Authors: Sajid Hussain, Muhammad Sohail, Nauman Ali Khan, Naima Iltaf, Ihtesham ul Islam
Abstract: Federated Learning (FL) has emerged as a transformative paradigm for distributed machine learning while preserving data privacy. However, existing approaches predominantly focus on model heterogeneity and aggregation techniques, largely overlooking the fundamental impact of dataset size characteristics on federated training dynamics. This paper introduces Size-Based Adaptive Federated Learning (SAFL), a novel progressive training framework that systematically organizes federated learning based on dataset size characteristics across heterogeneous multi-modal data. Our comprehensive experimental evaluation across 13 diverse datasets spanning 7 modalities (vision, text, time series, audio, sensor, medical vision, and multimodal) reveals critical insights: 1) an optimal dataset size range of 1000-1500 samples for federated learning effectiveness; 2) a clear modality performance hierarchy with structured data (time series, sensor) significantly outperforming unstructured data (text, multimodal); and 3) systematic performance degradation for large datasets exceeding 2000 samples. SAFL achieves an average accuracy of 87.68% across all datasets, with structured data modalities reaching 99%+ accuracy. The framework demonstrates superior communication efficiency, reducing total data transfer to 7.38 GB across 558 communications while maintaining high performance. Our real-time monitoring framework provides unprecedented insights into system resource utilization, network efficiency, and training dynamics. This work fills critical gaps in understanding how data characteristics should drive federated learning strategies, providing both theoretical insights and practical guidance for real-world FL deployments in neural network and learning systems.
Authors: Ugo Lomoio, Tommaso Mazza, Pierangelo Veltri, Pietro Hiram Guzzi
Abstract: The increasing availability of large-scale omics data calls for robust analytical frameworks capable of handling complex gene expression datasets while offering interpretable results. Recent advances in artificial intelligence have enabled the identification of aberrant molecular patterns distinguishing disease states from healthy controls. Coupled with improvements in model interpretability, these tools now support the identification of genes potentially driving disease phenotypes. However, current approaches to gene anomaly detection often remain limited to single datasets and lack accessible graphical interfaces. Here, we introduce E-ABIN, a general-purpose, explainable framework for Anomaly detection in Biological Networks. E-ABIN combines classical machine learning and graph-based deep learning techniques within a unified, user-friendly platform, enabling the detection and interpretation of anomalies from gene expression or methylation-derived networks. By integrating algorithms such as Support Vector Machines, Random Forests, Graph Autoencoders (GAEs), and Graph Adversarial Attributed Networks (GAANs), E-ABIN ensures a high predictive accuracy while maintaining interpretability. We demonstrate the utility of E-ABIN through case studies of bladder cancer and coeliac disease, where it effectively uncovers biologically relevant anomalies and offers insights into disease mechanisms.
Authors: Xin Li
Abstract: The Context-Content Uncertainty Principle (CCUP) proposes that inference under uncertainty is governed by an entropy asymmetry between context and content: high-entropy contexts must be interpreted through alignment with low-entropy, structured content. In this paper, we develop a layered computational framework that derives operational principles from this foundational asymmetry. At the base level, CCUP formalizes inference as directional entropy minimization, establishing a variational gradient that favors content-first structuring. Building upon this, we identify four hierarchical layers of operational principles: (\textbf{L1}) \emph{Core Inference Constraints}, including structure-before-specificity, asymmetric inference flow, cycle-consistent bootstrapping, and conditional compression, all shown to be mutually reducible; (\textbf{L2}) \emph{Resource Allocation Principles}, such as precision-weighted attention, asymmetric learning rates, and attractor-based memory encoding; (\textbf{L3}) \emph{Temporal Bootstrapping Dynamics}, which organize learning over time via structure-guided curricula; and (\textbf{L4}) \emph{Spatial Hierarchical Composition}, which integrates these mechanisms into self-organizing cycles of memory, inference, and planning. We present formal equivalence theorems, a dependency lattice among principles, and computational simulations demonstrating the efficiency gains of CCUP-aligned inference. This work provides a unified theoretical foundation for understanding how brains and machines minimize uncertainty through recursive structure-specificity alignment. The brain is not just an inference machine. It is a cycle-consistent entropy gradient resolver, aligning structure and specificity via path-dependent, content-seeded simulation.
Authors: Vineet Jain, Kusha Sareen, Mohammad Pedramfar, Siamak Ravanbakhsh
Abstract: Adapting a pretrained diffusion model to new objectives at inference time remains an open problem in generative modeling. Existing steering methods suffer from inaccurate value estimation, especially at high noise levels, which biases guidance. Moreover, information from past runs is not reused to improve sample quality, resulting in inefficient use of compute. Inspired by the success of Monte Carlo Tree Search, we address these limitations by casting inference-time alignment as a search problem that reuses past computations. We introduce a tree-based approach that samples from the reward-aligned target density by propagating terminal rewards back through the diffusion chain and iteratively refining value estimates with each additional generation. Our proposed method, Diffusion Tree Sampling (DTS), produces asymptotically exact samples from the target distribution in the limit of infinite rollouts, and its greedy variant, Diffusion Tree Search (DTS$^\star$), performs a global search for high reward samples. On MNIST and CIFAR-10 class-conditional generation, DTS matches the FID of the best-performing baseline with up to $10\times$ less compute. In text-to-image generation and language completion tasks, DTS$^\star$ effectively searches for high reward samples that match best-of-N with up to $5\times$ less compute. By reusing information from previous generations, we get an anytime algorithm that turns additional compute into steadily better samples, providing a scalable approach for inference-time alignment of diffusion models.
Authors: Kin Kwan Leung, Rasa Hosseinzadeh, Gabriel Loaiza-Ganem
Abstract: The manifold hypothesis asserts that data of interest in high-dimensional ambient spaces, such as image data, lies on unknown low-dimensional submanifolds. Diffusion models (DMs) -- which operate by convolving data with progressively larger amounts of Gaussian noise and then learning to revert this process -- have risen to prominence as the most performant generative models, and are known to be able to learn distributions with low-dimensional support. For a given datum in one of these submanifolds, we should thus intuitively expect DMs to have implicitly learned its corresponding local intrinsic dimension (LID), i.e. the dimension of the submanifold it belongs to. Kamkari et al. (2024b) recently showed that this is indeed the case by linking this LID to the rate of change of the log marginal densities of the DM with respect to the amount of added noise, resulting in an LID estimator known as FLIPD. LID estimators such as FLIPD have a plethora of uses, among others they quantify the complexity of a given datum, and can be used to detect outliers, adversarial examples and AI-generated text. FLIPD achieves state-of-the-art performance at LID estimation, yet its theoretical underpinnings are incomplete since Kamkari et al. (2024b) only proved its correctness under the highly unrealistic assumption of affine submanifolds. In this work we bridge this gap by formally proving the correctness of FLIPD under realistic assumptions. Additionally, we show that an analogous result holds when Gaussian convolutions are replaced with uniform ones, and discuss the relevance of this result.
Authors: Zhiqi Gao, Tianyi Li, Yurii Kvasiuk, Sai Chaitanya Tadepalli, Maja Rudolph, Daniel J. H. Chung, Frederic Sala, Moritz M\"unchmeyer
Abstract: Large language models (LLMs) have shown strong capabilities in complex reasoning, and test-time scaling techniques can enhance their performance with comparably low cost. Many of these methods have been developed and evaluated on mathematical reasoning benchmarks such as AIME. This paper investigates whether the lessons learned from these benchmarks generalize to the domain of advanced theoretical physics. We evaluate a range of common test-time scaling methods on the TPBench physics dataset and compare their effectiveness with results on AIME. To better leverage the structure of physics problems, we develop a novel, symbolic weak-verifier framework to improve parallel scaling results. Our empirical results demonstrate that this method significantly outperforms existing test-time scaling approaches on TPBench. We also evaluate our method on AIME, confirming its effectiveness in solving advanced mathematical problems. Our findings highlight the power of step-wise symbolic verification for tackling complex scientific problems.
Authors: Minh-Hao Van, Prateek Verma, Chen Zhao, Xintao Wu
Abstract: Foundation models (FMs) are catalyzing a transformative shift in materials science (MatSci) by enabling scalable, general-purpose, and multimodal AI systems for scientific discovery. Unlike traditional machine learning models, which are typically narrow in scope and require task-specific engineering, FMs offer cross-domain generalization and exhibit emergent capabilities. Their versatility is especially well-suited to materials science, where research challenges span diverse data types and scales. This survey provides a comprehensive overview of foundation models, agentic systems, datasets, and computational tools supporting this growing field. We introduce a task-driven taxonomy encompassing six broad application areas: data extraction, interpretation and Q\&A; atomistic simulation; property prediction; materials structure, design and discovery; process planning, discovery, and optimization; and multiscale modeling. We discuss recent advances in both unimodal and multimodal FMs, as well as emerging large language model (LLM) agents. Furthermore, we review standardized datasets, open-source tools, and autonomous experimental platforms that collectively fuel the development and integration of FMs into research workflows. We assess the early successes of foundation models and identify persistent limitations, including challenges in generalizability, interpretability, data imbalance, safety concerns, and limited multimodal fusion. Finally, we articulate future research directions centered on scalable pretraining, continual learning, data governance, and trustworthiness.
Authors: Todd Nief, David Reber, Sean Richardson, Ari Holtzman
Abstract: When an LLM learns a relation during finetuning (e.g., new movie releases, corporate mergers, etc.), where does this information go? Is it extracted when the model processes an entity, recalled just-in-time before a prediction, or are there multiple separate heuristics? Existing localization approaches (e.g. activation patching) are ill-suited for this analysis because they tend to replace parts of the residual stream, potentially deleting information. To fill this gap, we propose dynamic weight-grafting between fine-tuned and pre-trained language models to show that fine-tuned language models both (1) extract relation information learned during finetuning while processing entities and (2) ``recall" this information in later layers while generating predictions. In some cases, models need both of these pathways to correctly generate finetuned information while, in other cases, a single ``enrichment" or ``recall" pathway alone is sufficient. We examine the necessity and sufficiency of these information pathways, examining what layers they occur at, how much redundancy they exhibit, and which model components are involved -- finding that the ``recall" pathway occurs via both task-specific attention mechanisms and a relation extraction step in the output of the attention and the feedforward networks at the final layers before next token prediction.
Authors: Huangyuan Su, Mujin Kwun, Stephanie Gil, Sham Kakade, Nikhil Anand
Abstract: Training large language models is an expensive, compute-bound process that must be repeated as models scale, algorithms improve, and new data is collected. To address this, next-generation hardware accelerators increasingly support lower-precision arithmetic formats, such as the Microscaling (MX) formats introduced in NVIDIA's Blackwell architecture. These formats use a shared scale within blocks of parameters to extend representable range and perform forward/backward GEMM operations in reduced precision for efficiency gains. In this work, we investigate the challenges and viability of block-scaled precision formats during model training. Across nearly one thousand language models trained from scratch -- spanning compute budgets from $2 \times 10^{17}$ to $4.8 \times 10^{19}$ FLOPs and sweeping over a broad range of weight-activation precision combinations -- we consistently observe that training in MX formats exhibits sharp, stochastic instabilities in the loss, particularly at larger compute scales. To explain this phenomenon, we conduct controlled experiments and ablations on a smaller proxy model that exhibits similar behavior as the language model, sweeping across architectural settings, hyperparameters, and precision formats. These experiments motivate a simple model in which multiplicative gradient bias introduced by the quantization of layer-norm affine parameters and a small fraction of activations can trigger runaway divergence. Through \emph{in situ} intervention experiments on our proxy model, we demonstrate that instabilities can be averted or delayed by modifying precision schemes mid-training. Guided by these findings, we evaluate stabilization strategies in the LLM setting and show that certain hybrid configurations recover performance competitive with full-precision training. We release our code at https://github.com/Hither1/systems-scaling.
Authors: Xinghao Dong, Huchen Yang, Jin-Long Wu
Abstract: We propose a latent score-based generative AI framework for learning stochastic, non-local closure models and constitutive laws in nonlinear dynamical systems of computational mechanics. This work addresses a key challenge of modeling complex multiscale dynamical systems without a clear scale separation, for which numerically resolving all scales is prohibitively expensive, e.g., for engineering turbulent flows. While classical closure modeling methods leverage domain knowledge to approximate subgrid-scale phenomena, their deterministic and local assumptions can be too restrictive in regimes lacking a clear scale separation. Recent developments of diffusion-based stochastic models have shown promise in the context of closure modeling, but their prohibitive computational inference cost limits practical applications for many real-world applications. This work addresses this limitation by jointly training convolutional autoencoders with conditional diffusion models in the latent spaces, significantly reducing the dimensionality of the sampling process while preserving essential physical characteristics. Numerical results demonstrate that the joint training approach helps discover a proper latent space that not only guarantees small reconstruction errors but also ensures good performance of the diffusion model in the latent space. When integrated into numerical simulations, the proposed stochastic modeling framework via latent conditional diffusion models achieves significant computational acceleration while maintaining comparable predictive accuracy to standard diffusion models in physical spaces.
Authors: Lucius Bushnaq, Dan Braun, Lee Sharkey
Abstract: A key step in reverse engineering neural networks is to decompose them into simpler parts that can be studied in relative isolation. Linear parameter decomposition -- a framework that has been proposed to resolve several issues with current decomposition methods -- decomposes neural network parameters into a sum of sparsely used vectors in parameter space. However, the current main method in this framework, Attribution-based Parameter Decomposition (APD), is impractical on account of its computational cost and sensitivity to hyperparameters. In this work, we introduce \textit{Stochastic Parameter Decomposition} (SPD), a method that is more scalable and robust to hyperparameters than APD, which we demonstrate by decomposing models that are slightly larger and more complex than was possible to decompose with APD. We also show that SPD avoids other issues, such as shrinkage of the learned parameters, and better identifies ground truth mechanisms in toy models. By bridging causal mediation analysis and network decomposition methods, this demonstration opens up new research possibilities in mechanistic interpretability by removing barriers to scaling linear parameter decomposition methods to larger models. We release a library for running SPD and reproducing our experiments at https://github.com/goodfire-ai/spd.
Authors: Martin Andrews, Sam Witteveen
Abstract: Optimizing GPU kernels for high performance is a complex task, often demanding deep architectural knowledge, extensive profiling, and iterative experimentation. This challenge is amplified when targeting newer or less-documented GPU architectures where traditional development aids are scarce. This paper introduces an LLM-powered "GPU Kernel Scientist," an automated methodology for iteratively refining accelerator kernels. Our methodology employs LLMs in a multi-stage, evolutionary process: (a) strategically selecting promising prior code versions as a basis for new iterations; (b) generating hypotheses for optimization experiments, based on existing code and assimilated knowledge from general GPU literature; and (c) autonomously implementing these experiments through code modification and subsequent submission to an external evaluation system, using only observed timing data as performance feedback. We detail how this approach navigates the challenges of the AMD MI300 target architecture and leverages LLMs to compensate for limited domain-specific human expertise. Since quantitative results from an ongoing performance competition were embargoed on paper submission date, we present the architectural design, operational workflow, and qualitative insights, highlighting the potential of LLM-driven agents to democratise and accelerate GPU kernel optimization, especially in resource-constrained or rapidly evolving hardware environments.
Authors: Shashwat Khandelwal, Jakoba Petri-Koenig, Thomas B. Preu{\ss}er, Michaela Blott, Shreejith Shanker
Abstract: Recurrent neural networks (RNNs), particularly LSTMs, are effective for time-series tasks like sentiment analysis and short-term stock prediction. However, their computational complexity poses challenges for real-time deployment in resource constrained environments. While FPGAs offer a promising platform for energy-efficient AI acceleration, existing tools mainly target feed-forward networks, and LSTM acceleration typically requires full custom implementation. In this paper, we address this gap by leveraging the open-source and extensible FINN framework to enable the generalized deployment of LSTMs on FPGAs. Specifically, we leverage the Scan operator from the Open Neural Network Exchange (ONNX) specification to model the recurrent nature of LSTM computations, enabling support for mixed quantisation within them and functional verification of LSTM-based models. Furthermore, we introduce custom transformations within the FINN compiler to map the quantised ONNX computation graph to hardware blocks from the HLS kernel library of the FINN compiler and Vitis HLS. We validate the proposed tool-flow by training a quantised ConvLSTM model for a mid-price stock prediction task using the widely used dataset and generating a corresponding hardware IP of the model using our flow, targeting the XCZU7EV device. We show that the generated quantised ConvLSTM accelerator through our flow achieves a balance between performance (latency) and resource consumption, while matching (or bettering) inference accuracy of state-of-the-art models with reduced precision. We believe that the generalisable nature of the proposed flow will pave the way for resource-efficient RNN accelerator designs on FPGAs.
Authors: Jakub Piwko, J\k{e}drzej Ruci\'nski, Dawid P{\l}udowski, Antoni Zajko, Patryzja \.Zak, Mateusz Zacharecki, Anna Kozak, Katarzyna Wo\'znica
Abstract: Ensemble learning has proven effective in boosting predictive performance, but traditional methods such as bagging, boosting, and dynamic ensemble selection (DES) suffer from high computational cost and limited adaptability to heterogeneous data distributions. To address these limitations, we propose Hellsemble, a novel and interpretable ensemble framework for binary classification that leverages dataset complexity during both training and inference. Hellsemble incrementally partitions the dataset into circles of difficulty by iteratively passing misclassified instances from simpler models to subsequent ones, forming a committee of specialised base learners. Each model is trained on increasingly challenging subsets, while a separate router model learns to assign new instances to the most suitable base model based on inferred difficulty. Hellsemble achieves strong classification accuracy while maintaining computational efficiency and interpretability. Experimental results on OpenML-CC18 and Tabzilla benchmarks demonstrate that Hellsemble often outperforms classical ensemble methods. Our findings suggest that embracing instance-level difficulty offers a promising direction for constructing efficient and robust ensemble systems.
Authors: Furkan Mumcu, Yasin Yilmaz
Abstract: Deep Neural Networks (DNNs) are notoriously vulnerable to adversarial input designs with limited noise budgets. While numerous successful attacks with subtle modifications to original input have been proposed, defense techniques against these attacks are relatively understudied. Existing defense approaches either focus on improving DNN robustness by negating the effects of perturbations or use a secondary model to detect adversarial data. Although equally important, the attack detection approach, which is studied in this work, provides a more practical defense compared to the robustness approach. We show that the existing detection methods are either ineffective against the state-of-the-art attack techniques or computationally inefficient for real-time processing. We propose a novel universal and efficient method to detect adversarial examples by analyzing the varying degrees of impact of attacks on different DNN layers. {Our method trains a lightweight regression model that predicts deeper-layer features from early-layer features, and uses the prediction error to detect adversarial samples.} Through theoretical arguments and extensive experiments, we demonstrate that our detection method is highly effective, computationally efficient for real-time processing, compatible with any DNN architecture, and applicable across different domains, such as image, video, and audio.
Authors: Xin Huang, Chul-Ho Lee
Abstract: Graph neural networks (GNNs) are powerful tools for solving graph-related problems. Distributed GNN frameworks and systems enhance the scalability of GNNs and accelerate model training, yet most are optimized for node classification. Their performance on link prediction remains underexplored. This paper demystifies distributed training of GNNs for link prediction by investigating the issue of performance degradation when each worker trains a GNN on its assigned partitioned subgraph without having access to the entire graph. We discover that the main sources of the issue come from not only the information loss caused by graph partitioning but also the ways of drawing negative samples during model training. While sharing the complete graph information with each worker resolves the issue and preserves link prediction accuracy, it incurs a high communication cost. We propose SpLPG, which effectively leverages graph sparsification to mitigate the issue of performance degradation at a reduced communication cost. Experiment results on several public real-world datasets demonstrate the effectiveness of SpLPG, which reduces the communication overhead by up to about 80% while mostly preserving link prediction accuracy.
Authors: Ziyang Lu, M. Cenk Gursoy, Chilukuri K. Mohan, Pramod K. Varshney
Abstract: In this paper, we tackle the task of adaptive time allocation in integrated sensing and communication systems equipped with radar and communication units. The dual-functional radar-communication system's task involves allocating dwell times for tracking multiple targets and utilizing the remaining time for data transmission towards estimated target locations. We introduce a novel constrained deep reinforcement learning (CDRL) approach, designed to optimize resource allocation between tracking and communication under time budget constraints, thereby enhancing target communication quality. Our numerical results demonstrate the efficiency of our proposed CDRL framework, confirming its ability to maximize communication quality in highly dynamic environments while adhering to time constraints.
