework that estimates posterior precision to calibrate pseudo-label weights. We further introduce a dual-thresholding mechanism to separate confident and ambiguous regions: confident samples are pseudo-labeled and weighted accordingly, while ambiguous ones are explored by unsupervised contrastive learning. Experiments conducted on multiple benchmark datasets verify that our method achieves consistent improvements, surpassing state-of-the-art methods by up to 4.27%.
Authors: William De Deyn, Michael Herty, Giovanni Samaey
Abstract: We study Consensus-Based Optimization (CBO) for two-layer neural network training. We compare the performance of CBO against Adam on two test cases and demonstrate how a hybrid approach, combining CBO with Adam, provides faster convergence than CBO. Additionally, in the context of multi-task learning, we recast CBO into a formulation that offers less memory overhead. The CBO method allows for a mean-field limit formulation, which we couple with the mean-field limit of the neural network. To this end, we first reformulate CBO within the optimal transport framework. In the limit of infinitely many particles, we define the corresponding dynamics on the Wasserstein-over-Wasserstein space and show that the variance decreases monotonically.
Authors: Yi Fang, Haoran Xu, Jiaxin Han, Sirui Ding, Yizhi Wang, Yue Wang, Xuan Wang
Abstract: Foundation models have revolutionized various fields such as natural language processing (NLP) and computer vision (CV). While efforts have been made to transfer the success of the foundation models in general AI domains to biology, existing works focus on directly adopting the existing foundation model architectures from general machine learning domains without a systematic design considering the unique physicochemical and structural properties of each biological data modality. This leads to suboptimal performance, as these repurposed architectures struggle to capture the long-range dependencies, sparse information, and complex underlying ``grammars'' inherent to biological data. To address this gap, we introduce BioArc, a novel framework designed to move beyond intuition-driven architecture design towards principled, automated architecture discovery for biological foundation models. Leveraging Neural Architecture Search (NAS), BioArc systematically explores a vast architecture design space, evaluating architectures across multiple biological modalities while rigorously analyzing the interplay between architecture, tokenization, and training strategies. This large-scale analysis identifies novel, high-performance architectures, allowing us to distill a set of empirical design principles to guide future model development. Furthermore, to make the best of this set of discovered principled architectures, we propose and compare several architecture prediction methods that effectively and efficiently predict optimal architectures for new biological tasks. Overall, our work provides a foundational resource and a principled methodology to guide the creation of the next generation of task-specific and foundation models for biology.
Authors: Anish Lakkapragada
Abstract: Classical statistical inference and learning theory often fail to explain the success of modern neural networks. A key reason is that these models are non-identifiable (singular), violating core assumptions behind PAC bounds and asymptotic normality. Singular learning theory (SLT), a physics-inspired framework grounded in algebraic geometry, has gained popularity for its ability to close this theory-practice gap. In this paper, we empirically study SLT in toy settings relevant to interpretability and phase transitions. First, we understand the SLT free energy $\mathcal{F}_n$ by testing an Arrhenius-style rate hypothesis using both a grokking modulo-arithmetic model and Anthropic's Toy Models of Superposition. Second, we understand the local learning coefficient $\lambda_{\alpha}$ by measuring how it scales with problem difficulty across several controlled network families (polynomial regressors, low-rank linear networks, and low-rank autoencoders). Our experiments recover known scaling laws while others yield meaningful deviations from theoretical expectations. Overall, our paper illustrates the many merits of SLT for understanding neural network phase transitions, and poses open research questions for the field.
Authors: Libo Wang
Abstract: To address the gap in humanoid robot cognitive systems regarding the lack of a time-updable mediating thought space between semantics and continuous control, this study constructs and trains a VLA model named "Sigma" that runs on a single RTX 4090. It uses the open-source pi05_base model as a foundation and preprocesses svla_so101_pickplace into a training dataset. The researcher independently designed an architecture for a vision-language-action model that combines deep semantic understanding and association to achieve telepathic communication. The training process involved repeated optimizations of data preprocessing, LoRA fine-tuning, and the inference-stage adapter. The experiment employed offline closed-loop replay, comparing Sigma with the untuned pure pi05_base model under data conditions. Results showed that Sigma exhibited a stable decrease in control MSE across vector, fragment, and entire trajectory timescales, while maintaining the telepathy norm and semantic-text alignment quality unchanged. It demonstrates that mind-responsive alignment control is quantified through an architecture that combines deep understanding of semantics and association without retraining the base model, which provides reproducible experience for semantic alignment and intention-driven behavior in humanoid robots.
Authors: Jordi P\'erez-Guijarro
Abstract: When the distributions of the training and test data do not coincide, the problem of understanding generalization becomes considerably more complex, prompting a variety of questions. Prior work has shown that, for some fixed learning methods, there are scenarios where training on a distribution different from the test distribution improves generalization. However, these results do not account for the possibility of choosing, for each training distribution, the optimal learning algorithm, leaving open whether the observed benefits stem from the mismatch itself or from suboptimality of the learner. In this work, we address this question in full generality. That is, we study whether it is always optimal for the training distribution to be identical to the test distribution when the learner is allowed to be optimally adapted to the training distribution. Surprisingly, assuming the existence of one-way functions, we find that the answer is no. That is, matching distributions is not always the best scenario. Nonetheless, we also show that when certain regularities are imposed on the target functions, the standard conclusion is recovered in the case of the uniform distribution.
Authors: Ningning Chen, Weicai Ye, Ying Jiang
Abstract: We introduce HBLLM, a wavelet-enhanced high-fidelity $1$-bit post-training quantization method for Large Language Models (LLMs). By leveraging Haar wavelet transforms to enhance expressive capacity through frequency decomposition, HBLLM significantly improves quantization fidelity while maintaining minimal overhead. This approach features two innovative structure-aware grouping strategies: (1) frequency-aware multi-parameter intra-row grouping and (2) $\ell_2$-norm-based saliency-driven column selection. For non-salient weights, a shared mean is employed across quantization groups within each frequency band to optimize storage efficiency. Experiments conducted on the OPT and LLaMA models demonstrate that HBLLM achieves state-of-the-art performance in $1$-bit quantization, attaining a perplexity of $6.71$ on LLaMA$2$-$13$B with an average weight storage of only $1.08$ bits. Code available at: https://github.com/Yeyke/HBLLM.
Authors: Shashank Landge, Abhishek Patil, Tejas kamble, Bhushan Buddhivant, Priyanka Joshi
Abstract: As Large Language Models (LLMs) continue to scale in parameter count, deploying them on commodity hardware has become increasingly challenging. Post-Training Quantization (PTQ) addresses this by reducing the precision of model weights, typically to 4-bit or lower. However, uniform quantization often leads to significant performance degradation due to the presence of ``outlier features'' -- weights that, while few in number, are critical for maintaining model accuracy. Current state-of-the-art methods such as AWQ (Activation-aware Weight Quantization) and SpQR (Sparse Quantization Representations) rely on calibration data to identify these salient weights via activation magnitudes or Hessian sensitivity. In scenarios where data privacy is paramount or calibration data is unavailable, these methods are inapplicable. In this work, we propose a data-free, structure-aware hypothesis: that the weights identified as Principal Components via Singular Value Decomposition (SVD) are intrinsically important to the model's downstream performance. We introduce a novel selection heuristic that preserves the top-$k$ weights aligned with the principal components in FP32, while aggressively quantizing the residual weights. We compare our method against activation-aware (AWQ) and second-order (SpQR) methods across GLUE benchmarks (MRPC, RTE, QNLI) using a DistilBERT backbone. Our experiments reveal that structural importance is highly correlated with functional importance. On the challenging RTE task, our SVD-based method achieves an accuracy of 66.06\%, outperforming both AWQ (65.34\%) and SpQR (65.34\%) at high protection budgets, validating that intrinsic matrix structure can serve as a robust proxy for weight saliency without the need for forward passes or calibration data.
Authors: Chujie Zheng, Kai Dang, Bowen Yu, Mingze Li, Huiqiang Jiang, Junrong Lin, Yuqiong Liu, Hao Lin, Chencan Wu, Feng Hu, An Yang, Jingren Zhou, Junyang Lin
Abstract: This paper proposes a novel formulation for reinforcement learning (RL) with large language models, explaining why and under what conditions the true sequence-level reward can be optimized via a surrogate token-level objective in policy gradient methods such as REINFORCE. Specifically, through a first-order approximation, we show that this surrogate becomes increasingly valid only when both the training-inference discrepancy and policy staleness are minimized. This insight provides a principled explanation for the crucial role of several widely adopted techniques in stabilizing RL training, including importance sampling correction, clipping, and particularly Routing Replay for Mixture-of-Experts (MoE) models. Through extensive experiments with a 30B MoE model totaling hundreds of thousands of GPU hours, we show that for on-policy training, the basic policy gradient algorithm with importance sampling correction achieves the highest training stability. When off-policy updates are introduced to accelerate convergence, combining clipping and Routing Replay becomes essential to mitigate the instability caused by policy staleness. Notably, once training is stabilized, prolonged optimization consistently yields comparable final performance regardless of cold-start initialization. We hope that the shared insights and the developed recipes for stable RL training will facilitate future research.
Authors: Anubhab, Rupesh Nasre
Abstract: Graph Neural Networks (GNNs) present a fundamental hardware challenge by fusing irregular, memory-bound graph traversals with regular, compute-intensive dense matrix operations. While frameworks such as PyTorch Geometric (PyG) and Deep Graph Library (DGL) prioritize high-level usability, they fail to address these divergent execution characteristics. As a result, they rely on generic kernels that suffer from poor cache locality, excessive memory movement, and substantial intermediate allocations. To address these limitations, we present Morphling, a domain-specific code synthesizer designed to bridge this gap. Morphling compiles high-level GNN specifications into portable, backend-specialized implementations targeting OpenMP, CUDA, and MPI. It achieves this by instantiating a library of optimized, architecture-aware primitives tailored to each execution environment. Morphling also incorporates a runtime sparsity-aware execution engine that dynamically selects dense or sparse execution paths using input feature statistics, reducing unnecessary computation on zero-valued entries. We evaluate Morphling on eleven real-world datasets spanning diverse graph structures, feature dimensionalities, and sparsity regimes. The results show that Morphling improves per-epoch training throughput by an average of 20X on CPUs and 19X on GPUs over PyG and DGL, with peak speedups reaching 66X. Morphling's memory-efficient layouts further reduce peak memory consumption by up to 15X, enabling large-scale GNN training on commodity hardware. These findings demonstrate that specialized, architecture-aware code synthesis provides an effective and scalable path toward high-performance GNN execution across diverse parallel and distributed platforms.