Authors: Ziyang Lu, Subodh Kalia, M. Cenk Gursoy, Chilukuri K. Mohan, Pramod K. Varshney
Abstract: The time allocation problem in multi-function cognitive radar systems focuses on the trade-off between scanning for newly emerging targets and tracking the previously detected targets. We formulate this as a multi-objective optimization problem and employ deep reinforcement learning to find Pareto-optimal solutions and compare deep deterministic policy gradient (DDPG) and soft actor-critic (SAC) algorithms. Our results demonstrate the effectiveness of both algorithms in adapting to various scenarios, with SAC showing improved stability and sample efficiency compared to DDPG. We further employ the NSGA-II algorithm to estimate an upper bound on the Pareto front of the considered problem. This work contributes to the development of more efficient and adaptive cognitive radar systems capable of balancing multiple competing objectives in dynamic environments.
Authors: Fei Wang, Baochun Li
Abstract: Memorization in large language models (LLMs) makes them vulnerable to data extraction attacks. While pre-training memorization has been extensively studied, fewer works have explored its impact in fine-tuning, particularly for LoRA fine-tuning, a widely adopted parameter-efficient method. In this work, we re-examine memorization in fine-tuning and uncover a surprising divergence from prior findings across different fine-tuning strategies. Factors such as model scale and data duplication, which strongly influence memorization in pre-training and full fine-tuning, do not follow the same trend in LoRA fine-tuning. Using a more relaxed similarity-based memorization metric, we demonstrate that LoRA significantly reduces memorization risks compared to full fine-tuning, while still maintaining strong task performance.
Authors: Zixian Wang, Cole Ramos, Muhammad A. Awad, Keith Lowery
Abstract: In recent years, the rapid advancement of deep neural networks (DNNs) has revolutionized artificial intelligence, enabling models with unprecedented capabilities in understanding, generating, and processing complex data. These powerful architectures have transformed a wide range of downstream applications, tackling tasks beyond human reach. In this paper, we introduce Omniwise, the first end-to-end, self-supervised fine-tuning pipeline that applies large language models (LLMs) to GPU kernel performance prediction--a novel use case in performance profiling. Omniwise is model-agnostic and lightweight, achieving strong results even with a small 3B-parameter model. It can predict key performance metrics, including memory bandwidth, cache hit rates, GFLOPs, and arithmetic intensity, directly from kernel code without the need for code execution or profiling tools. Our approach achieves over 90% of predictions within 10% relative error on GPU kernels executed on AMD MI250 and MI300X architectures. In addition to the pipeline, we develop an online inference server and a Visual Studio Code plugin that seamlessly integrate LLM-based performance prediction into developers' workflows.
Authors: Yian Wang, Ali Ebrahimpour-Boroojeny, Hari Sundaram
Abstract: In this work, we introduce an output-reweighting unlearning method, RWFT, a lightweight technique that erases an entire class from a trained classifier without full retraining. Forgetting specific classes from trained models is essential for enforcing user deletion rights and mitigating harmful or biased predictions. The full retraining is costly and existing unlearning methods fail to replicate the behavior of the retrained models when predicting samples from the unlearned class. We prove this failure by designing a variant of membership inference attacks, MIA-NN that successfully reveals the unlearned class for any of these methods. We propose a simple redistribution of the probability mass for the prediction on the samples in the forgotten class which is robust to MIA-NN. We also introduce a new metric based on the total variation (TV) distance of the prediction probabilities to quantify residual leakage to prevent future methods from susceptibility to the new attack. Through extensive experiments with state of the art baselines in machine unlearning, we show that our approach matches the results of full retraining in both metrics used for evaluation by prior work and the new metric we propose in this work. Compare to state-of-the-art methods, we gain 2.79% in previously used metrics and 111.45% in our new TV-based metric over the best existing method.
Authors: Erfan Hajihashemi, Yanning Shen
Abstract: Online conformal prediction has demonstrated its capability to construct a prediction set for each incoming data point that covers the true label with a predetermined probability. To cope with potential distribution shift, multi-model online conformal prediction has been introduced to select and leverage different models from a preselected candidate set. Along with the improved flexibility, the choice of the preselected set also brings challenges. A candidate set that includes a large number of models may increase the computational complexity. In addition, the inclusion of irrelevant models with poor performance may negatively impact the performance and lead to unnecessarily large prediction sets. To address these challenges, we propose a novel multi-model online conformal prediction algorithm that identifies a subset of effective models at each time step by collecting feedback from a bipartite graph, which is refined upon receiving new data. A model is then selected from this subset to construct the prediction set, resulting in reduced computational complexity and smaller prediction sets. Additionally, we demonstrate that using prediction set size as feedback, alongside model loss, can significantly improve efficiency by constructing smaller prediction sets while still satisfying the required coverage guarantee. The proposed algorithms are proven to ensure valid coverage and achieve sublinear regret. Experiments on real and synthetic datasets validate that the proposed methods construct smaller prediction sets and outperform existing multi-model online conformal prediction approaches.
Authors: Matthew Zurek, Guy Zamir, Yudong Chen
Abstract: We study offline reinforcement learning in average-reward MDPs, which presents increased challenges from the perspectives of distribution shift and non-uniform coverage, and has been relatively underexamined from a theoretical perspective. While previous work obtains performance guarantees under single-policy data coverage assumptions, such guarantees utilize additional complexity measures which are uniform over all policies, such as the uniform mixing time. We develop sharp guarantees depending only on the target policy, specifically the bias span and a novel policy hitting radius, yielding the first fully single-policy sample complexity bound for average-reward offline RL. We are also the first to handle general weakly communicating MDPs, contrasting restrictive structural assumptions made in prior work. To achieve this, we introduce an algorithm based on pessimistic discounted value iteration enhanced by a novel quantile clipping technique, which enables the use of a sharper empirical-span-based penalty function. Our algorithm also does not require any prior parameter knowledge for its implementation. Remarkably, we show via hard examples that learning under our conditions requires coverage assumptions beyond the stationary distribution of the target policy, distinguishing single-policy complexity measures from previously examined cases. We also develop lower bounds nearly matching our main result.
Authors: Ziyang Lu, M. Cenk Gursoy, Chilukuri K. Mohan, Pramod K. Varshney
Abstract: Deep reinforcement learning has been extensively studied in decision-making processes and has demonstrated superior performance over conventional approaches in various fields, including radar resource management (RRM). However, a notable limitation of neural networks is their ``black box" nature and recent research work has increasingly focused on explainable AI (XAI) techniques to describe the rationale behind neural network decisions. One promising XAI method is local interpretable model-agnostic explanations (LIME). However, the sampling process in LIME ignores the correlations between features. In this paper, we propose a modified LIME approach that integrates deep learning (DL) into the sampling process, which we refer to as DL-LIME. We employ DL-LIME within deep reinforcement learning for radar resource management. Numerical results show that DL-LIME outperforms conventional LIME in terms of both fidelity and task performance, demonstrating superior performance with both metrics. DL-LIME also provides insights on which factors are more important in decision making for radar resource management.
Authors: Tong Zeng, Srivathsan Badrinarayanan, Janghoon Ock, Cheng-Kai Lai, Amir Barati Farimani
Abstract: Chemical process optimization is crucial to maximize production efficiency and economic performance. Traditional methods, including gradient-based solvers, evolutionary algorithms, and parameter grid searches, become impractical when operating constraints are ill-defined or unavailable, requiring engineers to rely on subjective heuristics to estimate feasible parameter ranges. To address this constraint definition bottleneck, we present a multi-agent framework of large language model (LLM) agents that autonomously infer operating constraints from minimal process descriptions, then collaboratively guide optimization using the inferred constraints. Our AutoGen-based agentic framework employs OpenAI's o3 model, with specialized agents for constraint generation, parameter validation, simulation execution, and optimization guidance. Through two phases - autonomous constraint generation using embedded domain knowledge, followed by iterative multi-agent optimization - the framework eliminates the need for predefined operational bounds. Validated on the hydrodealkylation process across cost, yield, and yield-to-cost ratio metrics, the framework demonstrated competitive performance with conventional optimization methods while achieving better computational efficiency, requiring fewer iterations to converge. Our approach converged in under 20 minutes, achieving a 31-fold speedup over grid search. Beyond computational efficiency, the framework's reasoning-guided search demonstrates sophisticated process understanding, correctly identifying utility trade-offs, and applying domain-informed heuristics. This approach shows significant potential for optimization scenarios where operational constraints are poorly characterized or unavailable, particularly for emerging processes and retrofit applications.
Authors: Shahrzad Behzadimanesh, Pierre Le Bodic, Geoffrey I. Webb, Mario Boley
Abstract: Small additive ensembles of symbolic rules offer interpretable prediction models. Traditionally, these ensembles use rule conditions based on conjunctions of simple threshold propositions $x \geq t$ on a single input variable $x$ and threshold $t$, resulting geometrically in axis-parallel polytopes as decision regions. While this form ensures a high degree of interpretability for individual rules and can be learned efficiently using the gradient boosting approach, it relies on having access to a curated set of expressive and ideally independent input features so that a small ensemble of axis-parallel regions can describe the target variable well. Absent such features, reaching sufficient accuracy requires increasing the number and complexity of individual rules, which diminishes the interpretability of the model. Here, we extend classical rule ensembles by introducing logical propositions with learnable sparse linear transformations of input variables, i.e., propositions of the form $\mathbf{x}^\mathrm{T}\mathbf{w} \geq t$, where $\mathbf{w}$ is a learnable sparse weight vector, enabling decision regions as general polytopes with oblique faces. We propose a learning method using sequential greedy optimization based on an iteratively reweighted formulation of logistic regression. Experimental results demonstrate that the proposed method efficiently constructs rule ensembles with the same test risk as state-of-the-art methods while significantly reducing model complexity across ten benchmark datasets.
Authors: Keivan Rezaei, Mehrdad Saberi, Abhilasha Ravichander, Soheil Feizi
Abstract: Large language models are trained on massive corpora of web data, which may include private data, copyrighted material, factually inaccurate data, or data that degrades model performance. Eliminating the influence of such problematic datapoints through complete retraining -- by repeatedly pretraining the model on datasets that exclude these specific instances -- is computationally prohibitive. For this reason, unlearning algorithms have emerged that aim to eliminate the influence of particular datapoints, while otherwise preserving the model -- at a low computational cost. However, precisely estimating and undoing the influence of individual datapoints has proved to be challenging. In this work, we propose a new algorithm, MSA, for estimating and undoing the influence of datapoints -- by leveraging model checkpoints i.e. artifacts capturing model states at different stages of pretraining. Our experimental results demonstrate that MSA consistently outperforms existing machine unlearning algorithms across multiple benchmarks, models, and evaluation metrics, suggesting that MSA could be an effective approach towards more flexible large language models that are capable of data erasure.
Authors: Jiameng Chen, Xiantao Cai, Jia Wu, Wenbin Hu
Abstract: Antibody design remains a critical challenge in therapeutic and diagnostic development, particularly for complex antigens with diverse binding interfaces. Current computational methods face two main limitations: (1) capturing geometric features while preserving symmetries, and (2) generalizing novel antigen interfaces. Despite recent advancements, these methods often fail to accurately capture molecular interactions and maintain structural integrity. To address these challenges, we propose \textbf{AbMEGD}, an end-to-end framework integrating \textbf{M}ulti-scale \textbf{E}quivariant \textbf{G}raph \textbf{D}iffusion for antibody sequence and structure co-design. Leveraging advanced geometric deep learning, AbMEGD combines atomic-level geometric features with residue-level embeddings, capturing local atomic details and global sequence-structure interactions. Its E(3)-equivariant diffusion method ensures geometric precision, computational efficiency, and robust generalizability for complex antigens. Furthermore, experiments using the SAbDab database demonstrate a 10.13\% increase in amino acid recovery, 3.32\% rise in improvement percentage, and a 0.062~\AA\ reduction in root mean square deviation within the critical CDR-H3 region compared to DiffAb, a leading antibody design model. These results highlight AbMEGD's ability to balance structural integrity with improved functionality, establishing a new benchmark for sequence-structure co-design and affinity optimization. The code is available at: https://github.com/Patrick221215/AbMEGD.
Authors: Yifan Yang, Zhen Zhang, Rupak Vignesh Swaminathan, Jing Liu, Nathan Susanj, Zheng Zhang
Abstract: Fine-tuning vision language models (VLMs) has achieved remarkable performance across various downstream tasks; yet, it requires access to model gradients through backpropagation (BP), making them unsuitable for memory-constrained, inference-only edge devices. To address this limitation, previous work has explored various BP-free fine-tuning methods. However, these approaches often rely on high-variance evolutionary strategies (ES) or zeroth-order (ZO) optimization, and often fail to achieve satisfactory performance. In this paper, we propose a hybrid Sharpness-aware Zeroth-order optimization (SharpZO) approach, specifically designed to enhance the performance of ZO VLM fine-tuning via a sharpness-aware warm-up training. SharpZO features a two-stage optimization process: a sharpness-aware ES stage that globally explores and smooths the loss landscape to construct a strong initialization, followed by a fine-grained local search via sparse ZO optimization. The entire optimization relies solely on forward passes. Detailed theoretical analysis and extensive experiments on CLIP models demonstrate that SharpZO significantly improves accuracy and convergence speed, achieving up to 7% average gain over state-of-the-art forward-only methods.
Authors: Steven Walton, Valeriy Klyukin, Maksim Artemev, Denis Derkach, Nikita Orlov, Humphrey Shi
Abstract: Explicit density learners are becoming an increasingly popular technique for generative models because of their ability to better model probability distributions. They have advantages over Generative Adversarial Networks due to their ability to perform density estimation and having exact latent-variable inference. This has many advantages, including: being able to simply interpolate, calculate sample likelihood, and analyze the probability distribution. The downside of these models is that they are often more difficult to train and have lower sampling quality. Normalizing flows are explicit density models, that use composable bijective functions to turn an intractable probability function into a tractable one. In this work, we present novel knowledge distillation techniques to increase sampling quality and density estimation of smaller student normalizing flows. We seek to study the capacity of knowledge distillation in Compositional Normalizing Flows to understand the benefits and weaknesses provided by these architectures. Normalizing flows have unique properties that allow for a non-traditional forms of knowledge transfer, where we can transfer that knowledge within intermediate layers. We find that through this distillation, we can make students significantly smaller while making substantial performance gains over a non-distilled student. With smaller models there is a proportionally increased throughput as this is dependent upon the number of bijectors, and thus parameters, in the network.
Authors: Feng Jiang, Mangal Prakash, Hehuan Ma, Jianyuan Deng, Yuzhi Guo, Amina Mollaysa, Tommaso Mansi, Rui Liao, Junzhou Huang
Abstract: Molecular property prediction aims to learn representations that map chemical structures to functional properties. While multimodal learning has emerged as a powerful paradigm to learn molecular representations, prior works have largely overlooked textual and taxonomic information of molecules for representation learning. We introduce TRIDENT, a novel framework that integrates molecular SMILES, textual descriptions, and taxonomic functional annotations to learn rich molecular representations. To achieve this, we curate a comprehensive dataset of molecule-text pairs with structured, multi-level functional annotations. Instead of relying on conventional contrastive loss, TRIDENT employs a volume-based alignment objective to jointly align tri-modal features at the global level, enabling soft, geometry-aware alignment across modalities. Additionally, TRIDENT introduces a novel local alignment objective that captures detailed relationships between molecular substructures and their corresponding sub-textual descriptions. A momentum-based mechanism dynamically balances global and local alignment, enabling the model to learn both broad functional semantics and fine-grained structure-function mappings. TRIDENT achieves state-of-the-art performance on 11 downstream tasks, demonstrating the value of combining SMILES, textual, and taxonomic functional annotations for molecular property prediction.
Authors: Haodong Lu, Chongyang Zhao, Jason Xue, Lina Yao, Kristen Moore, Dong Gong
Abstract: Continual learning (CL) with large pre-trained models is challenged by catastrophic forgetting and task interference. Existing LoRA-based Mixture-of-Experts (MoE) approaches mitigate forgetting by assigning and freezing task-specific adapters, but suffer from interference, redundancy, and ambiguous routing due to coarse adapter-level selection. However, this design introduces three key challenges: 1) Interference: Activating full LoRA experts per input leads to subspace interference and prevents selective reuse of useful components across tasks. 2) Redundancy: Newly added experts often duplicate or contradict existing knowledge due to unnecessary activation of unrelated ranks and insufficient reuse of relevant ones. 3) Ambiguity: Overlapping features across tasks confuse the router, resulting in unstable expert assignments. As more experts accumulate, earlier task routing degrades, accelerating forgetting. We propose MoRA, a Mixture-of-Rank Adaptive learning approach with self-activated and sparse rank activation for CL. Unlike mixing multiple low-rank matrices, MoRA decomposes each rank-r update into r rank-1 components, each treated as an independent expert, enabling fine-grained mixture of rank-1 expert utilization while mitigating interference and redundancy. To avoid ambiguous routing, we propose that each rank-1 expert can infer its own relevance via intermediate activations. Coupled with our proposed rank pruning and activation budgets, MoRA adaptively selects a sparse mixture of ranks per input. We validate MoRA on continual learning tasks with CLIP and large language models (LLMs), analyzing both in-domain learning and out-of-domain forgetting/generalization during fine-tuning. MoRA shows significant effectiveness on enhancing CL with PTMs, and improving generalization while mitigating forgetting.
Authors: Fu Peng, Meng Zhang, Ming Tang
Abstract: Recent studies in federated learning (FL) commonly train models on static datasets. However, real-world data often arrives as streams with shifting distributions, causing performance degradation known as concept drift. This paper analyzes FL performance under concept drift using information theory and proposes an algorithm to mitigate the performance degradation. We model concept drift as a Markov chain and introduce the \emph{Stationary Generalization Error} to assess a model's capability to capture characteristics of future unseen data. Its upper bound is derived using KL divergence and mutual information. We study three drift patterns (periodic, gradual, and random) and their impact on FL performance. Inspired by this, we propose an algorithm that regularizes the empirical risk minimization approach with KL divergence and mutual information, thereby enhancing long-term performance. We also explore the performance-cost tradeoff by identifying a Pareto front. To validate our approach, we build an FL testbed using Raspberry Pi4 devices. Experimental results corroborate with theoretical findings, confirming that drift patterns significantly affect performance. Our method consistently outperforms existing approaches for these three patterns, demonstrating its effectiveness in adapting concept drift in FL.
Authors: Suorong Yang, Peijia Li, Furao Shen, Jian Zhao
Abstract: Modern deep architectures often rely on large-scale datasets, but training on these datasets incurs high computational and storage overhead. Real-world datasets often contain substantial redundancies, prompting the need for more data-efficient training paradigms. Data selection has shown promise to mitigate redundancy by identifying the most representative samples, thereby reducing training costs without compromising performance. Existing methods typically rely on static scoring metrics or pretrained models, overlooking the combined effect of selected samples and their evolving dynamics during training. We introduce the concept of epsilon-sample cover, which quantifies sample redundancy based on inter-sample relationships, capturing the intrinsic structure of the dataset. Based on this, we reformulate data selection as a reinforcement learning (RL) process and propose RL-Selector, where a lightweight RL agent optimizes the selection policy by leveraging epsilon-sample cover derived from evolving dataset distribution as a reward signal. Extensive experiments across benchmark datasets and diverse architectures demonstrate that our method consistently outperforms existing state-of-the-art baselines. Models trained with our selected datasets show enhanced generalization performance with improved training efficiency.
Authors: Jaebak Hwang, Sanghyeon Lee, Jeongmo Kim, Seungyul Han
Abstract: Long-horizon goal-conditioned tasks pose fundamental challenges for reinforcement learning (RL), particularly when goals are distant and rewards are sparse. While hierarchical and graph-based methods offer partial solutions, they often suffer from subgoal infeasibility and inefficient planning. We introduce Strict Subgoal Execution (SSE), a graph-based hierarchical RL framework that enforces single-step subgoal reachability by structurally constraining high-level decision-making. To enhance exploration, SSE employs a decoupled exploration policy that systematically traverses underexplored regions of the goal space. Furthermore, a failure-aware path refinement, which refines graph-based planning by dynamically adjusting edge costs according to observed low-level success rates, thereby improving subgoal reliability. Experimental results across diverse long-horizon benchmarks demonstrate that SSE consistently outperforms existing goal-conditioned RL and hierarchical RL approaches in both efficiency and success rate.
Authors: He Zhang, Ming Zhou, Shaopeng Zhai, Ying Sun, Hui Xiong
Abstract: Unsupervised skill discovery aims to learn diverse and distinguishable behaviors in open-ended reinforcement learning. For existing methods, they focus on improving diversity through pure exploration, mutual information optimization, and learning temporal representation. Despite that they perform well on exploration, they remain limited in terms of efficiency, especially for the high-dimensional situations. In this work, we frame skill discovery as a min-max game of skill generation and policy learning, proposing a regret-aware method on top of temporal representation learning that expands the discovered skill space along the direction of upgradable policy strength. The key insight behind the proposed method is that the skill discovery is adversarial to the policy learning, i.e., skills with weak strength should be further explored while less exploration for the skills with converged strength. As an implementation, we score the degree of strength convergence with regret, and guide the skill discovery with a learnable skill generator. To avoid degeneration, skill generation comes from an up-gradable population of skill generators. We conduct experiments on environments with varying complexities and dimension sizes. Empirical results show that our method outperforms baselines in both efficiency and diversity. Moreover, our method achieves a 15% zero shot improvement in high-dimensional environments, compared to existing methods.