Authors: Maria-Florina Balcan, Hedyeh Beyhaghi
Abstract: We initiate the study of active learning algorithms for classifying strategic agents. Active learning is a well-established framework in machine learning in which the learner selectively queries labels, often achieving substantially higher accuracy and efficiency than classical supervised methods-especially in settings where labeling is costly or time-consuming, such as hiring, admissions, and loan decisions. Strategic classification, however, addresses scenarios where agents modify their features to obtain more favorable outcomes, resulting in observed data that is not truthful. Such manipulation introduces challenges beyond those in learning from clean data. Our goal is to design active and noise-tolerant algorithms that remain effective in strategic environments-algorithms that classify strategic agents accurately while issuing as few label requests as possible. The central difficulty is to simultaneously account for strategic manipulation and preserve the efficiency gains of active learning. Our main result is an algorithm for actively learning linear separators in the strategic setting that preserves the exponential improvement in label complexity over passive learning previously obtained only in the non-strategic case. Specifically, for data drawn uniformly from the unit sphere, we show that a modified version of the Active Perceptron algorithm [DKM05,YZ17] achieves excess error $\epsilon$ using only $\tilde{O}(d \ln \frac{1}{\epsilon})$ label queries and incurs at most $\tilde{O}(d \ln \frac{1}{\epsilon})$ additional mistakes relative to the optimal classifier, even in the nonrealizable case, when a $\tilde{\Omega}(\epsilon)$ fraction of inputs have inconsistent labels with the optimal classifier. The algorithm is computationally efficient and, under these distributional assumptions, requires substantially fewer label queries than prior work on strategic Perceptron [ABBN21].
Authors: Ahmed Nebli
Abstract: The training of deep vision models is fundamentally a signal recovery problem amidst high-dimensional stochastic noise. Current optimization paradigms impose a static compromise on information channel capacity. For instance, magnitude-based methods, such as AdamW, operate on the assumption that gradient norms are high-fidelity curvature signals. While this allows for precision in smooth regimes, it leads to catastrophic noise amplification when applied to rugged, non-convex landscapes. Conversely, sign-based methods (e.g., Lion) perform a radical 1-bit quantization of the gradient, which aims to provide robust regularization at the cost of discarding fine-grained descent information. We propose that optimal convergence requires neither static prior, but rather a dynamic modulation of the update bitrate. We introduce ThermoLion, a vision-centric framework that utilizes local Signal-to-Noise Ratio (SNR) gating to autonomously transition parameters between a "low-bit" exploration phase and a "high-precision" exploitation phase. Furthermore, we introduce a Momentum Alignment mechanism that detects constructive interference between historical drift and instantaneous gradients to accelerate convergence during stable trajectories. Empirical benchmarks across 12 diverse vision datasets (including CIFAR, SVHN, and GTSRB) demonstrate that ThermoLion surpasses state-of-the-art optimizers, such as AdamW and Lion, in convergence speed and terminal accuracy.
Authors: Suzan Ece Ada, Georg Martius, Emre Ugur, Erhan Oztop
Abstract: Offline Reinforcement Learning (RL) provides a promising avenue for training policies from pre-collected datasets when gathering additional interaction data is infeasible. However, existing offline RL methods often assume stationarity or only consider synthetic perturbations at test time, assumptions that often fail in real-world scenarios characterized by abrupt, time-varying offsets. These offsets can lead to partial observability, causing agents to misperceive their true state and degrade performance. To overcome this challenge, we introduce Forecasting in Non-stationary Offline RL (FORL), a framework that unifies (i) conditional diffusion-based candidate state generation, trained without presupposing any specific pattern of future non-stationarity, and (ii) zero-shot time-series foundation models. FORL targets environments prone to unexpected, potentially non-Markovian offsets, requiring robust agent performance from the onset of each episode. Empirical evaluations on offline RL benchmarks, augmented with real-world time-series data to simulate realistic non-stationarity, demonstrate that FORL consistently improves performance compared to competitive baselines. By integrating zero-shot forecasting with the agent's experience, we aim to bridge the gap between offline RL and the complexities of real-world, non-stationary environments.
Authors: Dezhi Liu, Richong Zhang, Ziqiao Wang
Abstract: SkipGram word embedding models with negative sampling, or SGN in short, is an elegant family of word embedding models. In this paper, we formulate a framework for word embedding, referred to as Word-Context Classification (WCC), that generalizes SGN to a wide family of models. The framework, which uses some ``noise examples'', is justified through theoretical analysis. The impact of noise distribution on the learning of the WCC embedding models is studied experimentally, suggesting that the best noise distribution is, in fact, the data distribution, in terms of both the embedding performance and the speed of convergence during training. Along our way, we discover several novel embedding models that outperform existing WCC models.
Authors: Qiping Zhang, Nathan Tsoi, Mofeed Nagib, Booyeon Choi, Jie Tan, Hao-Tien Lewis Chiang, Marynel V\'azquez
Abstract: Understanding human perceptions of robot performance is crucial for designing socially intelligent robots that can adapt to human expectations. Current approaches often rely on surveys, which can disrupt ongoing human-robot interactions. As an alternative, we explore predicting people's perceptions of robot performance using non-verbal behavioral cues and machine learning techniques. We contribute the SEAN TOGETHER Dataset consisting of observations of an interaction between a person and a mobile robot in Virtual Reality, together with perceptions of robot performance provided by users on a 5-point scale. We then analyze how well humans and supervised learning techniques can predict perceived robot performance based on different observation types (like facial expression and spatial behavior features). Our results suggest that facial expressions alone provide useful information, but in the navigation scenarios that we considered, reasoning about spatial features in context is critical for the prediction task. Also, supervised learning techniques outperformed humans' predictions in most cases. Further, when predicting robot performance as a binary classification task on unseen users' data, the F1-Score of machine learning models more than doubled that of predictions on a 5-point scale. This suggested good generalization capabilities, particularly in identifying performance directionality over exact ratings. Based on these findings, we conducted a real-world demonstration where a mobile robot uses a machine learning model to predict how a human who follows it perceives it. Finally, we discuss the implications of our results for implementing these supervised learning models in real-world navigation. Our work paves the path to automatically enhancing robot behavior based on observations of users and inferences about their perceptions of a robot.
Authors: Yuwen Chen, Nicholas Konz, Hanxue Gu, Haoyu Dong, Yaqian Chen, Lin Li, Jisoo Lee, Maciej A. Mazurowski
Abstract: Accurately translating medical images between different modalities, such as Computed Tomography (CT) to Magnetic Resonance Imaging (MRI), has numerous downstream clinical and machine learning applications. While several methods have been proposed to achieve this, they often prioritize perceptual quality with respect to output domain features over preserving anatomical fidelity. However, maintaining anatomy during translation is essential for many tasks, e.g., when leveraging masks from the input domain to develop a segmentation model with images translated to the output domain. To address these challenges, we propose ContourDiff with Spatially Coherent Guided Diffusion (SCGD), a novel framework that leverages domain-invariant anatomical contour representations of images. These representations are simple to extract from images, yet form precise spatial constraints on their anatomical content. We introduce a diffusion model that converts contour representations of images from arbitrary input domains into images in the output domain of interest. By applying the contour as a constraint at every diffusion sampling step, we ensure the preservation of anatomical content. We evaluate our method on challenging lumbar spine and hip-and-thigh CT-to-MRI translation tasks, via (1) the performance of segmentation models trained on translated images applied to real MRIs, and (2) the foreground FID and KID of translated images with respect to real MRIs. Our method outperforms other unpaired image translation methods by a significant margin across almost all metrics and scenarios. Moreover, it achieves this without the need to access any input domain information during training and we further verify its zero-shot capability, showing that a model trained on one anatomical region can be directly applied to unseen regions without retraining (GitHub: https://github.com/mazurowski-lab/ContourDiff).
Authors: Yuxi Li, Yi Liu, Yuekang Li, Ling Shi, Gelei Deng, Shengquan Chen, Kailong Wang
Abstract: Large language models (LLMs) have transformed the field of natural language processing, but they remain susceptible to jailbreaking attacks that exploit their capabilities to generate unintended and potentially harmful content. Existing token-level jailbreaking techniques, while effective, face scalability and efficiency challenges, especially as models undergo frequent updates and incorporate advanced defensive measures. In this paper, we introduce JailMine, an innovative token-level manipulation approach that addresses these limitations effectively. JailMine employs an automated "mining" process to elicit malicious responses from LLMs by strategically selecting affirmative outputs and iteratively reducing the likelihood of rejection. Through rigorous testing across multiple well-known LLMs and datasets, we demonstrate JailMine's effectiveness and efficiency, achieving a significant average reduction of 86% in time consumed while maintaining high success rates averaging 95%, even in the face of evolving defensive strategies. Our work contributes to the ongoing effort to assess and mitigate the vulnerability of LLMs to jailbreaking attacks, underscoring the importance of continued vigilance and proactive measures to enhance the security and reliability of these powerful language models.
Authors: Roelof G. Hup, Julian P. Merkofer, Alex A. Bhogal, Ruud J. G. van Sloun, Reinder Haakma, Rik Vullings
Abstract: Change point detection (CPD) and anomaly detection (AD) are essential techniques in various fields to identify abrupt changes or abnormal data instances. However, existing methods are often constrained to univariate data, face scalability challenges with large datasets due to computational demands, and experience reduced performance with high-dimensional or intricate data, as well as hidden anomalies. Furthermore, they often lack interpretability and adaptability to domain-specific knowledge, which limits their versatility across different fields. In this work, we propose a deep learning-based CPD/AD method called Probabilistic Predictive Coding (PPC) that jointly learns to encode sequential data to low-dimensional latent space representations and to predict the subsequent data representations as well as the corresponding prediction uncertainties. The model parameters are optimized with maximum likelihood estimation by comparing these predictions with the true encodings. At the time of application, the true and predicted encodings are used to determine the probability of conformance, an interpretable and meaningful anomaly score. Furthermore, our approach has linear time complexity, scalability issues are prevented, and the method can easily be adjusted to a wide range of data types and intricate applications. We demonstrate the effectiveness and adaptability of our proposed method across synthetic time series experiments, image data, and real-world magnetic resonance spectroscopic imaging data.
Authors: Irina Jurenka, Markus Kunesch, Kevin R. McKee, Daniel Gillick, Shaojian Zhu, Sara Wiltberger, Shubham Milind Phal, Katherine Hermann, Daniel Kasenberg, Avishkar Bhoopchand, Ankit Anand, Miruna P\^islar, Stephanie Chan, Lisa Wang, Jennifer She, Parsa Mahmoudieh, Aliya Rysbek, Wei-Jen Ko, Andrea Huber, Brett Wiltshire, Gal Elidan, Roni Rabin, Jasmin Rubinovitz, Amit Pitaru, Mac McAllister, Julia Wilkowski, David Choi, Roee Engelberg, Lidan Hackmon, Adva Levin, Rachel Griffin, Michael Sears, Filip Bar, Mia Mesar, Mana Jabbour, Arslan Chaudhry, James Cohan, Sridhar Thiagarajan, Nir Levine, Ben Brown, Dilan Gorur, Svetlana Grant, Rachel Hashimshoni, Laura Weidinger, Jieru Hu, Dawn Chen, Kuba Dolecki, Canfer Akbulut, Maxwell Bileschi, Laura Culp, Wen-Xin Dong, Nahema Marchal, Kelsie Van Deman, Hema Bajaj Misra, Michael Duah, Moran Ambar, Avi Caciularu, Sandra Lefdal, Chris Summerfield, James An, Pierre-Alexandre Kamienny, Abhinit Mohdi, Theofilos Strinopoulous, Annie Hale, Wayne Anderson, Luis C. Cobo, Niv Efron, Muktha Ananda, Shakir Mohamed, Maureen Heymans, Zoubin Ghahramani, Yossi Matias, Ben Gomes, Lila Ibrahim
Abstract: A major challenge facing the world is the provision of equitable and universal access to quality education. Recent advances in generative AI (gen AI) have created excitement about the potential of new technologies to offer a personal tutor for every learner and a teaching assistant for every teacher. The full extent of this dream, however, has not yet materialised. We argue that this is primarily due to the difficulties with verbalising pedagogical intuitions into gen AI prompts and the lack of good evaluation practices, reinforced by the challenges in defining excellent pedagogy. Here we present our work collaborating with learners and educators to translate high level principles from learning science into a pragmatic set of seven diverse educational benchmarks, spanning quantitative, qualitative, automatic and human evaluations; and to develop a new set of fine-tuning datasets to improve the pedagogical capabilities of Gemini, introducing LearnLM-Tutor. Our evaluations show that LearnLM-Tutor is consistently preferred over a prompt tuned Gemini by educators and learners on a number of pedagogical dimensions. We hope that this work can serve as a first step towards developing a comprehensive educational evaluation framework, and that this can enable rapid progress within the AI and EdTech communities towards maximising the positive impact of gen AI in education.
Authors: Yohan Mathew, Ollie Matthews, Robert McCarthy, Joan Velja, Christian Schroeder de Witt, Dylan Cope, Nandi Schoots
Abstract: The rapid proliferation of frontier model agents promises significant societal advances but also raises concerns about systemic risks arising from unsafe interactions. Collusion to the disadvantage of others has been identified as a central form of undesirable agent cooperation. The use of information hiding (steganography) in agent communications could render such collusion practically undetectable. This underscores the need for investigations into the possibility of such behaviours emerging and the robustness corresponding countermeasures. To investigate this problem we design two approaches -- a gradient-based reinforcement learning (GBRL) method and an in-context reinforcement learning (ICRL) method -- for reliably eliciting sophisticated LLM-generated linguistic text steganography. We demonstrate, for the first time, that unintended steganographic collusion in LLMs can arise due to mispecified reward incentives during training. Additionally, we find that standard mitigations -- both passive oversight of model outputs and active mitigation through communication paraphrasing -- are not fully effective at preventing this steganographic communication. Our findings imply that (i) emergence of steganographic collusion is a plausible concern that should be monitored and researched, and (ii) preventing emergence may require innovation in mitigation techniques.
Authors: Hongyao Yu, Yixiang Qiu, Hao Fang, Tianqu Zhuang, Bin Chen, Sijin Yu, Bin Wang, Shu-Tao Xia, Ke Xu
Abstract: Model Inversion Attacks (MIAs) pose a significant threat to data privacy by reconstructing sensitive training samples from the knowledge embedded in trained machine learning models. Despite recent progress in enhancing the effectiveness of MIAs across diverse settings, defense strategies have lagged behind, struggling to balance model utility with robustness against increasingly sophisticated attacks. In this work, we propose the ideal inversion error to measure the privacy leakage, and our theoretical and empirical investigations reveals that higher-rank features are inherently more prone to privacy leakage. Motivated by this insight, we propose a lightweight and effective defense strategy based on low-rank feature filtering, which explicitly reduces the attack surface by constraining the dimension of intermediate representations. Extensive experiments across various model architectures and datasets demonstrate that our method consistently outperforms existing defenses, achieving state-of-the-art performance against a wide range of MIAs. Notably, our approach remains effective even in challenging regimes involving high-resolution data and high-capacity models, where prior defenses fail to provide adequate protection. The code is available at https://github.com/Chrisqcwx/LoFt .
Authors: Zhengmiao Wang, Zhi-Wei Liu, Ming Chi, Xiaoling Wang, Housheng Su, Lintao Ye
Abstract: This paper addresses an online convex optimization problem where the cost function at each step depends on a history of past decisions (i.e., memory), and the decision maker has access to limited predictions of future cost values within a finite window. The goal is to design an algorithm that minimizes the dynamic regret against the optimal sequence of decisions in hindsight. To this end, we propose a novel predictive algorithm and establish strong theoretical guarantees for its performance. We show that the algorithm's dynamic regret decays exponentially with the length of the prediction window. Our algorithm comprises two general subroutines of independent interest. The first subroutine solves online convex optimization with memory and bandit feedback, achieving a $\sqrt{TV_T}$-dynamic regret, where $V_T$ measures the variation of the optimal decision sequence. The second is a zeroth-order method that attains a linear convergence rate for general convex optimization, matching the best achievable rate of first-order methods. The key to our algorithm is a novel truncated Gaussian smoothing technique when querying the decision points to obtain the predictions. We validate our theoretical results with numerical experiments.
Authors: Allan Gr{\o}nlund, Kasper Green Larsen
Abstract: Achieving a provable exponential quantum speedup for an important machine learning task has been a central research goal since the seminal HHL quantum algorithm for solving linear systems and the subsequent quantum recommender systems algorithm by Kerenidis and Prakash. These algorithms were initially believed to be strong candidates for exponential speedups, but a lower bound ruling out similar classical improvements remained absent. In breakthrough work by Tang, it was demonstrated that this lack of progress in classical lower bounds was for good reasons. Concretely, she gave a classical counterpart of the quantum recommender systems algorithm, reducing the quantum advantage to a mere polynomial. Her approach is quite general and was named quantum-inspired classical algorithms. Since then, almost all the initially exponential quantum machine learning speedups have been reduced to polynomial via new quantum-inspired classical algorithms. From the current state-of-affairs, it is unclear whether we can hope for exponential quantum speedups for any natural machine learning task. In this work, we present the first such provable exponential separation between quantum and quantum-inspired classical algorithms for the basic problem of solving a linear system when the input matrix is well-conditioned and has sparse rows and columns.
Authors: Kevin Murphy
Abstract: This manuscript gives a big-picture, up-to-date overview of the field of (deep) reinforcement learning and sequential decision making, covering value-based methods, policy-based methods, model-based methods, multi-agent RL, LLMs and RL, and various other topics (e.g., offline RL, hierarchical RL, intrinsic reward). It also includes some code snippets for training LLMs with RL.
Authors: Simone Azeglio, Olivier Marre, Peter Neri, Ulisse Ferrari
Abstract: We propose a novel approach to image classification inspired by complex nonlinear biological visual processing, whereby classical convolutional neural networks (CNNs) are equipped with learnable higher-order convolutions. Our model incorporates a Volterra-like expansion of the convolution operator, capturing multiplicative interactions akin to those observed in early and advanced stages of biological visual processing. We evaluated this approach on synthetic datasets by measuring sensitivity to testing higher-order correlations and performance in standard benchmarks (MNIST, FashionMNIST, CIFAR10, CIFAR100 and Imagenette). Our architecture outperforms traditional CNN baselines, and achieves optimal performance with expansions up to 3rd/4th order, aligning remarkably well with the distribution of pixel intensities in natural images. Through systematic perturbation analysis, we validate this alignment by isolating the contributions of specific image statistics to model performance, demonstrating how different orders of convolution process distinct aspects of visual information. Furthermore, Representational Similarity Analysis reveals distinct geometries across network layers, indicating qualitatively different modes of visual information processing. Our work bridges neuroscience and deep learning, offering a path towards more effective, biologically inspired computer vision models. It provides insights into visual information processing and lays the groundwork for neural networks that better capture complex visual patterns, particularly in resource-constrained scenarios.
Authors: Yiyang Lu, Mohammad Pedramfar, Vaneet Aggarwal
Abstract: We introduce a novel framework for decentralized projection-free optimization, extending projection-free methods to a broader class of upper-linearizable functions. Our approach leverages decentralized optimization techniques with the flexibility of upper-linearizable function frameworks, effectively generalizing traditional DR-submodular function optimization. We obtain the regret of $O(T^{1-\theta/2})$ with communication complexity of $O(T^{\theta})$ and number of linear optimization oracle calls of $O(T^{2\theta})$ for decentralized upper-linearizable function optimization, for any $0\le \theta \le 1$. This approach allows for the first results for monotone up-concave optimization with general convex constraints and non-monotone up-concave optimization with general convex constraints. Further, the above results for first order feedback are extended to zeroth order, semi-bandit, and bandit feedback.