Authors: Fu Peng, Ming Tang
Abstract: In federated learning (FL), the data distribution of each client may change over time, introducing both temporal and spatial data heterogeneity, known as concept drift. Data heterogeneity arises from three drift sources: real drift (a shift in the conditional distribution P(y|x)), virtual drift (a shift in the input distribution P(x)), and label drift (a shift in the label distribution P(y)). However, most existing FL methods addressing concept drift primarily focus on real drift. When clients experience virtual or label drift, these methods often fail to selectively retain useful historical knowledge, leading to catastrophic forgetting. A key challenge lies in distinguishing different sources of drift, as they require distinct adaptation strategies: real drift calls for discarding outdated data, while virtual or label drift benefits from retaining historical data. Without explicitly identifying the drift sources, a general adaptation strategy is suboptimal and may harm generalization. To address this challenge, we propose FedDAA, a dynamic clustered FL framework designed to adapt to multi-source concept drift while preserving valuable historical knowledge. Specifically, FedDAA integrates three modules: a cluster number determination module to find the optimal number of clusters; a real drift detection module to distinguish real drift from virtual/label drift; and a concept drift adaptation module to adapt to new data while retaining useful historical information. We provide theoretical convergence guarantees, and experiments show that FedDAA achieves 7.84% to 8.52% accuracy improvements over state-of-the-art methods on Fashion-MNIST, CIFAR-10, and CIFAR-100.
Authors: Jingwei Wang, Zai Zhang, Hao Qian, Chunjing Gan, Binbin Hu, Ziqi Liu, Zhiqiang Zhang, Jun Zhou, Bin Shi, Bo Dong
Abstract: Teaching large language models (LLMs) to use tools is crucial for improving their problem-solving abilities and expanding their applications. However, effectively using tools is challenging because it requires a deep understanding of tool functionalities and user intentions. Previous methods relied mainly on LLMs to generate instruction data, but the quality of these data was often insufficient. In this paper, we propose a new method that uses knowledge graphs to generate high-quality instruction data for LLMs. Knowledge graphs are manually curated datasets rich in semantic information. We begin by extracting various query pathways from a given knowledge graph, which are transformed into a broad spectrum of user queries. We then translate the relationships between entities into actionable tools and parse the pathways of each query into detailed solution steps, thereby creating high-quality instruction data. Our experiments show that fine-tuning on just a small sample of this synthetic data can significantly improve the tool utilization and overall capabilities of LLMs.
Authors: Li Fan, Peng Wang, Jing Yang, Cong Shen
Abstract: Transformers have shown potential in solving wireless communication problems, particularly via in-context learning (ICL), where models adapt to new tasks through prompts without requiring model updates. However, prior ICL-based Transformer models rely on deep architectures with many layers to achieve satisfactory performance, resulting in substantial storage and computational costs. In this work, we propose CHain Of thOught Symbol dEtection (CHOOSE), a CoT-enhanced shallow Transformer framework for wireless symbol detection. By introducing autoregressive latent reasoning steps within the hidden space, CHOOSE significantly improves the reasoning capacity of shallow models (1-2 layers) without increasing model depth. This design enables lightweight Transformers to achieve detection performance comparable to much deeper models, making them well-suited for deployment on resource-constrained mobile devices. Experimental results demonstrate that our approach outperforms conventional shallow Transformers and achieves performance comparable to that of deep Transformers, while maintaining storage and computational efficiency. This represents a promising direction for implementing Transformer-based algorithms in wireless receivers with limited computational resources.
Authors: Xenia Heilmann, Luca Corbucci, Mattia Cerrato, Anna Monreale
Abstract: Federated Learning (FL) enables collaborative model training across multiple clients without sharing clients' private data. However, fairness remains a key concern, as biases in local clients' datasets can impact the entire federated system. Heterogeneous data distributions across clients may lead to models that are fairer for some clients than others. Although several fairness-enhancing solutions are present in the literature, most focus on mitigating bias for a single sensitive attribute, typically binary, overlooking the diverse and sometimes conflicting fairness needs of different clients. This limited perspective can limit the effectiveness of fairness interventions for the different clients. To support more robust and reproducible fairness research in FL, we aim to enable a consistent benchmarking of fairness-aware FL methods at both the global and client levels. In this paper, we contribute in three ways: (1) We introduce FeDa4Fair, a library to generate tabular datasets tailored to evaluating fair FL methods under heterogeneous client bias; (2) we release four bias-heterogeneous datasets and corresponding benchmarks to compare fairness mitigation methods in a controlled environment; (3) we provide ready-to-use functions for evaluating fairness outcomes for these datasets.
Authors: David Debot, Pietro Barbiero, Gabriele Dominici, Giuseppe Marra
Abstract: Concept-Based Models (CBMs) are a class of deep learning models that provide interpretability by explaining predictions through high-level concepts. These models first predict concepts and then use them to perform a downstream task. However, current CBMs offer interpretability only for the final task prediction, while the concept predictions themselves are typically made via black-box neural networks. To address this limitation, we propose Hierarchical Concept Memory Reasoner (H-CMR), a new CBM that provides interpretability for both concept and task predictions. H-CMR models relationships between concepts using a learned directed acyclic graph, where edges represent logic rules that define concepts in terms of other concepts. During inference, H-CMR employs a neural attention mechanism to select a subset of these rules, which are then applied hierarchically to predict all concepts and the final task. Experimental results demonstrate that H-CMR matches state-of-the-art performance while enabling strong human interaction through concept and model interventions. The former can significantly improve accuracy at inference time, while the latter can enhance data efficiency during training when background knowledge is available.
Authors: Tim Lawson, Laurence Aitchison
Abstract: Conditional computation is a popular strategy to make Transformers more efficient. Existing methods often target individual modules (e.g., mixture-of-experts layers) or skip layers independently of one another. However, interpretability research has demonstrated that the middle layers of Transformers exhibit greater redundancy, and that early layers aggregate information into token positions. Guided by these insights, we propose a novel architecture that dynamically skips a variable number of layers from the middle outward. In particular, a learned gating mechanism determines whether to bypass a symmetric span of central blocks based on the input, and a gated attention mechanism prevents subsequent tokens from attending to skipped token positions. Residual norms are controlled with a 'sandwich' or 'perilayernorm' scheme and gate sparsity with an adaptive regularization loss. We had aimed to reduce compute requirements for 'simpler' tokens and potentially foster an emergent multi-level representational hierarchy but, at the scales investigated, our approach does not achieve improvements in the trade-off between validation cross-entropy and estimated FLOPs compared to dense baselines with fewer layers. We release our code at https://github.com/tim-lawson/skip-middle.
Authors: Changxi Chi, Jun Xia, Yufei Huang, Jingbo Zhou, Siyuan Li, Yunfan Liu, Chang Yu, Stan Z. Li
Abstract: Estimating single-cell responses across various perturbations facilitates the identification of key genes and enhances drug screening, significantly boosting experimental efficiency. However, single-cell sequencing is a destructive process, making it impossible to capture the same cell's phenotype before and after perturbation. Consequently, data collected under perturbed and unperturbed conditions are inherently unpaired. Existing methods either attempt to forcibly pair unpaired data using random sampling, or neglect the inherent relationship between unperturbed and perturbed cells during the modeling. In this work, we propose a framework based on Dual Diffusion Implicit Bridges (DDIB) to learn the mapping between different data distributions, effectively addressing the challenge of unpaired data. We further interpret this framework as a form of data augmentation. We integrate gene regulatory network (GRN) information to propagate perturbation signals in a biologically meaningful way, and further incorporate a masking mechanism to predict silent genes, improving the quality of generated profiles. Moreover, gene expression under the same perturbation often varies significantly across cells, frequently exhibiting a bimodal distribution that reflects intrinsic heterogeneity. To capture this, we introduce a more suitable evaluation metric. We propose Unlasting, dual conditional diffusion models that overcome the problem of unpaired single-cell perturbation data and strengthen the model's insight into perturbations under the guidance of the GRN, with a dedicated mask model designed to improve generation quality by predicting silent genes. In addition, we introduce a biologically grounded evaluation metric that better reflects the inherent heterogeneity in single-cell responses.
Authors: Deepak Kumar Panda, Weisi Guo
Abstract: The increasing automation of navigation for unmanned aerial vehicles (UAVs) has exposed them to adversarial attacks that exploit vulnerabilities in reinforcement learning (RL) through sensor manipulation. Although existing robust RL methods aim to mitigate such threats, their effectiveness has limited generalization to out-of-distribution shifts from the optimal value distribution, as they are primarily designed to handle fixed perturbation. To address this limitation, this paper introduces an antifragile RL framework that enhances adaptability to broader distributional shifts by incorporating a switching mechanism based on discounted Thompson sampling (DTS). This mechanism dynamically selects among multiple robust policies to minimize adversarially induced state-action-value distribution shifts. The proposed approach first derives a diverse ensemble of action robust policies by accounting for a range of perturbations in the policy space. These policies are then modeled as a multiarmed bandit (MAB) problem, where DTS optimally selects policies in response to nonstationary Bernoulli rewards, effectively adapting to evolving adversarial strategies. Theoretical framework has also been provided where by optimizing the DTS to minimize the overall regrets due to distributional shift, results in effective adaptation against unseen adversarial attacks thus inducing antifragility. Extensive numerical simulations validate the effectiveness of the proposed framework in complex navigation environments with multiple dynamic three-dimensional obstacles and with stronger projected gradient descent (PGD) and spoofing attacks. Compared to conventional robust, non-adaptive RL methods, the antifragile approach achieves superior performance, demonstrating shorter navigation path lengths and a higher rate of conflict-free navigation trajectories compared to existing robust RL techniques
Authors: Deepak Kumar Panda, Adolfo Perrusquia, Weisi Guo
Abstract: Reinforcement learning (RL) policies deployed in safety-critical systems, such as unmanned aerial vehicle (UAV) navigation in dynamic airspace, are vulnerable to out-ofdistribution (OOD) adversarial attacks in the observation space. These attacks induce distributional shifts that significantly degrade value estimation, leading to unsafe or suboptimal decision making rendering the existing policy fragile. To address this vulnerability, we propose an antifragile RL framework designed to adapt against curriculum of incremental adversarial perturbations. The framework introduces a simulated attacker which incrementally increases the strength of observation-space perturbations which enables the RL agent to adapt and generalize across a wider range of OOD observations and anticipate previously unseen attacks. We begin with a theoretical characterization of fragility, formally defining catastrophic forgetting as a monotonic divergence in value function distributions with increasing perturbation strength. Building on this, we define antifragility as the boundedness of such value shifts and derive adaptation conditions under which forgetting is stabilized. Our method enforces these bounds through iterative expert-guided critic alignment using Wasserstein distance minimization across incrementally perturbed observations. We empirically evaluate the approach in a UAV deconfliction scenario involving dynamic 3D obstacles. Results show that the antifragile policy consistently outperforms standard and robust RL baselines when subjected to both projected gradient descent (PGD) and GPS spoofing attacks, achieving up to 15% higher cumulative reward and over 30% fewer conflict events. These findings demonstrate the practical and theoretical viability of antifragile reinforcement learning for secure and resilient decision-making in environments with evolving threat scenarios.
Authors: Weikang Meng, Yadan Luo, Liangyu Huo, Yaowei Wang, Xin Li, Zheng Zhang
Abstract: Linear attention has emerged as a viable alternative to softmax attention by reducing complexity from quadratic to linear in sequence length. To preserve two fundamental properties of softmax, non-negativity and entropy reduction, current works employ various linearly separatable kernel functions with $L1$ normalization instead of softmax operator. However, query norms are neglected by the normalization operation in linear attention, such degradation heavily leads to an entropy gap. Meanwhile, existing works inhibit negative values of query and key vectors resulting in a missing inner-product interactions after being mapped. To address these dual challenges, we propose a novel Norm-Aware Linear Attention mechanism serving to restore norm-guided dynamic spikiness and recover kernel-perturbed norm distributions. Specifically, we first decouple query and key matrices into two components: norm and direction, to achieve norm-aware spikiness control and norm consistency, respectively. We mathematically reveal that the extent of entropy reduction varies with the query norm in softmax normalization, motivating a query-norm aware kernel function for dynamic control over entropy reduction. Furthermore, to ensure norm consistency and enforce non-negativity constraints, we employ a norm-preserving mapping to project all elements of the angular matrix into positive values, leveraging cosine similarity to inhibit dimensions with opposite directions. We conduct extensive experiments demonstrating that the NaLaFormer improves performance on vision and language tasks, enhancing both expressiveness and efficiency by up to 4.2\%.
Authors: Ziwei Wang, Hongbin Wang, Tianwang Jia, Xingyi He, Siyang Li, Dongrui Wu
Abstract: Electroencephalography (EEG)-based brain-computer interfaces (BCIs) transform spontaneous/evoked neural activity into control commands for external communication. While convolutional neural networks (CNNs) remain the mainstream backbone for EEG decoding, their inherently short receptive field makes it difficult to capture long-range temporal dependencies and global inter-channel relationships. Recent CNN-Transformer (Conformers) hybrids partially address this issue, but most adopt a serial design, resulting in suboptimal integration of local and global features, and often overlook explicit channel-wise modeling. To address these limitations, we propose DBConformer, a dual-branch convolutional Transformer network tailored for EEG decoding. It integrates a temporal Conformer to model long-range temporal dependencies and a spatial Conformer to extract inter-channel interactions, capturing both temporal dynamics and spatial patterns in EEG signals. A lightweight channel attention module further refines spatial representations by assigning data-driven importance to EEG channels. Extensive experiments on five motor imagery (MI) datasets and two seizure detection datasets under three evaluation settings demonstrate that DBConformer consistently outperforms 10 competitive baseline models, with over eight times fewer parameters than the high-capacity EEG Conformer baseline. Further, the visualization results confirm that the features extracted by DBConformer are physiologically interpretable and aligned with sensorimotor priors in MI. The superior performance and interpretability of DBConformer make it reliable for robust and explainable EEG decoding. Code is publicized at https://github.com/wzwvv/DBConformer.
Authors: Deepak Kumar Panda, Weisi Guo
Abstract: The growing integration of UAVs into civilian airspace underscores the need for resilient and intelligent intrusion detection systems (IDS), as traditional anomaly detection methods often fail to identify novel threats. A common approach treats unfamiliar attacks as out-of-distribution (OOD) samples; however, this leaves systems vulnerable when mitigation is inadequate. Moreover, conventional OOD detectors struggle to distinguish stealthy adversarial attacks from genuine OOD events. This paper introduces a conditional generative adversarial network (cGAN)-based framework for crafting stealthy adversarial attacks that evade IDS mechanisms. We first design a robust multi-class IDS classifier trained on benign UAV telemetry and known cyber-attacks, including Denial of Service (DoS), false data injection (FDI), man-in-the-middle (MiTM), and replay attacks. Using this classifier, our cGAN perturbs known attacks to generate adversarial samples that misclassify as benign while retaining statistical resemblance to OOD distributions. These adversarial samples are iteratively refined to achieve high stealth and success rates. To detect such perturbations, we implement a conditional variational autoencoder (CVAE), leveraging negative log-likelihood to separate adversarial inputs from authentic OOD samples. Comparative evaluation shows that CVAE-based regret scores significantly outperform traditional Mahalanobis distance-based detectors in identifying stealthy adversarial threats. Our findings emphasize the importance of advanced probabilistic modeling to strengthen IDS capabilities against adaptive, generative-model-based cyber intrusions.
Authors: Yuguang Zhang, Kuangpu Guo, Zhihe Lu, Yunbo Wang, Jian Liang
Abstract: Federated learning (FL) enables collaborative model training across decentralized clients without sharing local data, but is challenged by heterogeneity in data, computation, and communication. Pretrained vision-language models (VLMs), with their strong generalization and lightweight tuning via prompts, offer a promising solution. However, existing federated prompt-learning methods rely only on text prompts and overlook joint label-domain distribution shifts. In this paper, we propose a personalized FL framework based on dual-prompt learning and cross fusion, termed pFedDC. Specifically, each client maintains both global and local prompts across vision and language modalities: global prompts capture common knowledge shared across the federation, while local prompts encode client-specific semantics and domain characteristics. Meanwhile, a cross-fusion module is designed to adaptively integrate prompts from different levels, enabling the model to generate personalized representations aligned with each client's unique data distribution. Extensive experiments across nine datasets with various types of heterogeneity show that pFedDC consistently outperforms state-of-the-art methods.
Authors: Silas Dobler, Florian Lemmerich
Abstract: In neural network compression, most current methods reduce unnecessary parameters by measuring importance and redundancy. To augment already highly optimized existing solutions, we propose linearity-based compression as a novel way to reduce weights in a neural network. It is based on the intuition that with ReLU-like activation functions, neurons that are almost always activated behave linearly, allowing for merging of subsequent layers. We introduce the theory underlying this compression and evaluate our approach experimentally. Our novel method achieves a lossless compression down to 1/4 of the original model size in over the majority of tested models. Applying our method on already importance-based pruned models shows very little interference between different types of compression, demonstrating the option of successful combination of techniques. Overall, our work lays the foundation for a new type of compression method that enables smaller and ultimately more efficient neural network models.
Authors: Hampus Gummesson Svensson, Ola Engkvist, Jon Paul Janet, Christian Tyrchan, Morteza Haghir Chehreghani
Abstract: In many real-world applications, evaluating the goodness of instances is often costly and time-consuming, e.g., human feedback and physics simulations, in contrast to proposing new instances. In particular, this is even more critical in reinforcement learning, as new interactions with the environment (i.e., new instances) need to be evaluated to provide a reward signal to learn from. As sufficient exploration is crucial, learning from a diverse mini-batch can have a large impact and help mitigate mode collapse. In this paper, we introduce diverse mini-batch selection for reinforcement learning and propose to use determinantal point processes for this task. We study this framework in the context of a real-world problem, namely drug discovery. We experimentally study how our proposed framework can improve the effectiveness of chemical exploration in de novo drug design, where finding diverse and high-quality solutions is essential. We conduct a comprehensive evaluation with three well-established molecular generation oracles over numerous generative steps. Our experiments conclude that our diverse mini-batch selection framework can substantially improve the diversity of the solutions, while still obtaining solutions of high quality. In drug discovery, such outcome can potentially lead to fulfilling unmet medication needs faster.
Authors: Apurva Shah, Axel Abels, Ann Now\'e, Tom Lenaerts
Abstract: Perpetual voting addresses fairness in sequential collective decision-making by evaluating representational equity over time. However, existing perpetual voting rules rely on full participation and complete approval information, assumptions that rarely hold in practice, where partial turnout is the norm. In this work, we study the integration of Artificial Delegates, preference-learning agents trained to represent absent voters, into perpetual voting systems. We examine how absenteeism affects fairness and representativeness under various voting methods and evaluate the extent to which Artificial Delegates can compensate for missing participation. Our findings indicate that while absenteeism significantly affects fairness, Artificial Delegates reliably mitigate these effects and enhance robustness across diverse scenarios.
Authors: Andrey Goncharov, Daniil Vyazhev, Petr Sychev, Edvard Khalafyan, Alexey Zaytsev
Abstract: General-purpose Large Language Models (LLMs) are frequently fine-tuned through supervised fine-tuning (SFT) to enhance performance in specific domains. Better results can be achieved by distilling the chain-of-thought of a larger model at the cost of numerous expensive calls and a much greater amount of data. We propose a novel blueprint for efficient fine-tuning that uses reasoning only for complex data identified by entropy. Specifically, across two small open models ($\approx 3B$) we split the training data into complexity categories by a single token answer entropy (ROC AUC $0.73$), fine-tune large language models (LLMs) via SFT and distillation, and show that our pipeline significantly outperforms the standard SFT approach ($0.55$ vs $0.43$ average accuracy) and provides comparable with distillation performance while using $62\%$ less data ($0.55$ average accuracy for both). We publish our code and data to facilitate further research in this direction.
Authors: Elie Saad, Aya Mrabah, Mariem Besbes, Marc Zolghadri, Victor Czmil, Claude Baron, Vincent Bourgeois
Abstract: Component obsolescence poses significant challenges in industries reliant on electronic components, causing increased costs and disruptions in the security and availability of systems. Accurate obsolescence risk prediction is essential but hindered by a lack of reliable data. This paper proposes a novel approach to forecasting obsolescence risk using zero-shot learning (ZSL) with large language models (LLMs) to address data limitations by leveraging domain-specific knowledge from tabular datasets. Applied to two real-world datasets, the method demonstrates effective risk prediction. A comparative evaluation of four LLMs underscores the importance of selecting the right model for specific forecasting tasks.
Authors: Ji Qi, WenPeng Zhu, Li Li, Ming Wu, YingJun Wu, Wu He, Xun Gao, Jason Zeng, Michael Heinrich
Abstract: The distributed training of foundation models, particularly large language models (LLMs), demands a high level of communication. Consequently, it is highly dependent on a centralized cluster with fast and reliable interconnects. Can we conduct training on slow networks and thereby unleash the power of decentralized clusters when dealing with models exceeding 100 billion parameters? In this paper, we propose DiLoCoX, a low-communication large-scale decentralized cluster training framework. It combines Pipeline Parallelism with Dual Optimizer Policy, One-Step-Delay Overlap of Communication and Local Training, and an Adaptive Gradient Compression Scheme. This combination significantly improves the scale of parameters and the speed of model pre-training. We justify the benefits of one-step-delay overlap of communication and local training, as well as the adaptive gradient compression scheme, through a theoretical analysis of convergence. Empirically, we demonstrate that DiLoCoX is capable of pre-training a 107B foundation model over a 1Gbps network. Compared to vanilla AllReduce, DiLoCoX can achieve a 357x speedup in distributed training while maintaining negligible degradation in model convergence. To the best of our knowledge, this is the first decentralized training framework successfully applied to models with over 100 billion parameters.
Authors: Guillaume Carri\`ere, Fr\'ed\'eric Cazals
Abstract: We revisit the randomized seeding techniques for k-means clustering and k-GMM (Gaussian Mixture model fitting with Expectation-Maximization), formalizing their three key ingredients: the metric used for seed sampling, the number of candidate seeds, and the metric used for seed selection. This analysis yields novel families of initialization methods exploiting a lookahead principle--conditioning the seed selection to an enhanced coherence with the final metric used to assess the algorithm, and a multipass strategy to tame down the effect of randomization. Experiments show a consistent constant factor improvement over classical contenders in terms of the final metric (SSE for k-means, log-likelihood for k-GMM), at a modest overhead. In particular, for k-means, our methods improve on the recently designed multi-swap strategy, which was the first one to outperform the greedy k-means++ seeding. Our experimental analysis also shed light on subtle properties of k-means often overlooked, including the (lack of) correlations between the SSE upon seeding and the final SSE, the variance reduction phenomena observed in iterative seeding methods, and the sensitivity of the final SSE to the pool size for greedy methods. Practically, our most effective seeding methods are strong candidates to become one of the--if not the--standard techniques. From a theoretical perspective, our formalization of seeding opens the door to a new line of analytical approaches.
Authors: Jiajie Yang
Abstract: Mixture-of-Experts (MoE) architectures have emerged as a key strategy for scaling large language models (LLMs) efficiently. However, current MoE systems suffer from severe load imbalance, where only a small subset of experts is consistently activated during training and inference, leading to significant underutilization of model capacity and computational resources. In this work, we revisit expert routing through a clustering perspective and propose Latent Prototype Routing (LPR), a novel routing framework that generalizes existing approaches while promoting balanced expert utilization without compromising downstream performance. Extensive experiments across multiple open-source MoE models -- including DeepSeek-V3, Qwen3-MoE, and Mixtral -- demonstrate that LPR reduces the Gini coefficient of expert load from 0.70 to 0.035 on average, improves the min-max expert load ratio from 1e-6 to 0.70, achieving near-perfect load balancing.
Authors: Galvin Brice S. Lim, Brian Godwin S. Lim, Argel A. Bandala, John Anthony C. Jose, Timothy Scott C. Chu, Edwin Sybingco
Abstract: Brain-computer interface (BCI) technology utilizing electroencephalography (EEG) marks a transformative innovation, empowering motor-impaired individuals to engage with their environment on equal footing. Despite its promising potential, developing subject-invariant and session-invariant BCI systems remains a significant challenge due to the inherent complexity and variability of neural activity across individuals and over time, compounded by EEG hardware constraints. While prior studies have sought to develop robust BCI systems, existing approaches remain ineffective in capturing the intricate spatiotemporal dependencies within multichannel EEG signals. This study addresses this gap by introducing the attentive graph-temporal convolutional network (AGTCNet), a novel graph-temporal model for motor imagery EEG (MI-EEG) classification. Specifically, AGTCNet leverages the topographic configuration of EEG electrodes as an inductive bias and integrates graph convolutional attention network (GCAT) to jointly learn expressive spatiotemporal EEG representations. The proposed model significantly outperformed existing MI-EEG classifiers, achieving state-of-the-art performance while utilizing a compact architecture, underscoring its effectiveness and practicality for BCI deployment. With a 49.87% reduction in model size, 64.65% faster inference time, and shorter input EEG signal, AGTCNet achieved a moving average accuracy of 66.82% for subject-independent classification on the BCI Competition IV Dataset 2a, which further improved to 82.88% when fine-tuned for subject-specific classification. On the EEG Motor Movement/Imagery Dataset, AGTCNet achieved moving average accuracies of 64.14% and 85.22% for 4-class and 2-class subject-independent classifications, respectively, with further improvements to 72.13% and 90.54% for subject-specific classifications.
Authors: Jingyao Li, Hao Sun, Zile Qiao, Yong Jiang, Pengjun Xie, Fei Huang, Hong Xu, Jiaya Jia
Abstract: Traditional benchmarks for large language models (LLMs) typically rely on static evaluations through storytelling or opinion expression, which fail to capture the dynamic requirements of real-time information processing in contemporary applications. To address this limitation, we present DynamicBench, a benchmark designed to evaluate the proficiency of LLMs in storing and processing up-to-the-minute data. DynamicBench utilizes a dual-path retrieval pipeline, integrating web searches with local report databases. It necessitates domain-specific knowledge, ensuring accurate responses report generation within specialized fields. By evaluating models in scenarios that either provide or withhold external documents, DynamicBench effectively measures their capability to independently process recent information or leverage contextual enhancements. Additionally, we introduce an advanced report generation system adept at managing dynamic information synthesis. Our experimental results confirm the efficacy of our approach, with our method achieving state-of-the-art performance, surpassing GPT4o in document-free and document-assisted scenarios by 7.0% and 5.8%, respectively. The code and data will be made publicly available.
Authors: Le Vu Anh, Mehmet Dik, Nguyen Viet Anh
Abstract: Persistent Laplacians are matrix operators that track how the shape and structure of data transform across scales and are popularly adopted in biology, physics, and machine learning. Their eigenvalues are concise descriptors of geometric and topological features in a filtration. Although earlier work established global algebraic stability for these operators, the precise change in a single eigenvalue when one simplex, such as a vertex, edge, or triangle, is added has remained unknown. This is important because downstream tools, including heat-kernel signatures and spectral neural networks, depend directly on these eigenvalues. We close this gap by proving a uniform Lipschitz bound: after inserting one simplex, every up-persistent Laplacian eigenvalue can vary by at most twice the Euclidean norm of that simplex's boundary, independent of filtration scale and complex size. This result delivers the first eigenvalue-level robustness guarantee for spectral topological data analysis. It guarantees that spectral features remain stable under local updates and enables reliable error control in dynamic data settings.
Authors: Melanie Rieff, Maya Varma, Ossian Rabow, Subathra Adithan, Julie Kim, Ken Chang, Hannah Lee, Nidhi Rohatgi, Christian Bluethgen, Mohamed S. Muneer, Jean-Benoit Delbrouck, Michael Moor
Abstract: Multimodal in-context learning (ICL) remains underexplored despite significant potential for domains such as medicine. Clinicians routinely encounter diverse, specialized tasks requiring adaptation from limited examples, such as drawing insights from a few relevant prior cases or considering a constrained set of differential diagnoses. While multimodal large language models (MLLMs) have shown advances in medical visual question answering (VQA), their ability to learn multimodal tasks from context is largely unknown. We introduce SMMILE, the first expert-driven multimodal ICL benchmark for medical tasks. Eleven medical experts curated problems, each including a multimodal query and multimodal in-context examples as task demonstrations. SMMILE encompasses 111 problems (517 question-image-answer triplets) covering 6 medical specialties and 13 imaging modalities. We further introduce SMMILE++, an augmented variant with 1038 permuted problems. A comprehensive evaluation of 15 MLLMs demonstrates that most models exhibit moderate to poor multimodal ICL ability in medical tasks. In open-ended evaluations, ICL contributes only 8% average improvement over zero-shot on SMMILE and 9.4% on SMMILE++. We observe a susceptibility for irrelevant in-context examples: even a single noisy or irrelevant example can degrade performance by up to 9.5%. Moreover, example ordering exhibits a recency bias, i.e., placing the most relevant example last can lead to substantial performance improvements by up to 71%. Our findings highlight critical limitations and biases in current MLLMs when learning multimodal medical tasks from context.
Authors: Sam Lerman, Jing Bi
Abstract: rQdia regularizes Q-value distributions with augmented images in pixel-based deep reinforcement learning. With a simple auxiliary loss, that equalizes these distributions via MSE, rQdia boosts DrQ and SAC on 9/12 and 10/12 tasks respectively in the MuJoCo Continuous Control Suite from pixels, and Data-Efficient Rainbow on 18/26 Atari Arcade environments. Gains are measured in both sample efficiency and longer-term training. Moreover, the addition of rQdia finally propels model-free continuous control from pixels over the state encoding baseline.
Authors: Vasileios Leon, Georgios Makris, Sotirios Xydis, Kiamal Pekmestzi, Dimitrios Soudris
Abstract: Nowadays, the rapid growth of Deep Neural Network (DNN) architectures has established them as the defacto approach for providing advanced Machine Learning tasks with excellent accuracy. Targeting low-power DNN computing, this paper examines the interplay of fine-grained error resilience of DNN workloads in collaboration with hardware approximation techniques, to achieve higher levels of energy efficiency. Utilizing the state-of-the-art ROUP approximate multipliers, we systematically explore their fine-grained distribution across the network according to our layer-, filter-, and kernel-level approaches, and examine their impact on accuracy and energy. We use the ResNet-8 model on the CIFAR-10 dataset to evaluate our approximations. The proposed solution delivers up to 54% energy gains in exchange for up to 4% accuracy loss, compared to the baseline quantized model, while it provides 2x energy gains with better accuracy versus the state-of-the-art DNN approximations.
Authors: Chao Zhou, Tom Jacobs, Advait Gadhikar, Rebekka Burkholz
Abstract: Finetuning large pretrained neural networks is known to be resource-intensive, both in terms of memory and computational cost. To mitigate this, a common approach is to restrict training to a subset of the model parameters. By analyzing the relationship between gradients and weights during finetuning, we observe a notable pattern: large gradients are often associated with small-magnitude weights. This correlation is more pronounced in finetuning settings than in training from scratch. Motivated by this observation, we propose NANOADAM, which dynamically updates only the small-magnitude weights during finetuning and offers several practical advantages: first, this criterion is gradient-free -- the parameter subset can be determined without gradient computation; second, it preserves large-magnitude weights, which are likely to encode critical features learned during pretraining, thereby reducing the risk of catastrophic forgetting; thirdly, it permits the use of larger learning rates and consistently leads to better generalization performance in experiments. We demonstrate this for both NLP and vision tasks.
Authors: Zhi Zheng, Bochuan Zhou, Yuping Song
Abstract: Cryptocurrency transaction fraud detection faces the dual challenges of increasingly complex transaction patterns and severe class imbalance. Traditional methods rely on manual feature engineering and struggle to capture temporal and structural dependencies in transaction networks. This paper proposes an Augmented Temporal-aware Graph Attention Network (ATGAT) that enhances detection performance through three modules: (1) designing an advanced temporal embedding module that fuses multi-scale time difference features with periodic position encoding; (2) constructing a temporal-aware triple attention mechanism that jointly optimizes structural, temporal, and global context attention; (3) employing weighted BCE loss to address class imbalance. Experiments on the Elliptic++ cryptocurrency dataset demonstrate that ATGAT achieves an AUC of 0.9130, representing a 9.2% improvement over the best traditional method XGBoost, 12.0% over GCN, and 10.0% over standard GAT. This method not only validates the enhancement effect of temporal awareness and triple attention mechanisms on graph neural networks, but also provides financial institutions with more reliable fraud detection tools, with its design principles generalizable to other temporal graph anomaly detection tasks.
Authors: Jaris K\"uken, Lennart Purucker, Frank Hutter
Abstract: Tabular foundation models have shown strong performance across various tabular learning tasks via in-context learning, offering robust generalization without any downstream finetuning. However, their inference-time costs remain high, particularly for larger datasets. To address this, we propose early-stopping the in-context learning process. We achieve this by dynamically evaluating whether to stop in-context learning after each Transformer encoder layer. Once stopped, we decode the embedding using a pre-trained layer-wise decoder. Experiments across 34 small classification tasks size show that early stopping in-context learning accelerates inference by up to x1.3 with negligible degradation in predictive performance. To assess scalability, we further evaluate our method on five larger classification tasks, achieving speedups of up to x2.2. Our results demonstrate the potential of early exiting as an effective and practical strategy for improving the efficiency of tabular in-context learning.
Authors: Colin Samplawski, Adam D. Cobb, Manoj Acharya, Ramneet Kaur, Susmit Jha
Abstract: Despite their widespread use, large language models (LLMs) are known to hallucinate incorrect information and be poorly calibrated. This makes the uncertainty quantification of these models of critical importance, especially in high-stakes domains, such as autonomy and healthcare. Prior work has made Bayesian deep learning-based approaches to this problem more tractable by performing inference over the low-rank adaptation (LoRA) parameters of a fine-tuned model. While effective, these approaches struggle to scale to larger LLMs due to requiring further additional parameters compared to LoRA. In this work we present $\textbf{Scala}$ble $\textbf{B}$ayesian $\textbf{L}$ow-Rank Adaptation via Stochastic Variational Subspace Inference (ScalaBL). We perform Bayesian inference in an $r$-dimensional subspace, for LoRA rank $r$. By repurposing the LoRA parameters as projection matrices, we are able to map samples from this subspace into the full weight space of the LLM. This allows us to learn all the parameters of our approach using stochastic variational inference. Despite the low dimensionality of our subspace, we are able to achieve competitive performance with state-of-the-art approaches while only requiring ${\sim}1000$ additional parameters. Furthermore, it allows us to scale up to the largest Bayesian LLM to date, with four times as a many base parameters as prior work.
Authors: Aristeidis Tsaris, Isaac Lyngaas, John Lagregren, Mohamed Wahib, Larry York, Prasanna Balaprakash, Dan Lu, Feiyi Wang, Xiao Wang
Abstract: Vision-based scientific foundation models hold significant promise for advancing scientific discovery and innovation. This potential stems from their ability to aggregate images from diverse sources such as varying physical groundings or data acquisition systems and to learn spatio-temporal correlations using transformer architectures. However, tokenizing and aggregating images can be compute-intensive, a challenge not fully addressed by current distributed methods. In this work, we introduce the Distributed Cross-Channel Hierarchical Aggregation (D-CHAG) approach designed for datasets with a large number of channels across image modalities. Our method is compatible with any model-parallel strategy and any type of vision transformer architecture, significantly improving computational efficiency. We evaluated D-CHAG on hyperspectral imaging and weather forecasting tasks. When integrated with tensor parallelism and model sharding, our approach achieved up to a 75% reduction in memory usage and more than doubled sustained throughput on up to 1,024 AMD GPUs on the Frontier Supercomputer.
Authors: Prajwal Koirala, Cody Fleming
Abstract: Generative models such as diffusion and flow-matching offer expressive policies for offline reinforcement learning (RL) by capturing rich, multimodal action distributions, but their iterative sampling introduces high inference costs and training instability due to gradient propagation across sampling steps. We propose the \textit{Single-Step Completion Policy} (SSCP), a generative policy trained with an augmented flow-matching objective to predict direct completion vectors from intermediate flow samples, enabling accurate, one-shot action generation. In an off-policy actor-critic framework, SSCP combines the expressiveness of generative models with the training and inference efficiency of unimodal policies, without requiring long backpropagation chains. Our method scales effectively to offline, offline-to-online, and online RL settings, offering substantial gains in speed and adaptability over diffusion-based baselines. We further extend SSCP to goal-conditioned RL, enabling flat policies to exploit subgoal structures without explicit hierarchical inference. SSCP achieves strong results across standard offline RL and behavior cloning benchmarks, positioning it as a versatile, expressive, and efficient framework for deep RL and sequential decision-making.
Authors: Franco Rugolon, Thomas Jack Samuels, Stephan Hau, Lennart H\"ogman
Abstract: This study investigates the efficacy of using multimodal machine learning techniques to detect deception in dyadic interactions, focusing on the integration of data from both the deceiver and the deceived. We compare early and late fusion approaches, utilizing audio and video data - specifically, Action Units and gaze information - across all possible combinations of modalities and participants. Our dataset, newly collected from Swedish native speakers engaged in truth or lie scenarios on emotionally relevant topics, serves as the basis for our analysis. The results demonstrate that incorporating both speech and facial information yields superior performance compared to single-modality approaches. Moreover, including data from both participants significantly enhances deception detection accuracy, with the best performance (71%) achieved using a late fusion strategy applied to both modalities and participants. These findings align with psychological theories suggesting differential control of facial and vocal expressions during initial interactions. As the first study of its kind on a Scandinavian cohort, this research lays the groundwork for future investigations into dyadic interactions, particularly within psychotherapy settings.
Authors: Jens P\"uttschneider, Simon Heilig, Asja Fischer, Timm Faulwasser
Abstract: We propose a training formulation for ResNets reflecting an optimal control problem that is applicable for standard architectures and general loss functions. We suggest bridging both worlds via penalizing intermediate outputs of hidden states corresponding to stage cost terms in optimal control. For standard ResNets, we obtain intermediate outputs by propagating the state through the subsequent skip connections and the output layer. We demonstrate that our training dynamic biases the weights of the unnecessary deeper residual layers to vanish. This indicates the potential for a theory-grounded layer pruning strategy.
Authors: Tamim Al Mahmud, Md Gulzar Hussain, Sumaiya Kabir, Hasnain Ahmad, Mahmudus Sobhan
Abstract: Evaluation is the method of assessing and determining the educational system through various techniques such as verbal or viva-voice test, subjective or objective written test. This paper presents an efficient solution to evaluate the subjective answer script electronically. In this paper, we proposed and implemented an integrated system that examines and evaluates the written answer script. This article focuses on finding the keywords from the answer script and then compares them with the keywords that have been parsed from both open and closed domain. The system also checks the grammatical and spelling errors in the answer script. Our proposed system tested with answer scripts of 100 students and gives precision score 0.91.
Authors: Gavin Lee Goodship, Luis Miralles-Pechuan, Stephen O'Sullivan
Abstract: Extended Stability Runge-Kutta (ESRK) methods are crucial for solving large-scale computational problems in science and engineering, including weather forecasting, aerodynamic analysis, and complex biological modelling. However, balancing accuracy, stability, and computational efficiency remains challenging, particularly for high-order, low-storage schemes. This study introduces a hybrid Genetic Algorithm (GA) and Reinforcement Learning (RL) approach for automated heuristic discovery, optimising low-storage ESRK methods. Unlike traditional approaches that rely on manually designed heuristics or exhaustive numerical searches, our method leverages GA-driven mutations for search-space exploration and an RL-inspired state transition mechanism to refine heuristic selection dynamically. This enables systematic parameter reduction, preserving fourth-order accuracy while significantly improving computational efficiency.The proposed GA-RL heuristic optimisation framework is validated through rigorous testing on benchmark problems, including the 1D and 2D Brusselator systems and the steady-state Navier-Stokes equations. The best-performing heuristic achieves a 25\% reduction in IPOPT runtime compared to traditional ESRK optimisation processes while maintaining numerical stability and accuracy. These findings demonstrate the potential of adaptive heuristic discovery to improve resource efficiency in high-fidelity simulations and broaden the applicability of low-storage Runge-Kutta methods in real-world computational fluid dynamics, physics simulations, and other demanding fields. This work establishes a new paradigm in heuristic optimisation for numerical methods, opening pathways for further exploration using Deep RL and AutoML-based heuristic search
Authors: Priyanka Avhad, Vedanti Kshirsagar, Urvi Ranjan, Mahek Nakhua
Abstract: According to the data, the percent of women who underwent screening for cervical cancer, breast and oral cancer in Telangana in the year 2020 was 3.3 percent, 0.3 percent and 2.3 percent respectively. Although early detection is the only way to reduce morbidity and mortality, people have very low awareness about cervical and breast cancer signs and symptoms and screening practices. We developed an ML classification model to predict if a person is susceptible to breast or cervical cancer based on demographic factors. We devised a system to provide suggestions for the nearest hospital or Cancer treatment centres based on the users location or address. In addition to this, we can integrate the health card to maintain medical records of all individuals and conduct awareness drives and campaigns. For ML classification models, we used decision tree classification and support vector classification algorithms for cervical cancer susceptibility and breast cancer susceptibility respectively. Thus, by devising this solution we come one step closer to our goal which is spreading cancer awareness, thereby, decreasing the cancer mortality and increasing cancer literacy among the people of Telangana.