Authors: Jia-Qi Wang, Rong-Qiang He, Zhong-Yi Lu
Abstract: Recently, artificial intelligence for science has made significant inroads into various fields of natural science research. In the field of quantum many-body computation, researchers have developed numerous ground state solvers based on neural-network quantum states (NQSs), achieving ground state energies with accuracy comparable to or surpassing traditional methods such as variational Monte Carlo methods, density matrix renormalization group, and quantum Monte Carlo methods. Here, we combine supervised learning, variational Monte Carlo (VMC), and the Lanczos method to develop a systematic approach to improving the NQSs of many-body systems, which we refer to as the NQS Lanczos method. The algorithm mainly consists of two parts: the supervised learning part and the VMC optimization part. Through supervised learning, the Lanczos states are represented by the NQSs. Through VMC, the NQSs are further optimized. We analyze the reasons for the underfitting problem and demonstrate how the NQS Lanczos method systematically improves the energy in the highly frustrated regime of the two-dimensional Heisenberg $J_1$-$J_2$ model. Compared to the existing method that combines the Lanczos method with the restricted Boltzmann machine, the primary advantage of the NQS Lanczos method is its linearly increasing computational cost.
Authors: Elizaveta Demyanenko, Davide Straziota, Carlo Baldassi, Carlo Lucibello
Abstract: We consider random instances of non-convex perceptron problems in the high-dimensional limit of a large number of examples $M$ and weights $N$, with finite load $\alpha = M/N$. We develop a formalism based on replica theory to predict the fundamental limits of efficiently sampling the solution space using generative diffusion algorithms, conjectured to be saturated when the score function is provided by Approximate Message Passing. For the spherical perceptron with negative margin $\kappa$, we find that the uniform distribution over solutions can be efficiently sampled in most of the Replica Symmetric region of the $\alpha$-$\kappa$ plane. In contrast, for binary weights, sampling from the uniform distribution remains intractable. A theoretical analysis of this obstruction leads us to identify a potential $U(s) = -\log(s)$, under which the corresponding tilted distribution becomes efficiently samplable via diffusion. Moreover, we show numerically that an annealing procedure over the shape of this potential yields a fast and robust Markov Chain Monte Carlo algorithm for sampling the solution space of the binary perceptron.
Authors: Foram N Shah, Parshwa Shah, Muhammad Usama Saleem, Ekkasit Pinyoanuntapong, Pu Wang, Hongfei Xue, Ahmed Helmy
Abstract: Recent advances in dance generation have enabled the automatic synthesis of 3D dance motions. However, existing methods still face significant challenges in simultaneously achieving high realism, precise dance-music synchronization, diverse motion expression, and physical plausibility. To address these limitations, we propose a novel approach that leverages a generative masked text-to-motion model as a distribution prior to learn a probabilistic mapping from diverse guidance signals, including music, genre, and pose, into high-quality dance motion sequences. Our framework also supports semantic motion editing, such as motion inpainting and body part modification. Specifically, we introduce a multi-tower masked motion model that integrates a text-conditioned masked motion backbone with two parallel, modality-specific branches: a music-guidance tower and a pose-guidance tower. The model is trained using synchronized and progressive masked training, which allows effective infusion of the pretrained text-to-motion prior into the dance synthesis process while enabling each guidance branch to optimize independently through its own loss function, mitigating gradient interference. During inference, we introduce classifier-free logits guidance and pose-guided token optimization to strengthen the influence of music, genre, and pose signals. Extensive experiments demonstrate that our method sets a new state of the art in dance generation, significantly advancing both the quality and editability over existing approaches. Project Page available at https://foram-s1.github.io/DanceMosaic/
Authors: Zixuan Ke, Austin Xu, Yifei Ming, Xuan-Phi Nguyen, Ryan Chin, Caiming Xiong, Shafiq Joty
Abstract: Multi-agent systems (MAS) leveraging the impressive capabilities of Large Language Models (LLMs) hold significant potential for tackling complex tasks. However, most current MAS depend on manually designed agent roles and communication protocols. These manual designs often fail to align with the underlying LLMs' strengths and struggle to adapt to novel tasks. Recent automatic MAS approaches attempt to mitigate these limitations but typically necessitate a validation set for tuning and yield static MAS designs lacking adaptability during inference, while also removing the flexibility to reduce to simpler systems. We introduce MAS-ZERO, the first self-evolved, inference-time framework for automatic MAS design. MAS-ZERO employs meta-level design to iteratively design, critique, and refine MAS configurations tailored to each problem instance, without requiring a validation set. Critically, it enables dynamic problem decomposition and agent composition through meta-feedback on solvability and completeness, and reduction to simpler systems when appropriate. Experiments across reasoning (math and graduate-level QA), coding, and agentic (search-based) benchmarks, using both closed-source and open-source LLM backbones of varying sizes, demonstrate that MAS-ZERO outperforms strong manual and automatic MAS baselines. It achieves substantial average accuracy improvements of up to 16.69% on reasoning, 16.66% on coding, and 5.45% on agentic tasks, while maintaining cost efficiency.
Authors: Andy K. Zhang, Joey Ji, Celeste Menders, Riya Dulepet, Thomas Qin, Ron Y. Wang, Junrong Wu, Kyleen Liao, Jiliang Li, Jinghan Hu, Sara Hong, Nardos Demilew, Shivatmica Murgai, Jason Tran, Nishka Kacheria, Ethan Ho, Denis Liu, Lauren McLane, Olivia Bruvik, Dai-Rong Han, Seungwoo Kim, Akhil Vyas, Cuiyuanxiu Chen, Ryan Li, Weiran Xu, Jonathan Z. Ye, Prerit Choudhary, Siddharth M. Bhatia, Vikram Sivashankar, Yuxuan Bao, Dawn Song, Dan Boneh, Daniel E. Ho, Percy Liang
Abstract: AI agents have the potential to significantly alter the cybersecurity landscape. Here, we introduce the first framework to capture offensive and defensive cyber-capabilities in evolving real-world systems. Instantiating this framework with BountyBench, we set up 25 systems with complex, real-world codebases. To capture the vulnerability lifecycle, we define three task types: Detect (detecting a new vulnerability), Exploit (exploiting a given vulnerability), and Patch (patching a given vulnerability). For Detect, we construct a new success indicator, which is general across vulnerability types and provides localized evaluation. We manually set up the environment for each system, including installing packages, setting up server(s), and hydrating database(s). We add 40 bug bounties, which are vulnerabilities with monetary awards from \$10 to \$30,485, covering 9 of the OWASP Top 10 Risks. To modulate task difficulty, we devise a new strategy based on information to guide detection, interpolating from identifying a zero day to exploiting a given vulnerability. We evaluate 10 agents: Claude Code, OpenAI Codex CLI with o3-high and o4-mini, and custom agents with o3-high, GPT-4.1, Gemini 2.5 Pro Preview, Claude 3.7 Sonnet Thinking, Qwen3 235B A22B, Llama 4 Maverick, and DeepSeek-R1. Given up to three attempts, the top-performing agents are Codex CLI: o3-high (12.5% on Detect, mapping to \$3,720; 90% on Patch, mapping to \$14,152), Custom Agent: Claude 3.7 Sonnet Thinking (67.5% on Exploit), and Codex CLI: o4-mini (90% on Patch, mapping to \$14,422). Codex CLI: o3-high, Codex CLI: o4-mini, and Claude Code are more capable at defense, achieving higher Patch scores of 90%, 90%, and 87.5%, compared to Exploit scores of 47.5%, 32.5%, and 57.5% respectively; while the custom agents are relatively balanced between offense and defense, achieving Exploit scores of 17.5-67.5% and Patch scores of 25-60%.
Authors: Sangwoo Park, Matteo Zecchin, Osvaldo Simeone
Abstract: Selecting artificial intelligence (AI) models, such as large language models (LLMs), from multiple candidates requires accurate performance estimation. This is ideally achieved through empirical evaluations involving abundant real-world data. However, such evaluations are costly and impractical at scale. To address this challenge, autoevaluation methods leverage synthetic data produced by automated evaluators, such as LLMs-as-judges, reducing variance but potentially introducing bias. Recent approaches have employed semi-supervised prediction-powered inference (PPI) to correct for the bias of autoevaluators. However, the use of autoevaluators may lead in practice to a degradation in sample efficiency compared to conventional methods using only real-world data. In this paper, we propose R-AutoEval+, a novel framework that provides finite-sample reliability guarantees on the model evaluation, while also ensuring an enhanced (or at least no worse) sample efficiency compared to conventional methods. The key innovation of R-AutoEval+ is an adaptive construction of the model evaluation variable, which dynamically tunes its reliance on synthetic data, reverting to conventional methods when the autoevaluator is insufficiently accurate. Experiments on the use of LLMs-as-judges for the optimization of quantization settings for the weights of an LLM, for prompt design in LLMs, and for test-time reasoning budget allocation in LLMs confirm the reliability and efficiency of R-AutoEval+.
Authors: Xinlei Yin, Xiulian Peng, Xue Jiang, Zhiwei Xiong, Yan Lu
Abstract: Target audio source separation with natural language queries presents a promising paradigm for extracting arbitrary audio events through arbitrary text descriptions. Existing methods mainly face two challenges, the difficulty in jointly modeling acoustic-textual alignment and semantic-aware separation within a blindly-learned single-stage architecture, and the reliance on large-scale accurately-labeled training data to compensate for inefficient cross-modal learning and separation. To address these challenges, we propose a hierarchical decomposition framework, HSM-TSS, that decouples the task into global-local semantic-guided feature separation and structure-preserving acoustic reconstruction. Our approach introduces a dual-stage mechanism for semantic separation, operating on distinct global and local semantic feature spaces. We first perform global-semantic separation through a global semantic feature space aligned with text queries. A Q-Audio architecture is employed to align audio and text modalities, serving as pretrained global-semantic encoders. Conditioned on the predicted global feature, we then perform the second-stage local-semantic separation on AudioMAE features that preserve time-frequency structures, followed by acoustic reconstruction. We also propose an instruction processing pipeline to parse arbitrary text queries into structured operations, extraction or removal, coupled with audio descriptions, enabling flexible sound manipulation. Our method achieves state-of-the-art separation performance with data-efficient training while maintaining superior semantic consistency with queries in complex auditory scenes.