Authors: Francesco Vitale, Nicola Dall'Ora, Sebastiano Gaiardelli, Enrico Fraccaroli, Nicola Mazzocca, Franco Fummi
Abstract: Fault diagnosis in Cyber-Physical Systems (CPSs) is essential for ensuring system dependability and operational efficiency by accurately detecting anomalies and identifying their root causes. However, the manual modeling of faulty behaviors often demands extensive domain expertise and produces models that are complex, error-prone, and difficult to interpret. To address this challenge, we present a novel unsupervised fault diagnosis methodology that integrates collective anomaly detection in multivariate time series, process mining, and stochastic simulation. Initially, collective anomalies are detected from low-level sensor data using multivariate time-series analysis. These anomalies are then transformed into structured event logs, enabling the discovery of interpretable process models through process mining. By incorporating timing distributions into the extracted Petri nets, the approach supports stochastic simulation of faulty behaviors, thereby enhancing root cause analysis and behavioral understanding. The methodology is validated using the Robotic Arm Dataset (RoAD), a widely recognized benchmark in smart manufacturing. Experimental results demonstrate its effectiveness in modeling, simulating, and classifying faulty behaviors in CPSs. This enables the creation of comprehensive fault dictionaries that support predictive maintenance and the development of digital twins for industrial environments.
Authors: Xiaona Zhou, Constantin Brif, Ismini Lourentzou
Abstract: Multivariate time series anomaly detection (MTS-AD) is critical in domains like healthcare, cybersecurity, and industrial monitoring, yet remains challenging due to complex inter-variable dependencies, temporal dynamics, and sparse anomaly labels. We introduce mTSBench, the largest benchmark to date for MTS-AD and unsupervised model selection, spanning 344 labeled time series across 19 datasets and 12 diverse application domains. mTSBench evaluates 24 anomaly detection methods, including large language model (LLM)-based detectors for multivariate time series, and systematically benchmarks unsupervised model selection techniques under standardized conditions. Consistent with prior findings, our results confirm that no single detector excels across datasets, underscoring the importance of model selection. However, even state-of-the-art selection methods remain far from optimal, revealing critical gaps. mTSBench provides a unified evaluation suite to enable rigorous, reproducible comparisons and catalyze future advances in adaptive anomaly detection and robust model selection.
Authors: Ziyue Li, Chenrui Fan, Tianyi Zhou
Abstract: Grokking, i.e., test performance keeps improving long after training loss converged, has been recently witnessed in neural network training, making the mechanism of generalization and other emerging capabilities such as reasoning mysterious. While prior studies usually train small models on a few toy or highly-specific tasks for thousands of epochs, we conduct the first study of grokking on checkpoints during one-pass pretraining of a 7B large language model (LLM), i.e., OLMoE. We compute the training loss and evaluate generalization on diverse benchmark tasks, including math reasoning, code generation, and commonsense/domain-specific knowledge retrieval tasks. Our study, for the first time, verifies that grokking still happens in the pretraining of large-scale foundation models, though different data may enter grokking stages asynchronously. We further demystify grokking's "emergence of generalization" by investigating LLM internal dynamics. Specifically, we find that training samples' pathways (i.e., expert choices across layers) evolve from random, instance-specific to more structured and shareable between samples during grokking. Also, the complexity of a sample's pathway reduces despite the converged loss. These indicate a memorization-to-generalization conversion, providing a mechanistic explanation of delayed generalization. In the study, we develop two novel metrics to quantify pathway distance and the complexity of a single pathway. We show their ability to predict the generalization improvement on diverse downstream tasks. They are efficient, simple to compute and solely dependent on training data. Hence, they have practical value for pretraining, enabling us to monitor the generalization performance without finetuning and test. Theoretically, we show that more structured pathways reduce model complexity and improve the generalization bound.
Authors: Jacopo Dapueto, Nicoletta Noceti, Francesca Odone
Abstract: Developing meaningful and efficient representations that separate the fundamental structure of the data generation mechanism is crucial in representation learning. However, Disentangled Representation Learning has not fully shown its potential on real images, because of correlated generative factors, their resolution and limited access to ground truth labels. Specifically on the latter, we investigate the possibility of leveraging synthetic data to learn general-purpose disentangled representations applicable to real data, discussing the effect of fine-tuning and what properties of disentanglement are preserved after the transfer. We provide an extensive empirical study to address these issues. In addition, we propose a new interpretable intervention-based metric, to measure the quality of factors encoding in the representation. Our results indicate that some level of disentanglement, transferring a representation from synthetic to real data, is possible and effective.
Authors: Matja\v{z} Omladi\v{c}, Martin Vuk, Alja\v{z} Zalar
Abstract: A recent survey, nicknamed "Hitchhiker's Guide", J.J. Arias-Garc{\i}a, R. Mesiar, and B. De Baets, A hitchhiker's guide to quasi-copulas, Fuzzy Sets and Systems 393 (2020) 1-28, has raised the rating of quasi-copula problems in the dependence modeling community in spite of the lack of statistical interpretation of quasi-copulas. In our previous work (arXiv:2410.19339, accepted in Fuzzy Sets and Systems), we addressed the question of extreme values of the mass distribution associated with multivariate quasi-copulas. Using a linear programming approach, we were able to solve Open Problem 5 of the "Guide" up to dimension d = 17 and disprove a recent conjecture on the solution to that problem. In this paper, we use an analytical approach to provide a complete answer to the original question.
Authors: Anish Saxena, Po-An Tsai, Hritvik Taneja, Aamer Jaleel, Moinuddin Qureshi
Abstract: GPU memory bandwidth is the main bottleneck for low-latency Large Language Model (LLM) inference. Speculative decoding leverages idle GPU compute by using a lightweight drafter to propose K tokens, which the LLM verifies in parallel, boosting token throughput. In conventional dense LLMs, all model weights are fetched each iteration, so speculation adds no latency overhead. Emerging Mixture of Experts (MoE) models activate only a subset of weights per token, greatly reducing data movement. However, we show that speculation is ineffective for MoEs: draft tokens collectively activate more weights, increasing data movement and verification time by 2-3x. When token throughput gains fail to offset this overhead, speculation causes slowdowns up to 1.5x, making it infeasible. Even when useful, the optimal K varies by task, model, and even between requests and iterations. Thus, despite widespread use in dense LLMs, speculation remains impractical in leading MoEs. We present Cascade, a utility-driven framework that selectively enables speculation to avoid slowdowns and dynamically tunes K to accelerate MoE serving. Cascade uses a lightweight metric, speculation utility, the ratio of token gains to verification cost, which shows iteration-level locality, enabling periodic decisions via short test and longer set phases. For each request, Cascade disables speculation if utility drops below one during testing, and when utility exceeds one, tests multiple K-values to choose the utility-maximizing K for the set phase. We implement Cascade in vLLM and evaluate it on five popular MoEs with workloads spanning code, math, extraction, and mixed tasks. Cascade limits slowdown to 5% (vs. 1.5x) and improves throughput by 7-14% over static K, making speculative decoding practical for MoEs.
Authors: Hoa La, Ahan Gupta, Alex Morehead, Jianlin Cheng, Minjia Zhang
Abstract: Protein structure prediction models such as AlphaFold3 (AF3) push the frontier of biomolecular modeling by incorporating science-informed architectural changes to the transformer architecture. However, these advances come at a steep system cost, introducing: compute- and memory-intensive operators, 2D attention mechanisms, and retrieval-augmented data pipelines, which collectively hinder the scalability of AF3 training. In this work, we present MegaFold, a cross-platform system to accelerate AF3 training. MegaFold tackles key bottlenecks through ahead-of-time caching to eliminate GPU idle time from the retrieval-augmented data pipeline, Triton-based kernels for memory-efficient EvoAttention on heterogeneous devices, and deep fusion for common and critical small operators in AF3. Evaluation on both NVIDIA H200 and AMD MI250 GPUs shows that MegaFold reduces peak memory usage of AF3 training by up to 1.23$\times$ and improves per-iteration training time by up-to 1.73$\times$ and 1.62$\times$ respectively. More importantly, MegaFold enables training on 1.35$\times$ longer sequence lengths compared to PyTorch baselines without running out-of-memory, significantly improving the scalability of modern protein folding models. We open source our code at https://github.com/Supercomputing-System-AI-Lab/MegaFold/.
URLs: https://github.com/Supercomputing-System-AI-Lab/MegaFold/.
Authors: Racheal Mukisa, Arvind K. Bansal
Abstract: Artificial intelligence, including deep learning models, will play a transformative role in automated medical image analysis for the diagnosis of cardiac disorders and their management. Automated accurate delineation of cardiac images is the first necessary initial step for the quantification and automated diagnosis of cardiac disorders. In this paper, we propose a deep learning based enhanced UNet model, U-R-Veda, which integrates convolution transformations, vision transformer, residual links, channel-attention, and spatial attention, together with edge-detection based skip-connections for an accurate fully-automated semantic segmentation of cardiac magnetic resonance (CMR) images. The model extracts local-features and their interrelationships using a stack of combination convolution blocks, with embedded channel and spatial attention in the convolution block, and vision transformers. Deep embedding of channel and spatial attention in the convolution block identifies important features and their spatial localization. The combined edge information with channel and spatial attention as skip connection reduces information-loss during convolution transformations. The overall model significantly improves the semantic segmentation of CMR images necessary for improved medical image analysis. An algorithm for the dual attention module (channel and spatial attention) has been presented. Performance results show that U-R-Veda achieves an average accuracy of 95.2%, based on DSC metrics. The model outperforms the accuracy attained by other models, based on DSC and HD metrics, especially for the delineation of right-ventricle and left-ventricle-myocardium.
Authors: Zhen Yuan, Shaoqing Jiao, Yihang Xiao, Jiajie Peng
Abstract: The advent of single-cell multi-omics technologies has enabled the simultaneous profiling of diverse omics layers within individual cells. Integrating such multimodal data provides unprecedented insights into cellular identity, regulatory processes, and disease mechanisms. However, it remains challenging, as current methods often rely on selecting highly variable genes or peaks during preprocessing, which may inadvertently discard crucial biological information. Here, we present scMamba, a foundation model designed to integrate single-cell multi-omics data without the need for prior feature selection while preserving genomic positional information. scMamba introduces a patch-based cell tokenization strategy that treats genomics regions as words (tokens) and cells as sentences. Building upon the concept of state space duality, scMamba distills rich biological insights from high-dimensional, sparse single-cell multi-omics data. Additionally, our novel contrastive learning approach, enhanced with cosine similarity regularization, enables superior alignment across omics layers compared to traditional methods. Systematic benchmarking across multiple datasets demonstrates that scMamba significantly outperforms state-of-the-art methods in preserving biological variation, aligning omics layers, and enhancing key downstream tasks such as clustering, cell type annotation, and trajectory inference. Our findings position scMamba as a powerful tool for large-scale single-cell multi-omics integration, capable of handling large-scale atlases and advancing biological discovery.
Authors: Alain Bensoussan, Minh-Binh Tran, Bangjie Wang
Abstract: Although there is a substantial body of literature on control and optimization problems for parabolic and hyperbolic systems, the specific problem of controlling and optimizing the coefficients of the associated operators within such systems has not yet been thoroughly explored. In this work, we aim to initiate a line of research in control theory focused on optimizing and controlling the coefficients of these operators-a problem that naturally arises in the context of neural networks and supervised learning. In supervised learning, the primary objective is to transport initial data toward target data through the layers of a neural network. We propose a novel perspective: neural networks can be interpreted as partial differential equations (PDEs). From this viewpoint, the control problem traditionally studied in the context of ordinary differential equations (ODEs) is reformulated as a control problem for PDEs, specifically targeting the optimization and control of coefficients in parabolic and hyperbolic operators. To the best of our knowledge, this specific problem has not yet been systematically addressed in the control theory of PDEs. To this end, we propose a dual system formulation for the control and optimization problem associated with parabolic PDEs, laying the groundwork for the development of efficient numerical schemes in future research. We also provide a theoretical proof showing that the control and optimization problem for parabolic PDEs admits minimizers. Finally, we investigate the control problem associated with hyperbolic PDEs and prove the existence of solutions for a corresponding approximated control problem.
Authors: Tongtong Liang, Dan Qiao, Yu-Xiang Wang, Rahul Parhi
Abstract: We study the implicit bias of flatness / low (loss) curvature and its effects on generalization in two-layer overparameterized ReLU networks with multivariate inputs -- a problem well motivated by the minima stability and edge-of-stability phenomena in gradient-descent training. Existing work either requires interpolation or focuses only on univariate inputs. This paper presents new and somewhat surprising theoretical results for multivariate inputs. On two natural settings (1) generalization gap for flat solutions, and (2) mean-squared error (MSE) in nonparametric function estimation by stable minima, we prove upper and lower bounds, which establish that while flatness does imply generalization, the resulting rates of convergence necessarily deteriorate exponentially as the input dimension grows. This gives an exponential separation between the flat solutions vis-\`a-vis low-norm solutions (i.e., weight decay), which knowingly do not suffer from the curse of dimensionality. In particular, our minimax lower bound construction, based on a novel packing argument with boundary-localized ReLU neurons, reveals how flat solutions can exploit a kind of ''neural shattering'' where neurons rarely activate, but with high weight magnitudes. This leads to poor performance in high dimensions. We corroborate these theoretical findings with extensive numerical simulations. To the best of our knowledge, our analysis provides the first systematic explanation for why flat minima may fail to generalize in high dimensions.
Authors: Marc Bara
Abstract: We present the first theoretical framework for applying spiking neural networks (SNNs) to synthetic aperture radar (SAR) interferometric phase unwrapping. Despite extensive research in both domains, our comprehensive literature review confirms that SNNs have never been applied to phase unwrapping, representing a significant gap in current methodologies. As Earth observation data volumes continue to grow exponentially (with missions like NISAR expected to generate 100PB in two years) energy-efficient processing becomes critical for sustainable data center operations. SNNs, with their event-driven computation model, offer potential energy savings of 30-100x compared to conventional approaches while maintaining comparable accuracy. We develop spike encoding schemes specifically designed for wrapped phase data, propose SNN architectures that leverage the spatial propagation nature of phase unwrapping, and provide theoretical analysis of computational complexity and convergence properties. Our framework demonstrates how the temporal dynamics inherent in SNNs can naturally model the spatial continuity constraints fundamental to phase unwrapping. This work opens a new research direction at the intersection of neuromorphic computing and SAR interferometry, offering a complementary approach to existing algorithms that could enable more sustainable large-scale InSAR processing.
Authors: Cristian L\'opez, Keegan J. Moore
Abstract: Estimating the governing equation parameter values is essential for integrating experimental data with scientific theory to understand, validate, and predict the dynamics of complex systems. In this work, we propose a new method for structural system identification (SI), uncertainty quantification, and validation directly from data. Inspired by generative modeling frameworks, a neural network maps random noise to physically meaningful parameters. These parameters are then used in the known equation of motion to obtain fake accelerations, which are compared to real training data via a mean square error loss. To simultaneously validate the learned parameters, we use independent validation datasets. The generated accelerations from these datasets are evaluated by a discriminator network, which determines whether the output is real or fake, and guides the parameter-generator network. Analytical and real experiments show the parameter estimation accuracy and model validation for different nonlinear structural systems.
Authors: Chenglei Si, Tatsunori Hashimoto, Diyi Yang
Abstract: Large Language Models (LLMs) have shown promise in accelerating the scientific research pipeline. A key capability for this process is the ability to generate novel research ideas, and prior studies have found settings in which LLM-generated research ideas were judged as more novel than human-expert ideas. However, a good idea should not simply appear to be novel, it should also result in better research after being executed. To test whether AI-generated ideas lead to better research outcomes, we conduct an execution study by recruiting 43 expert researchers to execute randomly-assigned ideas, either written by experts or generated by an LLM. Each expert spent over 100 hours implementing the idea and wrote a 4-page short paper to document the experiments. All the executed projects are then reviewed blindly by expert NLP researchers. Comparing the review scores of the same ideas before and after execution, the scores of the LLM-generated ideas decrease significantly more than expert-written ideas on all evaluation metrics (novelty, excitement, effectiveness, and overall; p < 0.05), closing the gap between LLM and human ideas observed at the ideation stage. When comparing the aggregated review scores from the execution study, we even observe that for many metrics there is a flip in rankings where human ideas score higher than LLM ideas. This ideation-execution gap highlights the limitations of current LLMs in generating truly effective research ideas and the challenge of evaluating research ideas in the absence of execution outcomes.
Authors: Sinan Rasiya Koya, Tirthankar Roy
Abstract: Combining attention with recurrence has shown to be valuable in sequence modeling, including hydrological predictions. Here, we explore the strength of Temporal Fusion Transformers (TFTs) over Long Short-Term Memory (LSTM) networks in rainfall-runoff modeling. We train ten randomly initialized models, TFT and LSTM, for 531 CAMELS catchments in the US. We repeat the experiment with five subsets of the Caravan dataset, each representing catchments in the US, Australia, Brazil, Great Britain, and Chile. Then, the performance of the models, their variability regarding the catchment attributes, and the difference according to the datasets are assessed. Our findings show that TFT slightly outperforms LSTM, especially in simulating the midsection and peak of hydrographs. Furthermore, we show the ability of TFT to handle longer sequences and why it can be a better candidate for higher or larger catchments. Being an explainable AI technique, TFT identifies the key dynamic and static variables, providing valuable scientific insights. However, both TFT and LSTM exhibit a considerable drop in performance with the Caravan dataset, indicating possible data quality issues. Overall, the study highlights the potential of TFT in improving hydrological modeling and understanding.
Authors: Jing Luo, Yejun Gu, Yanfei Wang, Xiaolong Ma, Jaafar. A El-Awady
Abstract: Machine learning has significantly advanced the understanding and application of structural materials, with an increasing emphasis on integrating existing data and quantifying uncertainties in predictive modeling. This study presents a comprehensive methodology utilizing a mixed density network (MDN) model, trained on extensive experimental data from literature. This approach uniquely predicts the distribution of dislocation density, inferred as a latent variable, and the resulting stress distribution at the grain level. The incorporation of statistical parameters of those predicted distributions into a dislocation-mediated plasticity model allows for accurate stress-strain predictions with explicit uncertainty quantification. This strategy not only improves the accuracy and reliability of mechanical property predictions but also plays a vital role in optimizing alloy design, thereby facilitating the development of new materials in a rapidly evolving industry.
Authors: Osama Zafar, Rosemarie Santa Gonz\'alez, Mina Namazi, Alfonso Morales, Erman Ayday
Abstract: Data-driven agriculture, which integrates technology and data into agricultural practices, has the potential to improve crop yield, disease resilience, and long-term soil health. However, privacy concerns, such as adverse pricing, discrimination, and resource manipulation, deter farmers from sharing data, as it can be used against them. To address this barrier, we propose a privacy-preserving framework that enables secure data sharing and collaboration for research and development while mitigating privacy risks. The framework combines dimensionality reduction techniques (like Principal Component Analysis (PCA)) and differential privacy by introducing Laplacian noise to protect sensitive information. The proposed framework allows researchers to identify potential collaborators for a target farmer and train personalized machine learning models either on the data of identified collaborators via federated learning or directly on the aggregated privacy-protected data. It also allows farmers to identify potential collaborators based on similarities. We have validated this on real-life datasets, demonstrating robust privacy protection against adversarial attacks and utility performance comparable to a centralized system. We demonstrate how this framework can facilitate collaboration among farmers and help researchers pursue broader research objectives. The adoption of the framework can empower researchers and policymakers to leverage agricultural data responsibly, paving the way for transformative advances in data-driven agriculture. By addressing critical privacy challenges, this work supports secure data integration, fostering innovation and sustainability in agricultural systems.
Authors: Kyanna Dagenais, Istvan David
Abstract: Model-driven engineering problems often require complex model transformations (MTs), i.e., MTs that are chained in extensive sequences. Pertinent examples of such problems include model synchronization, automated model repair, and design space exploration. Manually developing complex MTs is an error-prone and often infeasible process. Reinforcement learning (RL) is an apt way to alleviate these issues. In RL, an autonomous agent explores the state space through trial and error to identify beneficial sequences of actions, such as MTs. However, RL methods exhibit performance issues in complex problems. In these situations, human guidance can be of high utility. In this paper, we present an approach and technical framework for developing complex MT sequences through RL, guided by potentially uncertain human advice. Our framework allows user-defined MTs to be mapped onto RL primitives, and executes them as RL programs to find optimal MT sequences. Our evaluation shows that human guidance, even if uncertain, substantially improves RL performance, and results in more efficient development of complex MTs. Through a trade-off between the certainty and timeliness of human advice, our method takes a step towards RL-driven human-in-the-loop engineering methods.
Authors: Matthew Zurek, Yudong Chen
Abstract: We study value-iteration (VI) algorithms for solving general (a.k.a. multichain) Markov decision processes (MDPs) under the average-reward criterion, a fundamental but theoretically challenging setting. Beyond the difficulties inherent to all average-reward problems posed by the lack of contractivity and non-uniqueness of solutions to the Bellman operator, in the multichain setting an optimal policy must solve the navigation subproblem of steering towards the best connected component, in addition to optimizing long-run performance within each component. We develop algorithms which better solve this navigational subproblem in order to achieve faster convergence for multichain MDPs, obtaining improved rates of convergence and sharper measures of complexity relative to prior work. Many key components of our results are of potential independent interest, including novel connections between average-reward and discounted problems, optimal fixed-point methods for discounted VI which extend to general Banach spaces, new sublinear convergence rates for the discounted value error, and refined suboptimality decompositions for multichain MDPs. Overall our results yield faster convergence rates for discounted and average-reward problems and expand the theoretical foundations of VI approaches.
Authors: Mina Namazi, Alexander Nemecek, Erman Ayday
Abstract: As the deployment of large language models (LLMs) grows in sensitive domains, ensuring the integrity of their computational provenance becomes a critical challenge, particularly in regulated sectors such as healthcare, where strict requirements are applied in dataset usage. We introduce ZKPROV, a novel cryptographic framework that enables zero-knowledge proofs of LLM provenance. It allows users to verify that a model is trained on a reliable dataset without revealing sensitive information about it or its parameters. Unlike prior approaches that focus on complete verification of the training process (incurring significant computational cost) or depend on trusted execution environments, ZKPROV offers a distinct balance. Our method cryptographically binds a trained model to its authorized training dataset(s) through zero-knowledge proofs while avoiding proof of every training step. By leveraging dataset-signed metadata and compact model parameter commitments, ZKPROV provides sound and privacy-preserving assurances that the result of the LLM is derived from a model trained on the claimed authorized and relevant dataset. Experimental results demonstrate the efficiency and scalability of the ZKPROV in generating this proof and verifying it, achieving a practical solution for real-world deployments. We also provide formal security guarantees, proving that our approach preserves dataset confidentiality while ensuring trustworthy dataset provenance.
Authors: Yuanxing Cheng, Lulu Kang, Yiwei Wang, Chun Liu
Abstract: This paper introduces an active learning framework for manifold Gaussian Process (GP) regression, combining manifold learning with strategic data selection to improve accuracy in high-dimensional spaces. Our method jointly optimizes a neural network for dimensionality reduction and a Gaussian process regressor in the latent space, supervised by an active learning criterion that minimizes global prediction error. Experiments on synthetic data demonstrate superior performance over randomly sequential learning. The framework efficiently handles complex, discontinuous functions while preserving computational tractability, offering practical value for scientific and engineering applications. Future work will focus on scalability and uncertainty-aware manifold learning.
Authors: Chi-Sheng Chen, Xinyu Zhang, Ya-Chuan Chen
Abstract: We propose a hybrid quantum-classical reinforcement learning framework for sector rotation in the Taiwan stock market. Our system employs Proximal Policy Optimization (PPO) as the backbone algorithm and integrates both classical architectures (LSTM, Transformer) and quantum-enhanced models (QNN, QRWKV, QASA) as policy and value networks. An automated feature engineering pipeline extracts financial indicators from capital share data to ensure consistent model input across all configurations. Empirical backtesting reveals a key finding: although quantum-enhanced models consistently achieve higher training rewards, they underperform classical models in real-world investment metrics such as cumulative return and Sharpe ratio. This discrepancy highlights a core challenge in applying reinforcement learning to financial domains -- namely, the mismatch between proxy reward signals and true investment objectives. Our analysis suggests that current reward designs may incentivize overfitting to short-term volatility rather than optimizing risk-adjusted returns. This issue is compounded by the inherent expressiveness and optimization instability of quantum circuits under Noisy Intermediate-Scale Quantum (NISQ) constraints. We discuss the implications of this reward-performance gap and propose directions for future improvement, including reward shaping, model regularization, and validation-based early stopping. Our work offers a reproducible benchmark and critical insights into the practical challenges of deploying quantum reinforcement learning in real-world finance.
Authors: Erik Jahn, Frederick Eberhardt, Leonard J. Schulman
Abstract: Causal discovery algorithms typically recover causal graphs only up to their Markov equivalence classes unless additional parametric assumptions are made. The sizes of these equivalence classes reflect the limits of what can be learned about the underlying causal graph from purely observational data. Under the assumptions of acyclicity, causal sufficiency, and a uniform model prior, Markov equivalence classes are known to be small on average. In this paper, we show that this is no longer the case when any of these assumptions is relaxed. Specifically, we prove exponentially large lower bounds for the expected size of Markov equivalence classes in three settings: sparse random directed acyclic graphs, uniformly random acyclic directed mixed graphs, and uniformly random directed cyclic graphs.
Authors: Hsin-Hsiung Huang, Hayden Hampton
Abstract: Forecasting geopolitical conflict from data sources like the Global Database of Events, Language, and Tone (GDELT) is a critical challenge for national security. The inherent sparsity, burstiness, and overdispersion of such data cause standard deep learning models, including the Temporal Fusion Transformer (TFT), to produce unreliable long-horizon predictions. We introduce STFT-VNNGP, a hybrid architecture that won the 2023 Algorithms for Threat Detection (ATD) competition by overcoming these limitations. Designed to bridge this gap, our model employs a two-stage process: first, a TFT captures complex temporal dynamics to generate multi-quantile forecasts. These quantiles then serve as informed inputs for a Variational Nearest Neighbor Gaussian Process (VNNGP), which performs principled spatiotemporal smoothing and uncertainty quantification. In a case study forecasting conflict dynamics in the Middle East and the U.S., STFT-VNNGP consistently outperforms a standalone TFT, showing a superior ability to predict the timing and magnitude of bursty event periods, particularly at long-range horizons. This work offers a robust framework for generating more reliable and actionable intelligence from challenging event data, with all code and workflows made publicly available to ensure reproducibility.
Authors: Fangyuan Zhang, Zhengjun Huang, Yingli Zhou, Qintian Guo, Zhixun Li, Wensheng Luo, Di Jiang, Yixiang Fang, Xiaofang Zhou
Abstract: Graph-based Retrieval-Augmented Generation (Graph-RAG) enhances large language models (LLMs) by structuring retrieval over an external corpus. However, existing approaches typically assume a static corpus, requiring expensive full-graph reconstruction whenever new documents arrive, limiting their scalability in dynamic, evolving environments. To address these limitations, we introduce EraRAG, a novel multi-layered Graph-RAG framework that supports efficient and scalable dynamic updates. Our method leverages hyperplane-based Locality-Sensitive Hashing (LSH) to partition and organize the original corpus into hierarchical graph structures, enabling efficient and localized insertions of new data without disrupting the existing topology. The design eliminates the need for retraining or costly recomputation while preserving high retrieval accuracy and low latency. Experiments on large-scale benchmarks demonstrate that EraRag achieves up to an order of magnitude reduction in update time and token consumption compared to existing Graph-RAG systems, while providing superior accuracy performance. This work offers a practical path forward for RAG systems that must operate over continually growing corpora, bridging the gap between retrieval efficiency and adaptability. Our code and data are available at https://github.com/EverM0re/EraRAG-Official.
Authors: Eric Zhang, Leshem Choshen, Jacob Andreas
Abstract: There are two primary ways of incorporating new information into a language model (LM): changing its prompt or changing its parameters, e.g. via fine-tuning. Parameter updates incur no long-term storage cost for model changes. However, for many model updates, prompting is significantly more effective: prompted models can generalize robustly from single examples and draw logical inferences that do not occur under standard fine-tuning. Can models be modified so that fine-tuning does emulate prompting? This paper describes a method for meta-training LMs such that gradient updates emulate the effects of conditioning on new information. Our approach uses tools from gradient-based meta-learning but uses an LM's own prompted predictions as targets, eliminating the need for ground-truth labels. Subsequent gradient descent training recovers some (and occasionally all) of prompted model performance -- showing improvement on the ``reversal curse'' tasks, and answering questions about text passages after a single gradient update. These results suggest that, with appropriate initialization, gradient descent can be surprisingly expressive. Our results suggest new avenues for long-context modeling and offer insight into the generalization capabilities of gradient-based learning.
Authors: Akio Hayakawa, Masato Ishii, Takashi Shibuya, Yuki Mitsufuji
Abstract: We propose a novel step-by-step video-to-audio generation method that sequentially produces individual audio tracks, each corresponding to a specific sound event in the video. Our approach mirrors traditional Foley workflows, aiming to capture all sound events induced by a given video comprehensively. Each generation step is formulated as a guided video-to-audio synthesis task, conditioned on a target text prompt and previously generated audio tracks. This design is inspired by the idea of concept negation from prior compositional generation frameworks. To enable this guided generation, we introduce a training framework that leverages pre-trained video-to-audio models and eliminates the need for specialized paired datasets, allowing training on more accessible data. Experimental results demonstrate that our method generates multiple semantically distinct audio tracks for a single input video, leading to higher-quality composite audio synthesis than existing baselines.
Authors: Qingyue Jiao, Kangyu Zheng, Yiyu Shi, Zhiding Liang
Abstract: Machine learning-assisted diagnosis is gaining traction in skin disease detection, but training effective models requires large amounts of high-quality data. Skin disease datasets often suffer from class imbalance, privacy concerns, and object bias, making data augmentation essential. While classical generative models are widely used, they demand extensive computational resources and lengthy training time. Quantum computing offers a promising alternative, but existing quantum-based image generation methods can only yield grayscale low-quality images. Through a novel classical-quantum latent space fusion technique, our work overcomes this limitation and introduces the first classical-quantum generative adversarial network (GAN) capable of generating color medical images. Our model outperforms classical deep convolutional GANs and existing hybrid classical-quantum GANs in both image generation quality and classification performance boost when used as data augmentation. Moreover, the performance boost is comparable with that achieved using state-of-the-art classical generative models, yet with over 25 times fewer parameters and 10 times fewer training epochs. Such results suggest a promising future for quantum image generation as quantum hardware advances. Finally, we demonstrate the robust performance of our model on real IBM quantum machine with hardware noise.
Authors: Hansam Cho, Seoung Bum Kim
Abstract: Text-guided diffusion models have become essential for high-quality image synthesis, enabling dynamic image editing. In image editing, two crucial aspects are editability, which determines the extent of modification, and faithfulness, which reflects how well unaltered elements are preserved. However, achieving optimal results is challenging because of the inherent trade-off between editability and faithfulness. To address this, we propose Faithfulness Guidance and Scheduling (FGS), which enhances faithfulness with minimal impact on editability. FGS incorporates faithfulness guidance to strengthen the preservation of input image information and introduces a scheduling strategy to resolve misalignment between editability and faithfulness. Experimental results demonstrate that FGS achieves superior faithfulness while maintaining editability. Moreover, its compatibility with various editing methods enables precise, high-quality image edits across diverse tasks.
Authors: Yann Kerzreho (ENS Paris Saclay)
Abstract: This paper introduces a new approach for approximating the learning dynamics of multiple reinforcement learning (RL) agents interacting in a finite-state Markov game. The idea is to rescale the learning process by simultaneously reducing the learning rate and increasing the update frequency, effectively treating the agent's parameters as a slow-evolving variable influenced by the fast-mixing game state. Under mild assumptions-ergodicity of the state process and continuity of the updates-we prove the convergence of this rescaled process to an ordinary differential equation (ODE). This ODE provides a tractable, deterministic approximation of the agent's learning dynamics. An implementation of the framework is available at\,: https://github.com/yannKerzreho/MarkovGameApproximation
URLs: https://github.com/yannKerzreho/MarkovGameApproximation
Authors: Sanjoy Chowdhury, Subrata Biswas, Sayan Nag, Tushar Nagarajan, Calvin Murdock, Ishwarya Ananthabhotla, Yijun Qian, Vamsi Krishna Ithapu, Dinesh Manocha, Ruohan Gao
Abstract: Modern perception models, particularly those designed for multisensory egocentric tasks, have achieved remarkable performance but often come with substantial computational costs. These high demands pose challenges for real-world deployment, especially in resource-constrained environments. In this paper, we introduce EgoAdapt, a framework that adaptively performs cross-modal distillation and policy learning to enable efficient inference across different egocentric perception tasks, including egocentric action recognition, active speaker localization, and behavior anticipation. Our proposed policy module is adaptable to task-specific action spaces, making it broadly applicable. Experimental results on three challenging egocentric datasets EPIC-Kitchens, EasyCom, and Aria Everyday Activities demonstrate that our method significantly enhances efficiency, reducing GMACs by up to 89.09%, parameters up to 82.02%, and energy up to 9.6x, while still on-par and in many cases outperforming, the performance of corresponding state-of-the-art models.
Authors: Yangzhe Peng, Kaiyuan Gao, Liang He, Yuheng Cong, Haiguang Liu, Kun He, Lijun Wu
Abstract: Molecular docking plays a crucial role in predicting the binding mode of ligands to target proteins, and covalent interactions, which involve the formation of a covalent bond between the ligand and the target, are particularly valuable due to their strong, enduring binding nature. However, most existing docking methods and deep learning approaches hardly account for the formation of covalent bonds and the associated structural changes. To address this gap, we introduce a comprehensive benchmark for covalent docking, CovDocker, which is designed to better capture the complexities of covalent binding. We decompose the covalent docking process into three main tasks: reactive location prediction, covalent reaction prediction, and covalent docking. By adapting state-of-the-art models, such as Uni-Mol and Chemformer, we establish baseline performances and demonstrate the effectiveness of the benchmark in accurately predicting interaction sites and modeling the molecular transformations involved in covalent binding. These results confirm the role of the benchmark as a rigorous framework for advancing research in covalent drug design. It underscores the potential of data-driven approaches to accelerate the discovery of selective covalent inhibitors and addresses critical challenges in therapeutic development.
Authors: Luosheng Xu, Dalin Zhang, Zhaohui Song
Abstract: Remote sensing change detection is essential for monitoring urban expansion, disaster assessment, and resource management, offering timely, accurate, and large-scale insights into dynamic landscape transformations. While deep learning has revolutionized change detection, the increasing complexity and computational demands of modern models have not necessarily translated into significant accuracy gains. Instead of following this trend, this study explores a more efficient approach, focusing on lightweight models that maintain high accuracy while minimizing resource consumption, which is an essential requirement for on-satellite processing. To this end, we propose FlickCD, which means quick flick then get great results, pushing the boundaries of the performance-resource trade-off. FlickCD introduces an Enhanced Difference Module (EDM) to amplify critical feature differences between temporal phases while suppressing irrelevant variations such as lighting and weather changes, thereby reducing computational costs in the subsequent change decoder. Additionally, the FlickCD decoder incorporates Local-Global Fusion Blocks, leveraging Shifted Window Self-Attention (SWSA) and Enhanced Global Self-Attention (EGSA) to efficiently capture semantic information at multiple scales, preserving both coarse- and fine-grained changes. Extensive experiments on four benchmark datasets demonstrate that FlickCD reduces computational and storage overheads by more than an order of magnitude while achieving state-of-the-art (SOTA) performance or incurring only a minor (<1\% F1) accuracy trade-off. The implementation code is publicly available at https://github.com/xulsh8/FlickCD.
Authors: He Li, Haoang Chi, Mingyu Liu, Wanrong Huang, Liyang Xu, Wenjing Yang
Abstract: The real world naturally has dimensions of time and space. Therefore, estimating the counterfactual outcomes with spatial-temporal attributes is a crucial problem. However, previous methods are based on classical statistical models, which still have limitations in performance and generalization. This paper proposes a novel framework for estimating counterfactual outcomes with spatial-temporal attributes using the Transformer, exhibiting stronger estimation ability. Under mild assumptions, the proposed estimator within this framework is consistent and asymptotically normal. To validate the effectiveness of our approach, we conduct simulation experiments and real data experiments. Simulation experiments show that our estimator has a stronger estimation capability than baseline methods. Real data experiments provide a valuable conclusion to the causal effect of conflicts on forest loss in Colombia. The source code is available at https://github.com/lihe-maxsize/DeppSTCI_Release_Version-master.
URLs: https://github.com/lihe-maxsize/DeppSTCI_Release_Version-master.
Authors: Jongyeon Park, Joonhee Lee, Do-Hyeon Lim, Hong Kook Kim, Hyeongcheol Geum, Jeong Eun Lim
Abstract: This technical report presents submission systems for Task 4 of the DCASE 2025 Challenge. This model incorporates additional audio features (spectral roll-off and chroma features) into the embedding feature extracted from the mel-spectral feature to im-prove the classification capabilities of an audio-tagging model in the spatial semantic segmentation of sound scenes (S5) system. This approach is motivated by the fact that mixed audio often contains subtle cues that are difficult to capture with mel-spectrograms alone. Thus, these additional features offer alterna-tive perspectives for the model. Second, an agent-based label correction system is applied to the outputs processed by the S5 system. This system reduces false positives, improving the final class-aware signal-to-distortion ratio improvement (CA-SDRi) metric. Finally, we refine the training dataset to enhance the classi-fication accuracy of low-performing classes by removing irrele-vant samples and incorporating external data. That is, audio mix-tures are generated from a limited number of data points; thus, even a small number of out-of-class data points could degrade model performance. The experiments demonstrate that the submit-ted systems employing these approaches relatively improve CA-SDRi by up to 14.7% compared to the baseline of DCASE 2025 Challenge Task 4.
Authors: Haoang Chi, He Li, Wenjing Yang, Feng Liu, Long Lan, Xiaoguang Ren, Tongliang Liu, Bo Han
Abstract: Causal reasoning capability is critical in advancing large language models (LLMs) toward strong artificial intelligence. While versatile LLMs appear to have demonstrated capabilities in understanding contextual causality and providing responses that obey the laws of causality, it remains unclear whether they perform genuine causal reasoning akin to humans. However, current evidence indicates the contrary. Specifically, LLMs are only capable of performing shallow (level-1) causal reasoning, primarily attributed to the causal knowledge embedded in their parameters, but they lack the capacity for genuine human-like (level-2) causal reasoning. To support this hypothesis, methodologically, we delve into the autoregression mechanism of transformer-based LLMs, revealing that it is not inherently causal. Empirically, we introduce a new causal Q&A benchmark called CausalProbe-2024, whose corpora are fresh and nearly unseen for the studied LLMs. The LLMs exhibit a significant performance drop on CausalProbe-2024 compared to earlier benchmarks, indicating the fact that they primarily engage in level-1 causal reasoning. To bridge the gap towards level-2 causal reasoning, we draw inspiration from the fact that human reasoning is usually facilitated by general knowledge and intended goals. We propose G^2-Reasoner, a method that incorporates general knowledge and goal-oriented prompts into LLMs' causal reasoning processes. Experiments demonstrate that G^2-Reasoner significantly enhances LLMs' causal reasoning capability, particularly in fresh and counterfactual contexts. This work sheds light on a new path for LLMs to advance towards genuine causal reasoning, going beyond level-1 and making strides towards level-2.
Authors: Giorgos Demosthenous, Chryssis Georgiou, Eliada Polydorou
Abstract: This study investigates the impact of data source diversity on the performance of cryptocurrency forecasting models by integrating various data categories, including technical indicators, on-chain metrics, sentiment and interest metrics, traditional market indices, and macroeconomic indicators. We introduce the Crypto100 index, representing the top 100 cryptocurrencies by market capitalization, and propose a novel feature reduction algorithm to identify the most impactful and resilient features from diverse data sources. Our comprehensive experiments demonstrate that data source diversity significantly enhances the predictive performance of forecasting models across different time horizons. Key findings include the paramount importance of on-chain metrics for both short-term and long-term predictions, the growing relevance of traditional market indices and macroeconomic indicators for longer-term forecasts, and substantial improvements in model accuracy when diverse data sources are utilized. These insights help demystify the short-term and long-term driving factors of the cryptocurrency market and lay the groundwork for developing more accurate and resilient forecasting models.
Authors: Lukas Sablica, Kurt Hornik
Abstract: We propose a novel variational autoencoder (VAE) architecture that employs a spherical Cauchy (spCauchy) latent distribution. Unlike traditional Gaussian latent spaces or the widely used von Mises-Fisher (vMF) distribution, spCauchy provides a more natural hyperspherical representation of latent variables, better capturing directional data while maintaining flexibility. Its heavy-tailed nature prevents over-regularization, ensuring efficient latent space utilization while offering a more expressive representation. Additionally, spCauchy circumvents the numerical instabilities inherent to vMF, which arise from computing normalization constants involving Bessel functions. Instead, it enables a fully differentiable and efficient reparameterization trick via M\"obius transformations, allowing for stable and scalable training. The KL divergence can be computed through a rapidly converging power series, eliminating concerns of underflow or overflow associated with evaluation of ratios of hypergeometric functions. These properties make spCauchy a compelling alternative for VAEs, offering both theoretical advantages and practical efficiency in high-dimensional generative modeling.