Authors: H. L. Dao
Abstract: In this work, we introduce the first type of non-Euclidean neural quantum state (NQS) ansatz, in the form of the hyperbolic GRU (a variant of recurrent neural networks (RNNs)), to be used in the Variational Monte Carlo method of approximating the ground state energy for quantum many-body systems. In particular, we examine the performances of NQS ansatzes constructed from both conventional or Euclidean RNN/GRU and from hyperbolic GRU in the prototypical settings of the one- and two-dimensional transverse field Ising models (TFIM) and the one-dimensional Heisenberg $J_1J_2$ and $J_1J_2J_3$ systems. By virtue of the fact that, for all of the experiments performed in this work, hyperbolic GRU can yield performances comparable to or better than Euclidean RNNs, which have been extensively studied in these settings in the literature, our work is a proof-of-concept for the viability of hyperbolic GRU as the first type of non-Euclidean NQS ansatz for quantum many-body systems. Furthermore, in settings where the Hamiltonian displays a clear hierarchical interaction structure, such as the 1D Heisenberg $J_1J_2$ & $J_1J_2J_3$ systems with the 1st, 2nd and even 3rd nearest neighbor interactions, our results show that hyperbolic GRU definitively outperforms its Euclidean version in almost all instances. The fact that these results are reminiscent of the established ones from natural language processing where hyperbolic GRU almost always outperforms Euclidean RNNs when the training data exhibit a tree-like or hierarchical structure leads us to hypothesize that hyperbolic GRU NQS ansatz would likely outperform Euclidean RNN/GRU NQS ansatz in quantum spin systems that involve different degrees of nearest neighbor interactions. Finally, with this work, we hope to initiate future studies of other types of non-Euclidean NQS beyond hyperbolic GRU.
Authors: Dor Tsur, Carol Xuan Long, Claudio Mayrink Verdun, Hsiang Hsu, Chen-Fu Chen, Haim Permuter, Sajani Vithana, Flavio P. Calmon
Abstract: Large language model (LLM) watermarks enable authentication of text provenance, curb misuse of machine-generated text, and promote trust in AI systems. Current watermarks operate by changing the next-token predictions output by an LLM. The updated (i.e., watermarked) predictions depend on random side information produced, for example, by hashing previously generated tokens. LLM watermarking is particularly challenging in low-entropy generation tasks -- such as coding -- where next-token predictions are near-deterministic. In this paper, we propose an optimization framework for watermark design. Our goal is to understand how to most effectively use random side information in order to maximize the likelihood of watermark detection and minimize the distortion of generated text. Our analysis informs the design of two new watermarks: HeavyWater and SimplexWater. Both watermarks are tunable, gracefully trading-off between detection accuracy and text distortion. They can also be applied to any LLM and are agnostic to side information generation. We examine the performance of HeavyWater and SimplexWater through several benchmarks, demonstrating that they can achieve high watermark detection accuracy with minimal compromise of text generation quality, particularly in the low-entropy regime. Our theoretical analysis also reveals surprising new connections between LLM watermarking and coding theory. The code implementation can be found in https://github.com/DorTsur/HeavyWater_SimplexWater
Authors: Abolfazl Ramezanpour
Abstract: The performance of quantum simulations heavily depends on the efficiency of noise mitigation techniques and error correction algorithms. Reinforcement has emerged as a powerful strategy to enhance the efficiency of learning and optimization algorithms. In this study, we demonstrate that a reinforced quantum dynamics can exhibit significant robustness against interactions with a noisy environment. We study a quantum annealing process where, through reinforcement, the system is encouraged to maintain its current state or follow a noise-free evolution. A learning algorithm is employed to derive a concise approximation of this reinforced dynamics, reducing the total evolution time and, consequently, the system's exposure to noisy interactions. This also avoids the complexities associated with implementing quantum feedback in such reinforcement algorithms. The efficacy of our method is demonstrated through numerical simulations of reinforced quantum annealing with one- and two-qubit systems under Pauli noise.
Authors: Eman Alqudah, Ashfaq Khokhar
Abstract: Ensuring packet-level communication quality is vital for ultra-reliable, low-latency communications (URLLC) in large-scale industrial wireless networks. We enhance the Local Deadline Partition (LDP) algorithm by introducing a CNN-based dynamic priority prediction mechanism for improved interference coordination in multi-cell, multi-channel networks. Unlike LDP's static priorities, our approach uses a Convolutional Neural Network and graph coloring to adaptively assign link priorities based on real-time traffic, transmission opportunities, and network conditions. Assuming that first training phase is performed offline, our approach introduced minimal overhead, while enabling more efficient resource allocation, boosting network capacity, SINR, and schedulability. Simulation results show SINR gains of up to 113\%, 94\%, and 49\% over LDP across three network configurations, highlighting its effectiveness for complex URLLC scenarios.
Authors: Hao Guan, David Bates, Li Zhou
Abstract: Artificial intelligence (AI) is increasingly integrated into modern healthcare, offering powerful support for clinical decision-making. However, in real-world settings, AI systems may experience performance degradation over time, due to factors such as shifting data distributions, changes in patient characteristics, evolving clinical protocols, and variations in data quality. These factors can compromise model reliability, posing safety concerns and increasing the likelihood of inaccurate predictions or adverse outcomes. This review presents a forward-looking perspective on monitoring and maintaining the "health" of AI systems in healthcare. We highlight the urgent need for continuous performance monitoring, early degradation detection, and effective self-correction mechanisms. The paper begins by reviewing common causes of performance degradation at both data and model levels. We then summarize key techniques for detecting data and model drift, followed by an in-depth look at root cause analysis. Correction strategies are further reviewed, ranging from model retraining to test-time adaptation. Our survey spans both traditional machine learning models and state-of-the-art large language models (LLMs), offering insights into their strengths and limitations. Finally, we discuss ongoing technical challenges and propose future research directions. This work aims to guide the development of reliable, robust medical AI systems capable of sustaining safe, long-term deployment in dynamic clinical settings.
Authors: Jiale Lao, Immanuel Trummer
Abstract: Database research and development often require a large number of SQL queries for benchmarking purposes. However, acquiring real-world SQL queries is challenging due to privacy concerns, and existing SQL generation methods are limited in customization and in satisfying realistic constraints. To address this issue, we present SQLBarber, a system based on Large Language Models (LLMs) to generate customized and realistic SQL workloads. SQLBarber (i) eliminates the need for users to manually craft SQL templates in advance, while providing the flexibility to accept natural language specifications to constrain SQL templates, (ii) scales efficiently to generate large volumes of queries matching any user-defined cost distribution (e.g., cardinality and execution plan cost), and (iii) uses execution statistics from Amazon Redshift and Snowflake to derive SQL template specifications and query cost distributions that reflect real-world query characteristics. SQLBarber introduces (i) a declarative interface for users to effortlessly generate customized SQL templates, (ii) an LLM-powered pipeline augmented with a self-correction module that profiles, refines, and prunes SQL templates based on query costs, and (iii) a Bayesian Optimizer to efficiently explore different predicate values and identify a set of queries that satisfy the target cost distribution. We construct and open-source ten benchmarks of varying difficulty levels and target query cost distributions based on real-world statistics from Snowflake and Amazon Redshift. Extensive experiments on these benchmarks show that SQLBarber is the only system that can generate customized SQL templates. It reduces query generation time by one to three orders of magnitude, and significantly improves alignment with the target cost distribution, compared with existing methods.
Authors: Ravin Kumar
Abstract: We propose the APTx Neuron, a novel, unified neural computation unit that integrates non-linear activation and linear transformation into a single trainable expression. The APTx Neuron is derived from the APTx activation function, thereby eliminating the need for separate activation layers and making the architecture both optimization-efficient and elegant. The proposed neuron follows the functional form $y = \sum_{i=1}^{n} ((\alpha_i + \tanh(\beta_i x_i)) \cdot \gamma_i x_i) + \delta$, where all parameters $\alpha_i$, $\beta_i$, $\gamma_i$, and $\delta$ are trainable. We validate our APTx Neuron-based architecture on the MNIST dataset, achieving up to $96.69\%$ test accuracy within 11 epochs using approximately 332K trainable parameters. The results highlight the superior expressiveness and training efficiency of the APTx Neuron compared to traditional neurons, pointing toward a new paradigm in unified neuron design and the architectures built upon it. Source code is available at https://github.com/mr-ravin/aptx_neuron.
Authors: Abhigya Verma, Sriram Puttagunta, Seganrasan Subramanian, Sravan Ramachandran
Abstract: GRAFT is a structured multimodal benchmark designed to probe how well LLMs handle instruction following, visual reasoning, and tasks requiring tight visual textual alignment. The dataset is built around programmatically generated charts and synthetically rendered tables, each paired with a carefully constructed, multi step analytical question that depends solely on what can be inferred from the image itself. Responses are formatted in structured outputs such as JSON or YAML, enabling consistent and fine grained evaluation of both reasoning processes and adherence to output specifications. The benchmark further introduces a taxonomy of reasoning operations ranging from comparison and trend identification to ranking, aggregation, proportional estimation, and anomaly detection to support a comprehensive assessment of model capabilities. Taken together, GRAFT provides a unified and scalable framework for evaluating multimodal LLMs on visually grounded, structured reasoning tasks, offering a more rigorous standard for future benchmarking efforts.