Authors: Istabrak Abbes, Gabriele Prato, Quentin Fournier, Fernando Rodriguez, Alaa Boukhary, Adam Elwood, Sarath Chandar
Abstract: Augmenting large language models (LLMs) with external context significantly improves their performance in natural language processing (NLP) tasks. However, LLMs struggle to answer queries reliably when the provided context lacks information, often resorting to ungrounded speculation or internal knowledge. Groundedness - generating responses strictly supported by the context - is essential for ensuring factual consistency and trustworthiness. This study focuses on detecting whether a given query is grounded in a document provided in context before the costly answer generation by LLMs. Such a detection mechanism can significantly reduce both inference time and resource consumption. We show that lightweight, task specific encoder models such as RoBERTa and NomicBERT, fine-tuned on curated datasets, can achieve accuracy comparable to state-of-the-art LLMs, such as Llama3 8B and GPT4o, in groundedness detection while reducing inference latency by orders of magnitude. The code is available at : https://github.com/chandarlab/Hallucinate-less
Authors: Atharva Mehta, Shivam Chauhan, Monojit Choudhury
Abstract: Fine-tuning large-scale music generation models, such as MusicGen and Mustango, is a computationally expensive process, often requiring updates to billions of parameters and, therefore, significant hardware resources. Parameter-Efficient Fine-Tuning (PEFT) techniques, particularly adapter-based methods, have emerged as a promising alternative, enabling adaptation with minimal trainable parameters while preserving model performance. However, the design choices for adapters, including their architecture, placement, and size, are numerous, and it is unclear which of these combinations would produce optimal adapters and why, for a given case of low-resource music genre. In this paper, we attempt to answer this question by studying various adapter configurations for two AI music models, MusicGen and Mustango, on two genres: Hindustani Classical and Turkish Makam music. Our findings reveal distinct trade-offs: convolution-based adapters excel in capturing fine-grained local musical details such as ornamentations and short melodic phrases, while transformer-based adapters better preserve long-range dependencies crucial for structured improvisation. Additionally, we analyze computational resource requirements across different adapter scales, demonstrating how mid-sized adapters (40M parameters) achieve an optimal balance between expressivity and quality. Furthermore, we find that Mustango, a diffusion-based model, generates more diverse outputs with better adherence to the description in the input prompt while lacking in providing stability in notes, rhythm alignment, and aesthetics. Also, it is computationally intensive and requires significantly more time to train. In contrast, autoregressive models like MusicGen offer faster training and are more efficient, and can produce better quality output in comparison, but have slightly higher redundancy in their generations.
Authors: Kingsley Yeon, Steven B. Damelin
Abstract: It is a classical result in rational approximation theory that certain non-smooth or singular functions, such as $|x|$ and $x^{1/p}$, can be efficiently approximated using rational functions with root-exponential convergence in terms of degrees of freedom \cite{Sta, GN}. In contrast, polynomial approximations admit only algebraic convergence by Jackson's theorem \cite{Lub2}. Recent work shows that composite polynomial architectures can recover exponential approximation rates even without smoothness \cite{KY}. In this work, we introduce and analyze a class of weighted deep polynomial approximants tailored for functions with asymmetric behavior-growing unbounded on one side and decaying on the other. By multiplying a learnable deep polynomial with a one-sided weight, we capture both local non-smoothness and global growth. We show numerically that this framework outperforms Taylor, Chebyshev, and standard deep polynomial approximants, even when all use the same number of parameters. To optimize these approximants in practice, we propose a stable graph-based parameterization strategy building on \cite{Jar}.
Authors: Jiechen Chen, Bipin Rajendran, Osvaldo Simeone
Abstract: Neuromorphic and quantum computing have recently emerged as promising paradigms for advancing artificial intelligence, each offering complementary strengths. Neuromorphic systems built on spiking neurons excel at processing time-series data efficiently through sparse, event-driven computation, consuming energy only upon input events. Quantum computing, on the other hand, leverages superposition and entanglement to explore feature spaces that are exponentially large in the number of qubits. Hybrid approaches combining these paradigms have begun to show potential, but existing quantum spiking models have important limitations. Notably, prior quantum spiking neuron implementations rely on classical memory mechanisms on single qubits, requiring repeated measurements to estimate firing probabilities, and they use conventional backpropagation on classical simulators for training. Here we propose a stochastic quantum spiking (SQS) neuron model that addresses these challenges. The SQS neuron uses multi-qubit quantum circuits to realize a spiking unit with internal quantum memory, enabling event-driven probabilistic spike generation in a single shot. Furthermore, we outline how networks of SQS neurons -- dubbed SQS neural networks (SQSNNs) -- can be trained via a hardware-friendly local learning rule, eliminating the need for global classical backpropagation. The proposed SQSNN model fuses the time-series efficiency of neuromorphic computing with the exponentially large inner state space of quantum computing, paving the way for quantum spiking neural networks that are modular, scalable, and trainable on quantum hardware.
Authors: Maxime Leiber, Yosra Marnissi, Axel Barrau, Sylvain Meignen, Laurent Massouli\'e
Abstract: The short-time Fourier transform (STFT) is widely used for analyzing non-stationary signals. However, its performance is highly sensitive to its parameters, and manual or heuristic tuning often yields suboptimal results. To overcome this limitation, we propose a unified differentiable formulation of the STFT that enables gradient-based optimization of its parameters. This approach addresses the limitations of traditional STFT parameter tuning methods, which often rely on computationally intensive discrete searches. It enables fine-tuning of the time-frequency representation (TFR) based on any desired criterion. Moreover, our approach integrates seamlessly with neural networks, allowing joint optimization of the STFT parameters and network weights. The efficacy of the proposed differentiable STFT in enhancing TFRs and improving performance in downstream tasks is demonstrated through experiments on both simulated and real-world data.
Authors: Cyrus Addy, Ajay Kumar Gurumadaiah, Yixiang Gao, Kwame Awuah-Offei
Abstract: Underground mining operations face significant safety challenges that make emergency response capabilities crucial. While robots have shown promise in assisting with search and rescue operations, their effectiveness depends on reliable miner detection capabilities. Deep learning algorithms offer potential solutions for automated miner detection, but require comprehensive training datasets, which are currently lacking for underground mining environments. This paper presents a novel thermal imaging dataset specifically designed to enable the development and validation of miner detection systems for potential emergency applications. We systematically captured thermal imagery of various mining activities and scenarios to create a robust foundation for detection algorithms. To establish baseline performance metrics, we evaluated several state-of-the-art object detection algorithms including YOLOv8, YOLOv10, YOLO11, and RT-DETR on our dataset. While not exhaustive of all possible emergency situations, this dataset serves as a crucial first step toward developing reliable thermal-based miner detection systems that could eventually be deployed in real emergency scenarios. This work demonstrates the feasibility of using thermal imaging for miner detection and establishes a foundation for future research in this critical safety application.
Authors: Martin J. Wainwright
Abstract: We describe and analyze a computionally efficient refitting procedure for computing high-probability upper bounds on the instance-wise mean-squared prediction error of penalized nonparametric estimates based on least-squares minimization. Requiring only a single dataset and black box access to the prediction method, it consists of three steps: computing suitable residuals, symmetrizing and scaling them with a pre-factor $\rho$, and using them to define and solve a modified prediction problem recentered at the current estimate. We refer to it as wild refitting, since it uses Rademacher residual symmetrization as in a wild bootstrap variant. Under relatively mild conditions allowing for noise heterogeneity, we establish a high probability guarantee on its performance, showing that the wild refit with a suitably chosen wild noise scale $\rho$ gives an upper bound on prediction error. This theoretical analysis provides guidance into the design of such procedures, including how the residuals should be formed, the amount of noise rescaling in the wild sub-problem needed for upper bounds, and the local stability properties of the block-box procedure. We illustrate the applicability of this procedure to various problems, including non-rigid structure-from-motion recovery with structured matrix penalties; plug-and-play image restoration with deep neural network priors; and randomized sketching with kernel methods.
Authors: Anne Wu, Laurent Mazar\'e, Neil Zeghidour, Alexandre D\'efossez
Abstract: We propose a novel preference alignment framework for improving spoken dialogue models on real-time conversations from user interactions. Current preference learning methods primarily focus on text-based language models, and are not directly suited to the complexities of real-time speech interactions, with richer dynamics (e.g. interruption, interjection) and no explicit segmentation between speaker turns.We create a large-scale dataset of more than 150,000 preference pairs from raw multi-turn speech conversations, annotated with AI feedback, to cover preferences over both linguistic content and temporal context variations. We leverage offline alignment methods to finetune a full-duplex autoregressive speech-to-speech model. Extensive experiments demonstrate that feedback on generic conversations can be consistently effective in improving spoken dialogue models to produce more factual, safer and more contextually aligned interactions. We deploy the finetuned model and conduct holistic human evaluations to assess the impact beyond single-turn conversations. Our findings shed light on the importance of a well-calibrated balance among various dynamics, crucial for natural real-time speech dialogue systems.
Authors: Mohammad Shokrolah Shirazi, Hung-Fu Chang
Abstract: The turning movement count data is crucial for traffic signal design, intersection geometry planning, traffic flow, and congestion analysis. This work proposes three methods called dynamic, static, and hybrid configuration for TMC-based traffic signals. A vision-based tracking system is developed to estimate the TMC of six intersections in Las Vegas using traffic cameras. The intersection design, route (e.g. vehicle movement directions), and signal configuration files with compatible formats are synthesized and imported into Simulation of Urban MObility for signal evaluation with realistic data. The initial experimental results based on estimated waiting times indicate that the cycle time of 90 and 120 seconds works best for all intersections. In addition, four intersections show better performance for dynamic signal timing configuration, and the other two with lower performance have a lower ratio of total vehicle count to total lanes of the intersection leg. Since daily traffic flow often exhibits a bimodal pattern, we propose a hybrid signal method that switches between dynamic and static methods, adapting to peak and off-peak traffic conditions for improved flow management. So, a built-in traffic generator module creates vehicle routes for 4 hours, including peak hours, and a signal design module produces signal schedule cycles according to static, dynamic, and hybrid methods. Vehicle count distributions are weighted differently for each zone (i.e., West, North, East, South) to generate diverse traffic patterns. The extended experimental results for 6 intersections with 4 hours of simulation time imply that zone-based traffic pattern distributions affect signal design selection. Although the static method works great for evenly zone-based traffic distribution, the hybrid method works well for highly weighted traffic at intersection pairs of the West-East and North-South zones.
Authors: Tobias J. Riedlinger, Kira Maag, Hanno Gottschalk
Abstract: Deep neural networks have set the state-of-the-art in computer vision tasks such as bounding box detection and semantic segmentation. Object detectors and segmentation models assign confidence scores to predictions, reflecting the model's uncertainty in object detection or pixel-wise classification. However, these confidence estimates are often miscalibrated, as their architectures and loss functions are tailored to task performance rather than probabilistic foundation. Even with well calibrated predictions, object detectors fail to quantify uncertainty outside detected bounding boxes, i.e., the model does not make a probability assessment of whether an area without detected objects is truly free of obstacles. This poses a safety risk in applications such as automated driving, where uncertainty in empty areas remains unexplored. In this work, we propose an object detection model grounded in spatial statistics. Bounding box data matches realizations of a marked point process, commonly used to describe the probabilistic occurrence of spatial point events identified as bounding box centers, where marks are used to describe the spatial extension of bounding boxes and classes. Our statistical framework enables a likelihood-based training and provides well-defined confidence estimates for whether a region is drivable, i.e., free of objects. We demonstrate the effectiveness of our method through calibration assessments and evaluation of performance.
Authors: Marek \v{S}uppa, Andrej Ridzik, Daniel Hl\'adek, Tom\'a\v{s} Jav\r{u}rek, Vikt\'oria Ondrejov\'a, Krist\'ina S\'asikov\'a, Martin Tamajka, Mari\'an \v{S}imko
Abstract: In this work, we introduce skLEP, the first comprehensive benchmark specifically designed for evaluating Slovak natural language understanding (NLU) models. We have compiled skLEP to encompass nine diverse tasks that span token-level, sentence-pair, and document-level challenges, thereby offering a thorough assessment of model capabilities. To create this benchmark, we curated new, original datasets tailored for Slovak and meticulously translated established English NLU resources. Within this paper, we also present the first systematic and extensive evaluation of a wide array of Slovak-specific, multilingual, and English pre-trained language models using the skLEP tasks. Finally, we also release the complete benchmark data, an open-source toolkit facilitating both fine-tuning and evaluation of models, and a public leaderboard at https://github.com/slovak-nlp/sklep in the hopes of fostering reproducibility and drive future research in Slovak NLU.
Authors: Michalis K. Titsias, Angelos Alexopoulos, Siran Liu, Petros Dellaportas
Abstract: We develop sampling methods, which consist of Gaussian invariant versions of random walk Metropolis (RWM), Metropolis adjusted Langevin algorithm (MALA) and second order Hessian or Manifold MALA. Unlike standard RWM and MALA we show that Gaussian invariant sampling can lead to ergodic estimators with improved statistical efficiency. This is due to a remarkable property of Gaussian invariance that allows us to obtain exact analytical solutions to the Poisson equation for Gaussian targets. These solutions can be used to construct efficient and easy to use control variates for variance reduction of estimators under any intractable target. We demonstrate the new samplers and estimators in several examples, including high dimensional targets in latent Gaussian models where we compare against several advanced methods and obtain state-of-the-art results. We also provide theoretical results regarding geometric ergodicity, and an optimal scaling analysis that shows the dependence of the optimal acceptance rate on the Gaussianity of the target.
Authors: Mohammed Baharoon, Jun Ma, Congyu Fang, Augustin Toma, Bo Wang
Abstract: Multimodal Large Language Models (MLLMs) have emerged as a promising way to automate Radiology Report Generation (RRG). In this work, we systematically investigate the design space of 3D MLLMs, including visual input representation, projectors, Large Language Models (LLMs), and fine-tuning techniques for 3D CT report generation. We also introduce two knowledge-based report augmentation methods that improve performance on the GREEN score by up to 10\%, achieving the 2nd place on the MICCAI 2024 AMOS-MM challenge. Our results on the 1,687 cases from the AMOS-MM dataset show that RRG is largely independent of the size of LLM under the same training protocol. We also show that larger volume size does not always improve performance if the original ViT was pre-trained on a smaller volume size. Lastly, we show that using a segmentation mask along with the CT volume improves performance. The code is publicly available at https://github.com/bowang-lab/AMOS-MM-Solution
Authors: Hani Alomari, Anushka Sivakumar, Andrew Zhang, Chris Thomas
Abstract: Cross-modal image-text retrieval is challenging because of the diverse possible associations between content from different modalities. Traditional methods learn a single-vector embedding to represent semantics of each sample, but struggle to capture nuanced and diverse relationships that can exist across modalities. Set-based approaches, which represent each sample with multiple embeddings, offer a promising alternative, as they can capture richer and more diverse relationships. In this paper, we show that, despite their promise, these set-based representations continue to face issues including sparse supervision and set collapse, which limits their effectiveness. To address these challenges, we propose Maximal Pair Assignment Similarity to optimize one-to-one matching between embedding sets which preserve semantic diversity within the set. We also introduce two loss functions to further enhance the representations: Global Discriminative Loss to enhance distinction among embeddings, and Intra-Set Divergence Loss to prevent collapse within each set. Our method achieves state-of-the-art performance on MS-COCO and Flickr30k without relying on external data.
Authors: Xinzhuo Li, Adheesh Juvekar, Xingyou Liu, Muntasir Wahed, Kiet A. Nguyen, Ismini Lourentzou
Abstract: Recent progress in vision-language segmentation has significantly advanced grounded visual understanding. However, these models often exhibit hallucinations by producing segmentation masks for objects not grounded in the image content or by incorrectly labeling irrelevant regions. Existing evaluation protocols for segmentation hallucination primarily focus on label or textual hallucinations without manipulating the visual context, limiting their capacity to diagnose critical failures. In response, we introduce HalluSegBench, the first benchmark specifically designed to evaluate hallucinations in visual grounding through the lens of counterfactual visual reasoning. Our benchmark consists of a novel dataset of 1340 counterfactual instance pairs spanning 281 unique object classes, and a set of newly introduced metrics that quantify hallucination sensitivity under visually coherent scene edits. Experiments on HalluSegBench with state-of-the-art vision-language segmentation models reveal that vision-driven hallucinations are significantly more prevalent than label-driven ones, with models often persisting in false segmentation, highlighting the need for counterfactual reasoning to diagnose grounding fidelity.
Authors: Yutong Bai, Danny Tran, Amir Bar, Yann LeCun, Trevor Darrell, Jitendra Malik
Abstract: We train models to Predict Ego-centric Video from human Actions (PEVA), given the past video and an action represented by the relative 3D body pose. By conditioning on kinematic pose trajectories, structured by the joint hierarchy of the body, our model learns to simulate how physical human actions shape the environment from a first-person point of view. We train an auto-regressive conditional diffusion transformer on Nymeria, a large-scale dataset of real-world egocentric video and body pose capture. We further design a hierarchical evaluation protocol with increasingly challenging tasks, enabling a comprehensive analysis of the model's embodied prediction and control abilities. Our work represents an initial attempt to tackle the challenges of modeling complex real-world environments and embodied agent behaviors with video prediction from the perspective of a human.
Authors: Saurabh Kumar, Henrik Marklund, Ashish Rao, Yifan Zhu, Hong Jun Jeon, Yueyang Liu, Benjamin Van Roy
Abstract: An agent that efficiently accumulates knowledge to develop increasingly sophisticated skills over a long lifetime could advance the frontier of artificial intelligence capabilities. The design of such agents, which remains a long-standing challenge of artificial intelligence, is addressed by the subject of continual learning. This monograph clarifies and formalizes concepts of continual learning, introducing a framework and set of tools to stimulate further research.
Authors: Youssef Allouah, Rachid Guerraoui, John Stephan
Abstract: This position paper argues that achieving robustness, privacy, and efficiency simultaneously in machine learning systems is infeasible under prevailing threat models. The tension between these goals arises not from algorithmic shortcomings but from structural limitations imposed by worst-case adversarial assumptions. We advocate for a systematic research agenda aimed at formalizing the robustness-privacy-efficiency trilemma, exploring how principled relaxations of threat models can unlock better trade-offs, and designing benchmarks that expose rather than obscure the compromises made. By shifting focus from aspirational universal guarantees to context-aware system design, the machine learning community can build models that are truly appropriate for real-world deployment.
Authors: Liam Madden, Curtis Fox, Christos Thrampoulidis
Abstract: Given a sequence of tokens, such as words, the task of next-token prediction is to predict the next-token conditional probability distribution. Decoder-only transformers have become effective models for this task, but their properties are still not fully understood. In particular, the largest number of distinct context sequences that a decoder-only transformer can interpolate next-token distributions for has not been established. To fill this gap, we prove upper and lower bounds on this number, which are equal up to a multiplicative constant. We prove these bounds in the general setting where next-token distributions can be arbitrary as well as the empirical setting where they are calculated from a finite number of document sequences. Our lower bounds are for one-layer multi-head decoder-only transformers and our proofs highlight an important injectivity property satisfied by self-attention. Furthermore, we provide numerical evidence that the minimal number of parameters for memorization is sufficient for being able to train the model to the entropy lower bound.
Authors: Qing-xin Meng, Jian-wei Liu
Abstract: This paper focuses on the online saddle point problem, which involves a sequence of two-player time-varying convex-concave games. Considering the nonstationarity of the environment, we adopt the duality gap and the dynamic Nash equilibrium regret as performance metrics for algorithm design. We present three variants of the proximal point method: the Online Proximal Point Method (OPPM), the Optimistic OPPM (OptOPPM), and the OptOPPM with multiple predictors. Each algorithm guarantees upper bounds for both the duality gap and dynamic Nash equilibrium regret, achieving near-optimality when measured against the duality gap. Specifically, in certain benign environments, such as sequences of stationary payoff functions, these algorithms maintain a nearly constant metric bound. Experimental results further validate the effectiveness of these algorithms. Lastly, this paper discusses potential reliability concerns associated with using dynamic Nash equilibrium regret as a performance metric. The technical appendix and code can be found at https://github.com/qingxin6174/PPM-for-OSP.
Authors: Soumyajit Gupta, Maria De-Arteaga, Matthew Lease
Abstract: Target-group detection is the task of detecting which group(s) a social media post is ``directed at or about'', with various applications, such as targeted-marketing. In this work, we focus on the fairness implications of target-group detection in the context of toxicity detection, where the perceived harm of a post often depends on which group(s) it targets. Because toxicity is highly contextual, language that appears benign in general may be harmful when targeting specific demographic groups. It is thus important to first detect which group(s) are being {\em targeted} by a post as a precursor to the subsequent task of determining whether the post is toxic given the group(s). Target-group detection is also challenging: a single post may simultaneously target one to many groups, and we must detect groups fairly in order to promote equitable treatment. We show that our proposed approach to {\em fairness-aware multi target-group detection} not only reduces bias across groups, but also achieves competitive predictive performance, outperforming existing fairness-aware baselines. To spur future research on fairness-aware target-group detection and support competitive benchmarking, we also share our code.