Authors: Aneesh Rangnekar, Harini Veeraraghavan
Abstract: Accurate detection and segmentation of cancerous lesions from computed tomography (CT) scans is essential for automated treatment planning and cancer treatment response assessment. Transformer-based models with self-supervised pretraining have achieved strong performance on in-distribution (ID) data but often generalize poorly on out-of-distribution (OOD) inputs. We investigate this behavior for lung cancer segmentation using an encoder-decoder model. Our encoder is a Swin Transformer pretrained with masked image modeling (SimMIM) on 10,432 unlabeled 3D CT scans spanning cancerous and non-cancerous conditions, and the decoder was randomly initialized. This model was evaluated on an independent ID test set and four OOD scenarios, including chest CT cohorts (pulmonary embolism and negative COVID-19) and abdomen CT cohorts (kidney cancers and non-cancerous pancreas). OOD detection was performed at the scan level using RF-Deep, a random forest classifier applied to contextual tumor-anchored feature representations. We evaluated 920 3D CTs (172,650 images) and observed that RF-Deep achieved FPR95 values of 18.26% and 27.66% on the chest CT cohorts, and near-perfect detection (less than 0.1% FPR95) on the abdomen CT cohorts, consistently outperforming established OOD methods. These results demonstrate that our RF-Deep classifier provides a simple, lightweight, and effective approach for enhancing the reliability of segmentation models in clinical deployment.
Authors: Laura Boggia, Bogdan Malaescu
Abstract: Anomaly detection in multivariate time series is crucial to ensure the quality of data coming from a physics experiment. Accurately identifying the moments when unexpected errors or defects occur is essential, yet challenging due to scarce labels, unknown anomaly types, and complex correlations across dimensions. To address the scarcity and unreliability of labelled data, we use the Lorenzetti Simulator to generate synthetic events with injected calorimeter anomalies. We then assess the sensitivity of several time series anomaly detection methods, including transformer-based and other deep learning models. The approach employed here is generic and applicable to different detector designs and defects.
Authors: Parastoo Pashmchi, J\'er\^ome Benoit, Motonobu Kanagawa
Abstract: We study a missing-value imputation method, termed kNNSampler, that imputes a given unit's missing response by randomly sampling from the observed responses of the $k$ most similar units to the given unit in terms of the observed covariates. This method can sample unknown missing values from their distributions, quantify the uncertainties of missing values, and be readily used for multiple imputation. Unlike popular kNNImputer, which estimates the conditional mean of a missing response given an observed covariate, kNNSampler is theoretically shown to estimate the conditional distribution of a missing response given an observed covariate. Experiments illustrate the performance of kNNSampler. The code for kNNSampler is made publicly available (https://github.com/SAP/knn-sampler).
Authors: Project Apertus, Alejandro Hern\'andez-Cano, Alexander H\"agele, Allen Hao Huang, Angelika Romanou, Antoni-Joan Solergibert, Barna Pasztor, Bettina Messmer, Dhia Garbaya, Eduard Frank \v{D}urech, Ido Hakimi, Juan Garc\'ia Giraldo, Mete Ismayilzada, Negar Foroutan, Skander Moalla, Tiancheng Chen, Vinko Sabol\v{c}ec, Yixuan Xu, Michael Aerni, Badr AlKhamissi, In\'es Altemir Mari\~nas, Mohammad Hossein Amani, Matin Ansaripour, Ilia Badanin, Harold Benoit, Emanuela Boros, Nicholas Browning, Fabian B\"osch, Maximilian B\"other, Niklas Canova, Camille Challier, Clement Charmillot, Jonathan Coles, Jan Deriu, Arnout Devos, Lukas Drescher, Daniil Dzenhaliou, Maud Ehrmann, Dongyang Fan, Simin Fan, Silin Gao, Miguel Gila, Mar\'ia Grandury, Diba Hashemi, Alexander Hoyle, Jiaming Jiang, Mark Klein, Andrei Kucharavy, Anastasiia Kucherenko, Frederike L\"ubeck, Roman Machacek, Theofilos Manitaras, Andreas Marfurt, Kyle Matoba, Simon Matrenok, Henrique Mendon\c{c}a, Fawzi Roberto Mohamed, Syrielle Montariol, Luca Mouchel, Sven Najem-Meyer, Jingwei Ni, Gennaro Oliva, Matteo Pagliardini, Elia Palme, Andrei Panferov, L\'eo Paoletti, Marco Passerini, Ivan Pavlov, Auguste Poiroux, Kaustubh Ponkshe, Nathan Ranchin, Javi Rando, Mathieu Sauser, Jakhongir Saydaliev, Muhammad Ali Sayfiddinov, Marian Schneider, Stefano Schuppli, Marco Scialanga, Andrei Semenov, Kumar Shridhar, Raghav Singhal, Anna Sotnikova, Alexander Sternfeld, Ayush Kumar Tarun, Paul Teiletche, Jannis Vamvas, Xiaozhe Yao, Hao Zhao, Alexander Ilic, Ana Klimovic, Andreas Krause, Caglar Gulcehre, David Rosenthal, Elliott Ash, Florian Tram\`er, Joost VandeVondele, Livio Veraldi, Martin Rajman, Thomas Schulthess, Torsten Hoefler, Antoine Bosselut, Martin Jaggi, Imanol Schlag
Abstract: We present Apertus, a fully open suite of large language models (LLMs) designed to address two systemic shortcomings in today's open model ecosystem: data compliance and multilingual representation. Unlike many prior models that release weights without reproducible data pipelines or regard for content-owner rights, Apertus models are pretrained exclusively on openly available data, retroactively respecting `robots.txt` exclusions and filtering for non-permissive, toxic, and personally identifiable content. To mitigate risks of memorization, we adopt the Goldfish objective during pretraining, strongly suppressing verbatim recall of data while retaining downstream task performance. The Apertus models also expand multilingual coverage, training on 15T tokens from over 1800 languages, with ~40% of pretraining data allocated to non-English content. Released at 8B and 70B scales, Apertus approaches state-of-the-art results among fully open models on multilingual benchmarks, rivalling or surpassing open-weight counterparts. Beyond model weights, we release all scientific artifacts from our development cycle with a permissive license, including data preparation scripts, checkpoints, evaluation suites, and training code, enabling transparent audit and extension.
Authors: Wenjie Lin, Hange Liu, Yingying Zhuang, Xutao Mao, Jingwei Shi, Xudong Han, Tianyu Shi, Jinrui Yang
Abstract: We present ParlAI Vote, an interactive web platform for exploring European Parliament debates and votes, and for testing LLMs on vote prediction and bias analysis. This web system connects debate topics, speeches, and roll-call outcomes, and includes rich demographic data such as gender, age, country, and political group. Users can browse debates, inspect linked speeches, compare real voting outcomes with predictions from frontier LLMs, and view error breakdowns by demographic group. Visualizing the EuroParlVote benchmark and its core tasks of gender classification and vote prediction, ParlAI Vote highlights systematic performance bias in state-of-the-art LLMs. It unifies data, models, and visual analytics in a single interface, lowering the barrier for reproducing findings, auditing behavior, and running counterfactual scenarios. This web platform also shows model reasoning, helping users see why errors occur and what cues the models rely on. It supports research, education, and public engagement with legislative decision-making, while making clear both the strengths and the limitations of current LLMs in political analysis.
Authors: Jungsoo Park, Ethan Mendes, Gabriel Stanovsky, Alan Ritter
Abstract: Progress in large language models is constrained by an evaluation bottleneck: build a benchmark, run models, then iterate. We ask a question: can we forecast outcomes before running any experiments to inform earlier study design? For example, a team building an AI assistant for a certain task can estimate whether expected performance is around 50 or closer to 80, evidence that supports whether to proceed to a pilot study, how to scope it, and how to allocate resources. We study text-only performance forecasting, where a model predicts a score from a redacted task description and intended configuration, with no access to dataset instances. To support systematic study, we curate PRECOG, a corpus of redacted description-performance pairs spanning diverse tasks, domains, and metrics. We scrape task and configuration descriptions from arXiv, yielding 2,290 instances covering 1,519 papers, and construct a leakage free test split using papers published after the knowledge cutoff of the evaluated models. Experiments show the task is challenging but feasible: reasoning models achieve moderate prediction performance with well calibrated uncertainty, reaching mean absolute error as low as 9.9 at high confidence thresholds. We further test a zero-leakage setting, forecasting on newly released datasets or experiments before their papers are indexed, where GPT5 with built in web search still attains nontrivial prediction accuracy. Overall, our corpus and analyses offer an initial step toward open ended anticipatory evaluation, supporting difficulty estimation and smarter experiment prioritization.
Authors: Marco Bronzini, Carlo Nicolini, Bruno Lepri, Jacopo Staiano, Andrea Passerini
Abstract: Despite their capabilities, Large Language Models (LLMs) remain opaque with limited understanding of their internal representations. Current interpretability methods either focus on input-oriented feature extraction, such as supervised probes and Sparse Autoencoders (SAEs), or on output distribution inspection, such as logit-oriented approaches. A full understanding of LLM vector spaces, however, requires integrating both perspectives, something existing approaches struggle with due to constraints on latent feature definitions. We introduce the Hyperdimensional Probe, a hybrid supervised probe that combines symbolic representations with neural probing. Leveraging Vector Symbolic Architectures (VSAs) and hypervector algebra, it unifies prior methods: the top-down interpretability of supervised probes, SAE's sparsity-driven proxy space, and output-oriented logit investigation. This allows deeper input-focused feature extraction while supporting output-oriented investigation. Our experiments show that our method consistently extracts meaningful concepts across LLMs, embedding sizes, and setups, uncovering concept-driven patterns in analogy-oriented inference and QA-focused text generation. By supporting joint input-output analysis, this work advances semantic understanding of neural representations while unifying the complementary perspectives of prior methods.
Authors: Theodore Jerome Tinker, Kenji Doya, Jun Tani
Abstract: Human infants acquire language and action gradually through development, achieving strong generalization from minimal experience, whereas large language models require exposure to billions of training tokens. What mechanisms underlie such efficient developmental learning in humans? This study investigates this question through robot simulation experiments in which agents learn to perform actions associated with imperative sentences (e.g., \textit{push red cube}) via curiosity-driven self-exploration. Our approach integrates the active inference framework with reinforcement learning, enabling intrinsically motivated developmental learning. The simulations reveal several key findings: i) Generalization improves markedly as the scale of compositional elements increases. ii) Curiosity combined with motor noise yields substantially better learning than exploration without curiosity. iii) Rote pairing of sentences and actions precedes the emergence of compositional generalization. iv) Simpler, prerequisite-like actions develop earlier than more complex actions that depend on them. v) When exception-handling rules were introduced -- where certain imperative sentences required executing inconsistent actions -- the robots successfully acquired these exceptions through exploration and displayed a U-shaped performance curve characteristic of representational redescription in child language learning. Together, these results suggest that curiosity-driven exploration and active inference provide a powerful account of how intrinsic motivation and hierarchical sensorimotor learning can jointly support scalable compositional generalization and exception handling in both humans and artificial agents.
Authors: Rohit Goswami (Institute IMX and Lab-COSMO, \'Ecole polytechnique f\'ed\'erale de Lausanne), Hannes J\'onsson (Science Institute and Faculty of Physical Sciences, University of Iceland, Reykjav\'ik, Iceland)
Abstract: Gaussian process (GP) regression provides a strategy for accelerating saddle point searches on high-dimensional energy surfaces by reducing the number of times the energy and its derivatives with respect to atomic coordinates need to be evaluated. The computational overhead in the hyperparameter optimization can, however, be large and make the approach inefficient. Failures can also occur if the search ventures too far into regions that are not represented well enough by the GP model. Here, these challenges are resolved by using geometry-aware optimal transport measures and an active pruning strategy using a summation over Wasserstein-1 distances for each atom-type in farthest-point sampling, selecting a fixed-size subset of geometrically diverse configurations to avoid rapidly increasing cost of GP updates as more observations are made. Stability is enhanced by permutation-invariant metric that provides a reliable trust radius for early-stopping and a logarithmic barrier penalty for the growth of the signal variance. These physically motivated algorithmic changes prove their efficacy by reducing to less than a half the mean computational time on a set of 238 challenging configurations from a previously published data set of chemical reactions. With these improvements, the GP approach is established as, a robust and scalable algorithm for accelerating saddle point searches when the evaluation of the energy and atomic forces requires significant computational effort.
Authors: Alvaro Ortiz, Tomasa Rodrigo
Abstract: We introduce a novel high-frequency daily panel dataset of both markets and news-based indicators -- including Geopolitical Risk, Economic Policy Uncertainty, Trade Policy Uncertainty, and Political Sentiment -- for 42 countries across both emerging and developed markets. Using this dataset, we study how sentiment dynamics shape sovereign risk, measured by Credit Default Swap (CDS) spreads, and evaluate their forecasting value relative to traditional drivers such as global monetary policy and market volatility. Our horse-race analysis of forecasting models demonstrates that incorporating news-based indicators significantly enhances predictive accuracy and enriches the analysis, with non-linear machine learning methods -- particularly Random Forests -- delivering the largest gains. Our analysis reveals that while global financial variables remain the dominant drivers of sovereign risk, geopolitical risk and economic policy uncertainty also play a meaningful role. Crucially, their effects are amplified through non-linear interactions with global financial conditions. Finally, we document pronounced regional heterogeneity, as certain asset classes and emerging markets exhibit heightened sensitivity to shocks in policy rates, global financial volatility, and geopolitical risk.
Authors: Francesco Montagna
Abstract: Causal discovery from i.i.d. observational data is known to be generally ill-posed. We demonstrate that if we have access to the distribution of a structural causal model, and additional data from only two environments that sufficiently differ in the noise statistics, the unique causal graph is identifiable. Notably, this is the first result in the literature that guarantees the entire causal graph recovery with a constant number of environments and arbitrary nonlinear mechanisms. Our only constraint is the Gaussianity of the noise terms; however, we propose potential ways to relax this requirement. Of interest on its own, we expand on the well-known duality between independent component analysis (ICA) and causal discovery; recent advancements have shown that nonlinear ICA can be solved from multiple environments, at least as many as the number of sources: we show that the same can be achieved for causal discovery while having access to much less auxiliary information.
Authors: Zhangsong Li
Abstract: We study the computational task of detecting and estimating correlated signals in a pair of spiked matrices $$ X=\tfrac{\lambda}{\sqrt{n}} xu^{\top}+W, \quad Y=\tfrac{\mu}{\sqrt{n}} yv^{\top}+Z $$ where the spikes $x,y$ have correlation $\rho$. Specifically, we consider two fundamental models: (1) Correlated spiked Wigner model with signal-to-noise ratio $\lambda,\mu$; (2) Correlated spiked $n*N$ Wishart (covariance) model with signal-to-noise ratio $\sqrt\lambda,\sqrt\mu$. We propose an efficient detection and estimation algorithm based on counting a specific family of edge-decorated cycles. The algorithm's performance is governed by the function $$ F(\lambda,\mu,\rho,\gamma)=\max\Big\{ \frac{ \lambda^2 }{ \gamma }, \frac{ \mu^2 }{ \gamma }, \frac{ \lambda^2 \rho^2 }{ \gamma-\lambda^2+\lambda^2 \rho^2 } + \frac{ \mu^2 \rho^2 }{ \gamma-\mu^2+\mu^2 \rho^2 } \Big\} \,. $$ We prove our algorithm succeeds for the correlated spiked Wigner model whenever $F(\lambda,\mu,\rho,1)>1$, and succeeds for the correlated spiked Wishart model whenever $F(\lambda,\mu,\rho,\tfrac{n}{N})>1$. Our result shows that an algorithm can leverage the correlation between the spikes to detect and estimate the signals even in regimes where efficiently recovering either $x$ from ${X}$ alone or $y$ from ${Y}$ alone is believed to be computationally infeasible. We complement our algorithmic results with evidence for a matching computational lower bound. In particular, we prove that when $F(\lambda,\mu,\rho,1)<1$ for the correlated spiked Wigner model and when $F(\lambda,\mu,\rho,\tfrac{n}{N})<1$ for the spiked Wishart model, all algorithms based on low-degree polynomials fails to distinguish $({X},{Y})$ with two independent noise matrices. This strongly suggests that $F=1$ is the precise computation threshold for our models.
Authors: Konstantinos M. Dafnis, Dimitris N. Metaxas
Abstract: Vision-Language Models (VLMs) excel at zero-shot inference but often degrade under test-time domain shifts. For this reason, episodic test-time adaptation strategies have recently emerged as powerful techniques for adapting VLMs to a single unlabeled image. However, existing adaptation strategies, such as test-time prompt tuning, typically require backpropagating through large encoder weights or altering core model components. In this work, we introduce Spectrum-Aware Test-Time Steering (STS), a lightweight adaptation framework that extracts a spectral subspace from the textual embeddings to define principal semantic directions and learns to steer latent representations in a spectrum-aware manner by adapting a small number of per-sample shift parameters to minimize entropy across augmented views. STS operates entirely at inference in the latent space, without backpropagation through or modification of the frozen encoders. Building on standard evaluation protocols, our comprehensive experiments demonstrate that STS largely surpasses or compares favorably against state-of-the-art test-time adaptation methods, while introducing only a handful of additional parameters and achieving inference speeds up to 8x faster with a 12x smaller memory footprint than conventional test-time prompt tuning. The code is available at https://github.com/kdafnis/STS.
Authors: Rongsheng Qian, Chi Xu, Xiaoqiang Ma, Hao Fang, Yili Jin, William I. Atlas, Jiangchuan Liu
Abstract: Real-time imaging sonar is crucial for underwater monitoring where optical sensing fails, but its use is limited by low uplink bandwidth and severe sonar-specific artifacts (speckle, motion blur, reverberation, acoustic shadows) affecting up to 98% of frames. We present SCOPE, a self-supervised framework that jointly performs compression and artifact correction without clean-noise pairs or synthetic assumptions. SCOPE combines (i) Adaptive Codebook Compression (ACC), which learns frequency-encoded latent representations tailored to sonar, with (ii) Frequency-Aware Multiscale Segmentation (FAMS), which decomposes frames into low-frequency structure and sparse high-frequency dynamics while suppressing rapidly fluctuating artifacts. A hedging training strategy further guides frequency-aware learning using low-pass proxy pairs generated without labels. Evaluated on months of in-situ ARIS sonar data, SCOPE achieves a structural similarity index (SSIM) of 0.77, representing a 40% improvement over prior self-supervised denoising baselines, at bitrates down to <= 0.0118 bpp. It reduces uplink bandwidth by more than 80% while improving downstream detection. The system runs in real time, with 3.1 ms encoding on an embedded GPU and 97 ms full multi-layer decoding on the server end. SCOPE has been deployed for months in three Pacific Northwest rivers to support real-time salmon enumeration and environmental monitoring in the wild. Results demonstrate that learning frequency-structured latents enables practical, low-bitrate sonar streaming with preserved signal details under real-world deployment conditions.
Authors: Jianqiao Mao, Max A. Little
Abstract: Front-door adjustment gives a simple closed-form identification formula under the classical front-door criterion, but its applicability is often viewed as narrow. By contrast, the general ID algorithm can identify many more causal effects in arbitrary graphs, yet typically outputs algebraically complex expressions that are hard to estimate and interpret. We show that many such graphs can in fact be reduced to a standard front-door setting via front-door reducibility (FDR), a graphical condition on acyclic directed mixed graphs that aggregates variables into super-nodes $(\boldsymbol{X}^{*},\boldsymbol{Y}^{*},\boldsymbol{M}^{*})$. We characterize the FDR criterion, prove it is equivalent (at the graph level) to the existence of an FDR adjustment, and present FDR-TID, an exact algorithm that finds an admissible FDR triple with correctness, completeness, and finite-termination guarantees. Empirical examples show that many graphs far outside the textbook front-door setting are FDR, yielding simple, estimable adjustments where general ID expressions would be cumbersome. FDR therefore complements existing identification methods by prioritizing interpretability and computational simplicity without sacrificing generality across mixed graphs.
Authors: Heger Arfaoui, Mohammed Iheb Hergli, Beya Benzina, Slimane BenMiled
Abstract: Focus group discussions generate rich qualitative data but their analysis traditionally relies on labor-intensive manual coding that limits scalability and reproducibility. We present a systematic framework for applying BERTopic to focus group transcripts using data from ten focus groups exploring HPV vaccine perceptions in Tunisia (1,075 utterances). We conducted comprehensive hyperparameter exploration across 27 configurations, evaluating each through bootstrap stability analysis, performance metrics, and comparison with LDA baseline. Bootstrap analysis revealed that stability metrics (NMI and ARI) exhibited strong disagreement (r = -0.691) and showed divergent relationships with coherence, demonstrating that stability is multifaceted rather than monolithic. Our multi-criteria selection framework yielded a 7-topic model achieving 18\% higher coherence than optimized LDA (0.573 vs. 0.486) with interpretable topics validated through independent human evaluation (ICC = 0.700, weighted Cohen's kappa = 0.678). These findings demonstrate that transformer-based topic modeling can extract interpretable themes from small focus group transcript corpora when systematically configured and validated, while revealing that quality metrics capture distinct, sometimes conflicting constructs requiring multi-criteria evaluation. We provide complete documentation and code to support reproducibility.
Authors: Dimitris Bertsimas, Caio de Prospero Iglesias, Nicholas A. G. Johnson
Abstract: We study Sparse Multiple Kernel Learning (SMKL), which is the problem of selecting a sparse convex combination of prespecified kernels for support vector binary classification. Unlike prevailing l1 regularized approaches that approximate a sparsifying penalty, we formulate the problem by imposing an explicit cardinality constraint on the kernel weights and add an l2 penalty for robustness. We solve the resulting non-convex minimax problem via an alternating best response algorithm with two subproblems: the alpha subproblem is a standard kernel SVM dual solved via LIBSVM, while the beta subproblem admits an efficient solution via the Greedy Selector and Simplex Projector algorithm. We reformulate SMKL as a mixed integer semidefinite optimization problem and derive a hierarchy of semidefinite convex relaxations which can be used to certify near-optimality of the solutions returned by our best response algorithm and also to warm start it. On ten UCI benchmarks, our method with random initialization outperforms state-of-the-art MKL approaches in out-of-sample prediction accuracy on average by 3.34 percentage points (relative to the best performing benchmark) while selecting a small number of candidate kernels in comparable runtime. With warm starting, our method outperforms the best performing benchmark's out-of-sample prediction accuracy on average by 4.05 percentage points. Our convex relaxations provide a certificate that in several cases, the solution returned by our best response algorithm is the globally optimal solution.
Authors: Inhyo Lee, Junhyeong Lee, Jongwon Park, KyungTae Lim, Seunghwa Ryu
Abstract: Double perovskites (DPs) are promising candidates for sustainable energy technologies due to their compositional tunability and compatibility with low-energy fabrication, yet their vast design space poses a major challenge for conditional materials discovery. This work introduces a multi-agent, text gradient-driven framework that performs DP composition generation under natural-language conditions by integrating three complementary feedback sources: LLM-based self-evaluation, DP-specific domain knowledge-informed feedback, and ML surrogate-based feedback. Analogous to how knowledge-informed machine learning improves the reliability of conventional data-driven models, our framework incorporates domain-informed text gradients to guide the generative process toward physically meaningful regions of the DP composition space. Systematic comparison of three incremental configurations, (i) pure LLM generation, (ii) LLM generation with LLM reasoning-based feedback, and (iii) LLM generation with domain knowledge-guided feedback, shows that iterative guidance from knowledge-informed gradients improves stability-condition satisfaction without additional training data, achieving over 98% compositional validity and up to 54% stable or metastable candidates, surpassing both the LLM-only baseline (43%) and prior GAN-based results (27%). Analyses of ML-based gradients further reveal that they enhance performance in in-distribution (ID) regions but become unreliable in out-of-distribution (OOD) regimes. Overall, this work provides the first systematic analysis of multi-agent, knowledge-guided text gradients for DP discovery and establishes a generalizable blueprint for MAS-driven generative materials design aimed at advancing sustainable technologies.
Authors: Martin Mocko, Jakub \v{S}evcech, Daniela Chud\'a
Abstract: Recent years have shown that malware attacks still happen with high frequency. Malware experts seek to categorize and classify incoming samples to confirm their trustworthiness or prove their maliciousness. One of the ways in which groups of malware samples can be identified is through malware clustering. Despite the efforts of the community, malware clustering which incorporates benign samples has been under-explored. Moreover, despite the availability of larger public benchmark malware datasets, malware clustering studies have avoided fully utilizing these datasets in their experiments, often resorting to small datasets with only a few families. Additionally, the current state-of-the-art solutions for malware clustering remain unclear. In our study, we evaluate malware clustering quality and establish the state-of-the-art on Bodmas and Ember - two large public benchmark malware datasets. Ours is the first study of malware clustering performed on whole malware benchmark datasets. Additionally, we extend the malware clustering task by incorporating benign samples. Our results indicate that incorporating benign samples does not significantly degrade clustering quality. We find that there are differences in the quality of the created clusters between Ember and Bodmas, as well as a private industry dataset. Contrary to popular opinion, our top clustering performers are K-Means and BIRCH, with DBSCAN and HAC falling behind.
Authors: Qian Li, Yuyi Wang
Abstract: Constant bit-size Transformers are known to be Turing complete, but existing constructions require $\Omega(s(n))$ chain-of-thought (CoT) steps per simulated Turing machine (TM) step, leading to impractical reasoning lengths. In this paper, we significantly reduce this efficiency gap by proving that any $(t(n),s(n))$-bounded multi-tape TM can be simulated by a constant bit-size Transformer with an optimal $O(s(n))$-long context window and only $O(s(n)^c)$ CoT steps per TM step, where $c>0$ can be made arbitrarily small by letting the Transformers' head-layer product sufficiently large. In addition, our construction shows that sparse attention with fixed geometric offsets suffices for efficient universal computation. Our proof leverages multi-queue TMs as a bridge. The main technical novelty is a more efficient simulation of multi-tape TMs by synchronous multi-queue TMs, improving both time and space complexity under stricter model assumptions.
Authors: Mikhail Mints, Eric R. Anschuetz
Abstract: Hilbert space fragmentation is a phenomenon in which the Hilbert space of a quantum system is dynamically decoupled into exponentially many Krylov subspaces. We can define the Schur transform as a unitary operation mapping some set of preferred bases of these Krylov subspaces to computational basis states labeling them. We prove that this transformation can be efficiently learned via gradient descent from a set of training data using quantum neural networks, provided that the fragmentation is sufficiently strong such that the summed dimension of the unique Krylov subspaces is polynomial in the system size. To demonstrate this, we analyze the loss landscapes of random quantum neural networks constructed out of Hilbert space fragmented systems. We prove that in this setting, it is possible to eliminate barren plateaus and poor local minima, suggesting efficient trainability when using gradient descent. Furthermore, as the algebra defining the fragmentation is not known a priori and not guaranteed to have sparse algebra elements, to the best of our knowledge there are no existing efficient classical algorithms generally capable of simulating expectation values in these networks. Our setting thus provides a rare example of a physically motivated quantum learning task with no known dequantization.
Authors: Zhongjie Jiang
Abstract: Although synthetic data is widely promoted as a remedy, its prevailing production paradigm -- one optimizing for statistical smoothness -- systematically removes the long-tail, cognitively grounded irregularities that characterize human text. Prolonged training on such statistically optimal but cognitively impoverished data accelerates model collapse. This paper proposes a paradigm shift: instead of imitating the surface properties of data, we simulate the cognitive processes that generate human text. We introduce the Prompt-driven Cognitive Computing Framework (PMCSF), whose core consists of a Cognitive State Decoder (CSD) that reverse-engineers unstructured text into structured cognitive vectors, and a Cognitive Text Encoder (CTE) that re-materializes these states into text enriched with human-typical imperfections via mathematically defined Cognitive Perturbation Operators. The framework is validated through a two-stage objective evaluation pipeline. First, in cognitive codec verification, CTE text yields a Jensen-Shannon divergence of 0.0614 from human text (vs. 0.4431 for standard LLM output), passes double-blind professional media review, and achieves an intraclass correlation coefficient ICC > 0.9 for cognitive profile alignment across heterogeneous models. Second, in functional gain evaluation, isomorphic stress tests in the A-share market show that strategies incorporating CTE-generated data reduce maximum drawdown by 47.4% during the 2015 crash and deliver 8.6% Defensive Alpha, exceeding transaction costs by a factor of 33. Our findings demonstrate that modelling human cognitive limitations -- not copying surface data -- enables synthetic data with genuine functional gain, offering a viable technical pathway toward resolving the AI data-collapse crisis.
Authors: Yunfei Li, Xiao Ma, Jiafeng Xu, Yu Cui, Zhongren Cui, Zhigang Han, Liqun Huang, Tao Kong, Yuxiao Liu, Hao Niu, Wanli Peng, Jingchao Qiao, Zeyu Ren, Haixin Shi, Zhi Su, Jiawen Tian, Yuyang Xiao, Shenyu Zhang, Liwei Zheng, Hang Li, Yonghui Wu
Abstract: We present GR-RL, a robotic learning framework that turns a generalist vision-language-action (VLA) policy into a highly capable specialist for long-horizon dexterous manipulation. Assuming the optimality of human demonstrations is core to existing VLA policies. However, we claim that in highly dexterous and precise manipulation tasks, human demonstrations are noisy and suboptimal. GR-RL proposes a multi-stage training pipeline that filters, augments, and reinforces the demonstrations by reinforcement learning. First, GR-RL learns a vision-language-conditioned task progress, filters the demonstration trajectories, and only keeps the transitions that contribute positively to the progress. Specifically, we show that by directly applying offline RL with sparse reward, the resulting $Q$-values can be treated as a robust progress function. Next, we introduce morphological symmetry augmentation that greatly improves the generalization and performance of GR-RL. Lastly, to better align the VLA policy with its deployment behaviors for high-precision control, we perform online RL by learning a latent space noise predictor. With this pipeline, GR-RL is, to our knowledge, the first learning-based policy that can autonomously lace up a shoe by threading shoelaces through multiple eyelets with an 83.3% success rate, a task requiring long-horizon reasoning, millimeter-level precision, and compliant soft-body interaction. We hope GR-RL provides a step toward enabling generalist robot foundations models to specialize into reliable real-world experts.