Authors: Attila Lischka, Filip Rydin, Jiaming Wu, Morteza Haghir Chehreghani, Bal\'azs Kulcs\'ar
Abstract: In the last years, many learning-based approaches have been proposed to tackle combinatorial optimization problems such as routing problems. Many of these approaches are based on graph neural networks (GNNs) or related transformers, operating on the Euclidean coordinates representing the routing problems. However, models operating on Euclidean coordinates are ill-suited for non-Euclidean, asymmetric problem instances that are often found in real-world settings. To overcome this limitation, we propose a novel GNN-based and edge-focused neural model called Graph Edge Attention Network (GREAT). Using GREAT as an encoder to capture the properties of a routing problem instance, we build a reinforcement learning framework which we apply to Euclidean and non-Euclidean variants of vehicle routing problems such as Traveling Salesman Problem, Capacitated Vehicle Routing Problem and Orienteering Problem. Our framework is among the first to tackle non-Euclidean variants of these problems and achieves competitive results among learning-based solvers.
Authors: Yair Stolero, Itzik Klein
Abstract: Low-cost gyroscope calibration is essential for ensuring the accuracy and reliability of gyroscope measurements. Stationary calibration estimates the deterministic parts of measurement errors. To this end, a common practice is to average the gyroscope readings during a predefined period and estimate the gyroscope bias. Calibration duration plays a crucial role in performance, therefore, longer periods are preferred. However, some applications require quick startup times and calibration is therefore allowed only for a short time. In this work, we focus on reducing low-cost gyroscope calibration time using deep learning methods. We propose an end-to-end convolutional neural network for the application of gyroscope calibration. We explore the possibilities of using multiple real and virtual gyroscopes to improve the calibration performance of single gyroscopes. To train and validate our approach, we recorded a dataset consisting of 186.6 hours of gyroscope readings, using 36 gyroscopes of four different brands. We also created a virtual dataset consisting of simulated gyroscope readings. The six datasets were used to evaluate our proposed approach. One of our key achievements in this work is reducing gyroscope calibration time by up to 89% using three low-cost gyroscopes. Our dataset is publicly available to allow reproducibility of our work and to increase research in the field.
Authors: Ruiyang Hong, Anastasis Kratsios
Abstract: The foundations of deep learning are supported by the seemingly opposing perspectives of approximation or learning theory. The former advocates for large/expressive models that need not generalize, while the latter considers classes that generalize but may be too small/constrained to be universal approximators. Motivated by real-world deep learning implementations that are both expressive and statistically reliable, we ask: "Is there a class of neural networks that is both large enough to be universal but structured enough to generalize?" This paper constructively provides a positive answer to this question by identifying a highly structured class of ReLU multilayer perceptions (MLPs), which are optimal function approximators and are statistically well-behaved. We show that any $(L,\alpha)$-H\"{o}lder function from $[0,1]^d$ to $[-n,n]$ can be approximated to a uniform $\mathcal{O}(1/n)$ error on $[0,1]^d$ with a sparsely connected ReLU MLP with the same H\"{o}lder exponent $\alpha$ and coefficient $L$, of width $\mathcal{O}(dn^{d/\alpha})$, depth $\mathcal{O}(\log(d))$, with $\mathcal{O}(dn^{d/\alpha})$ nonzero parameters, and whose weights and biases take values in $\{0,\pm 1/2\}$ except in the first and last layers which instead have magnitude at-most $n$. Further, our class of MLPs achieves a near-optimal sample complexity of $\mathcal{O}(\log(N)/\sqrt{N})$ when given $N$ i.i.d. normalized sub-Gaussian training samples. We achieve this through a new construction that perfectly fits together linear pieces using Kuhn triangulations, along with a new proof technique which shows that our construction preserves the regularity of not only the H\"{o}lder functions, but also any uniformly continuous function. Our results imply that neural networks can solve the McShane extension problem on suitable finite sets.
Authors: Xinyu Zhou, Simin Fan, Martin Jaggi
Abstract: Influence functions provide a principled method to assess the contribution of individual training samples to a specific target. Yet, their high computational costs limit their applications on large-scale models and datasets. Existing methods proposed for influence function approximation have significantly reduced the computational overheads. However, they mostly suffer from inaccurate estimation due to the lack of strong convergence guarantees from the algorithm. The family of hyperpower methods are well-known for their rigorous convergence guarantees on matrix inverse approximation, while the matrix multiplication operation can involve intractable memory and computation costs on large-scale models. We propose HyperINF, an efficient and accurate influence function approximation method which leverages the hyperpower method, specifically Schulz's iterative algorithm. To deal with the computation-intensive matrix multiplication, we incorporate the generalized fisher information (GFIM) as a low-rank approximation of the Hessian matrix, which reduces the memory and computation overheads to constant costs independent of ranks on LoRA-tuned models. We first demonstrate the superior accuracy and stability of HyperINF compared to other baselines through a synthetic convergence simulation for matrix inversion. We further validate the efficacy of HyperINF through extensive real-world data attribution tasks, including mislabeled data detection and data selection for LLM and VLM fine-tuning. On LoRA-tuned models, HyperINF achieves superior downstream performance with minimal memory and computational overhead, while other baselines suffer from significant degradation. Our codebase is available at https://github.com/Blackzxy/HyperINF.
Authors: Aryo Lotfi, Enrico Fini, Samy Bengio, Moin Nabi, Emmanuel Abbe
Abstract: Modern vision models have achieved remarkable success in benchmarks where local features provide critical information about the target. There is now a growing interest in tackling tasks requiring more global reasoning, where local features do not provide significant information. Minsky and Papert put forward such tasks in 1969 with their connectivity study, exposing the limitations of the perceptron model. In this paper, we introduce an expanded set of global visual datasets involving graphs, strings, mazes, and image grids. We show that large vision models still struggle to learn these tasks efficiently. Similarly, state-of-the-art multi-modal LLMs perform poorly on these datasets. We explain this learning inefficiency by means of the 'globality degree' measure. To mitigate this, we propose a method called chain-of-sketch (CoS). Similar to the chain-of-thought and scratchpad techniques used in language models, CoS breaks the original task into intermediate visual steps to help learn a complex task. In addition, we show that not all CoS strategies perform equally well. Our key insight is to impose a Markovian structure on the CoS frames. This leads to the introduction of 'inductive CoS' which achieves better out-of-distribution generalization and performs well even with smaller models compared to non-inductive variants.
Authors: Ben Adcock, Michael Griebel, Gregor Maier
Abstract: Operator learning, the approximation of mappings between infinite-dimensional function spaces using machine learning, has gained increasing research attention in recent years. Approximate operators, learned from data, can serve as efficient surrogate models for problems in computational science and engineering, complementing traditional methods. However, despite their empirical success, our understanding of the underlying mathematical theory is in large part still incomplete. In this paper, we study the approximation of Lipschitz operators with respect to Gaussian measures. We prove higher Gaussian Sobolev regularity of Lipschitz operators and establish lower and upper bounds on the Hermite polynomial approximation error. We then study general reconstruction strategies of Lipschitz operators from $m$ arbitrary (potentially adaptive) linear samples. As a key finding, we tightly characterize the corresponding sample complexity, that is, the smallest achievable worst-case error among all possible choices of (adaptive) sampling and reconstruction strategies in terms of $m$. As a consequence, we identify an inherent curse of sample complexity: No method to approximate Lipschitz operators based on $m$ linear samples can achieve algebraic convergence rates in $m$. On the positive side, we prove that a sufficiently fast spectral decay of the covariance operator of the underlying Gaussian measure guarantees convergence rates which are arbitrarily close to any algebraic rate. Overall, by tightly characterizing the sample complexity, our work confirms the intrinsic difficulty of learning Lipschitz operators, regardless of the data or learning technique.
Authors: Mohammad Hossein Moslemi, Mostafa Milani
Abstract: Record matching is the task of identifying records that refer to the same real-world entity across datasets. While most existing models optimize for accuracy, fairness has become an important concern due to the potential for unequal outcomes across demographic groups. Prior work typically focuses on binary outcomes evaluated at fixed decision thresholds. However, such evaluations can miss biases in matching scores--biases that persist across thresholds and affect downstream tasks. We propose a threshold-independent framework for measuring and reducing score bias, defined as disparities in the distribution of matching scores across groups. We show that several state-of-the-art matching methods exhibit substantial score bias, even when appearing fair under standard threshold-based metrics. To address this, we introduce two post-processing score calibration algorithms. The first, calib, aligns group-wise score distributions using the Wasserstein barycenter, targeting demographic parity. The second, ccalib, conditions on predicted labels to further reduce label-dependent biases, such as equal opportunity. Both methods are model-agnostic and require no access to model training data. calib also offers theoretical guarantees, ensuring reduced bias with minimal deviation from original scores. Experiments across real-world datasets and matching models confirm that calib and ccalib substantially reduce score bias while minimally impacting model accuracy.
Authors: Advik Raj Basani, Xiao Zhang
Abstract: LLMs have shown impressive capabilities across various natural language processing tasks, yet remain vulnerable to input prompts, known as jailbreak attacks, carefully designed to bypass safety guardrails and elicit harmful responses. Traditional methods rely on manual heuristics but suffer from limited generalizability. Despite being automatic, optimization-based attacks often produce unnatural prompts that can be easily detected by safety filters or require high computational costs due to discrete token optimization. In this paper, we introduce Generative Adversarial Suffix Prompter (GASP), a novel automated framework that can efficiently generate human-readable jailbreak prompts in a fully black-box setting. In particular, GASP leverages latent Bayesian optimization to craft adversarial suffixes by efficiently exploring continuous latent embedding spaces, gradually optimizing the suffix prompter to improve attack efficacy while balancing prompt coherence via a targeted iterative refinement procedure. Through comprehensive experiments, we show that GASP can produce natural adversarial prompts, significantly improving jailbreak success over baselines, reducing training times, and accelerating inference speed, thus making it an efficient and scalable solution for red-teaming LLMs.
Authors: Peiyan Hu, Rui Wang, Xiang Zheng, Tao Zhang, Haodong Feng, Ruiqi Feng, Long Wei, Yue Wang, Zhi-Ming Ma, Tailin Wu
Abstract: Simulating and controlling physical systems described by partial differential equations (PDEs) are crucial tasks across science and engineering. Recently, diffusion generative models have emerged as a competitive class of methods for these tasks due to their ability to capture long-term dependencies and model high-dimensional states. However, diffusion models typically struggle with handling system states with abrupt changes and generalizing to higher resolutions. In this work, we propose Wavelet Diffusion Neural Operator (WDNO), a novel PDE simulation and control framework that enhances the handling of these complexities. WDNO comprises two key innovations. Firstly, WDNO performs diffusion-based generative modeling in the wavelet domain for the entire trajectory to handle abrupt changes and long-term dependencies effectively. Secondly, to address the issue of poor generalization across different resolutions, which is one of the fundamental tasks in modeling physical systems, we introduce multi-resolution training. We validate WDNO on five physical systems, including 1D advection equation, three challenging physical systems with abrupt changes (1D Burgers' equation, 1D compressible Navier-Stokes equation and 2D incompressible fluid), and a real-world dataset ERA5, which demonstrates superior performance on both simulation and control tasks over state-of-the-art methods, with significant improvements in long-term and detail prediction accuracy. Remarkably, in the challenging context of the 2D high-dimensional and indirect control task aimed at reducing smoke leakage, WDNO reduces the leakage by 78% compared to the second-best baseline. The code can be found at https://github.com/AI4Science-WestlakeU/wdno.git.
Authors: Wan Jiang, He Wang, Xin Zhang, Dan Guo, Zhaoxin Fan, Yunfeng Diao, Richang Hong
Abstract: Score-based Generative Models (SGMs) have demonstrated remarkable generalization abilities, e.g. generating unseen, but natural data. However, the greater the generalization power, the more likely the unintended generalization, and the more dangerous the abuse. Research on moderated generalization in SGMs remains limited. To fill this gap, we first examine the current 'gold standard' in Machine Unlearning (MU), i.e., re-training the model after removing the undesirable training data, and find it does not work in SGMs. Further analysis of score functions reveals that the MU 'gold standard' does not alter the original score function, which explains its ineffectiveness. Based on this insight, we propose the first Moderated Score-based Generative Model (MSGM), which introduces a novel score adjustment strategy that redirects the score function away from undesirable data during the continuous-time stochastic differential equation process. Extensive experimental results demonstrate that MSGM significantly reduces the likelihood of generating undesirable content while preserving high visual quality for normal image generation. Albeit designed for SGMs, MSGM is a general and flexible MU framework that is compatible with diverse diffusion architectures (SGM and DDPM) and training strategies (re-training and fine-tuning), and enables zero-shot transfer of the pre-trained models to downstream tasks, e.g. image inpainting and reconstruction. The code will be shared upon acceptance.
Authors: Andy Yang, Lena Strobl, David Chiang, Dana Angluin
Abstract: We study conditions under which transformers using soft attention can simulate hard attention, that is, effectively focus all attention on a subset of positions. First, we examine several subclasses of languages recognized by hard-attention transformers, which can be defined in variants of linear temporal logic. We demonstrate how soft-attention transformers can compute formulas of these logics using unbounded positional embeddings or temperature scaling. Second, we demonstrate how temperature scaling allows softmax transformers to simulate general hard-attention transformers, using a temperature that depends on the minimum gap between the maximum attention scores and other attention scores.
Authors: Andrea Perin, Stephane Deny
Abstract: Symmetries (transformations by group actions) are present in many datasets, and leveraging them holds considerable promise for improving predictions in machine learning. In this work, we aim to understand when and how deep networks -- with standard architectures trained in a standard, supervised way -- learn symmetries from data. Inspired by real-world scenarios, we study a classification paradigm where data symmetries are only partially observed during training: some classes include all transformations of a cyclic group, while others -- only a subset. In the infinite-width limit, where kernel analogies apply, we derive a neural kernel theory of symmetry learning. The group-cyclic nature of the dataset allows us to analyze the Gram matrix of neural kernels in the Fourier domain; here we find a simple characterization of the generalization error as a function of class separation (signal) and class-orbit density (noise). This characterization reveals that generalization can only be successful when the local structure of the data prevails over its non-local, symmetry-induced structure, in the kernel space defined by the architecture. We extend our theoretical treatment to any finite group, including non-abelian groups. Our framework also applies to equivariant architectures (e.g., CNNs), and recovers their success in the special case where the architecture matches the inherent symmetry of the data. Empirically, our theory reproduces the generalization failure of finite-width networks (MLP, CNN, ViT) trained on partially observed versions of rotated-MNIST. We conclude that conventional deep networks lack a mechanism to learn symmetries that have not been explicitly embedded in their architecture a priori. Our framework could be extended to guide the design of architectures and training procedures able to learn symmetries from data.
Authors: Konstantin Avrachenkov, Vivek S. Borkar, Pratik Shah
Abstract: We study the Lagrange Index Policy (LIP) for restless multi-armed bandits with long-run average reward. In particular, we compare the performance of LIP with the performance of the Whittle Index Policy (WIP), both heuristic policies known to be asymptotically optimal under certain natural conditions. Even though in most cases their performances are very similar, in the cases when WIP shows bad performance, LIP continues to perform very well. We then propose reinforcement learning algorithms, both tabular and NN-based, to obtain online learning schemes for LIP in the model-free setting. The proposed reinforcement learning schemes for LIP require significantly less memory than the analogous schemes for WIP. We calculate analytically the Lagrange index for the restart model, which applies to the optimal web crawling and the minimization of the weighted age of information. We also give a new proof of asymptotic optimality in case of homogeneous arms as the number of arms goes to infinity, based on exchangeability and de Finetti's theorem.
Authors: David Restrepo, Chenwei Wu, Yueran Jia, Jaden K. Sun, Jack Gallifant, Catherine G. Bielick, Yugang Jia, Leo A. Celi
Abstract: Accurate imputation of missing laboratory values in electronic health records (EHRs) is critical to enable robust clinical predictions and reduce biases in AI systems in healthcare. Existing methods, such as XGBoost, softimpute, GAIN, Expectation Maximization (EM), and MICE, struggle to model the complex temporal and contextual dependencies in EHR data, particularly in underrepresented groups. In this work, we propose Lab-MAE, a novel transformer-based masked autoencoder framework that leverages self-supervised learning for the imputation of continuous sequential lab values. Lab-MAE introduces a structured encoding scheme that jointly models laboratory test values and their corresponding timestamps, enabling explicit capturing temporal dependencies. Empirical evaluation on the MIMIC-IV dataset demonstrates that Lab-MAE significantly outperforms state-of-the-art baselines such as XGBoost, softimpute, GAIN, EM, and MICE across multiple metrics, including root mean square error (RMSE), R-squared (R2), and Wasserstein distance (WD). Notably, Lab-MAE achieves equitable performance across demographic groups of patients, advancing fairness in clinical predictions. We further investigate the role of follow-up laboratory values as potential shortcut features, revealing Lab-MAE's robustness in scenarios where such data is unavailable. The findings suggest that our transformer-based architecture, adapted to the characteristics of EHR data, offers a foundation model for more accurate and fair clinical imputation. In addition, we measure and compare the carbon footprint of Lab-MAE with the a XGBoost model, highlighting its environmental requirements.
Authors: Kutay B\"olat, Simon Tindemans
Abstract: Probabilistic forecasting in power systems often involves multi-entity datasets like households, feeders, and wind turbines, where generating reliable entity-specific forecasts presents significant challenges. Traditional approaches require training individual models for each entity, making them inefficient and hard to scale. This study addresses this problem using GUIDE-VAE, a conditional variational autoencoder that allows entity-specific probabilistic forecasting using a single model. GUIDE-VAE provides flexible outputs, ranging from interpretable point estimates to full probability distributions, thanks to its advanced covariance composition structure. These distributions capture uncertainty and temporal dependencies, offering richer insights than traditional methods. To evaluate our GUIDE-VAE-based forecaster, we use household electricity consumption data as a case study due to its multi-entity and highly stochastic nature. Experimental results demonstrate that GUIDE-VAE outperforms conventional quantile regression techniques across key metrics while ensuring scalability and versatility. These features make GUIDE-VAE a powerful and generalizable tool for probabilistic forecasting tasks, with potential applications beyond household electricity consumption.
Authors: Hao Zhu, Yuan Zhang, Joschka Boedecker
Abstract: We consider the inverse reinforcement learning (IRL) problem, where an unknown reward function of some Markov decision process is estimated based on observed expert demonstrations. In most existing approaches, IRL is formulated and solved as a nonconvex optimization problem, posing challenges in scenarios where robustness and reproducibility are critical. We discuss a convex formulation of the IRL problem (CIRL) initially proposed by Ng and Russel, and reformulate the problem such that the domain-specific language CVXPY can be applied directly to specify and solve the convex problem. We also extend the CIRL problem to scenarios where the expert policy is not given analytically but by trajectory as state-action pairs, which can be strongly inconsistent with optimality, by augmenting some of the constraints. Theoretical analysis and practical implementation for hyperparameter auto-selection are introduced. This note helps the users to easily apply CIRL for their problems, without background knowledge on convex optimization.
Authors: Qingchuan Lyu
Abstract: This paper proposes Genetic Algorithm with Border Trades (GAB), a novel modification of the standard genetic algorithm that enhances exploration by incorporating new chromosome patterns in the breeding process. This approach significantly mitigates premature convergence and improves search diversity. Empirically, GAB achieves up to 8x higher fitness and 10x faster convergence on complex job scheduling problems compared to standard Genetic Algorithms, reaching average fitness scores of 888 versus 106 in under 20 seconds. On the classic Flip-Flop problem, GAB consistently finds optimal or near-optimal solutions in fewer generations, even as input sizes scale to thousands of bits. These results highlight GAB as a highly effective and computationally efficient alternative for solving large-scale combinatorial optimization problems.
Authors: Michalis K. Titsias
Abstract: Sparse variational Gaussian processes (GPs) construct tractable posterior approximations to GP models. At the core of these methods is the assumption that the true posterior distribution over training function values ${\bf f}$ and inducing variables ${\bf u}$ is approximated by a variational distribution that incorporates the conditional GP prior $p({\bf f} | {\bf u})$ in its factorization. While this assumption is considered as fundamental, we show that for model training we can relax it through the use of a more general variational distribution $q({\bf f} | {\bf u})$ that depends on $N$ extra parameters, where $N$ is the number of training examples. In GP regression, we can analytically optimize the evidence lower bound over the extra parameters and express a tractable collapsed bound that is tighter than the previous bound. The new bound is also amenable to stochastic optimization and its implementation requires minor modifications to existing sparse GP code. Further, we also describe extensions to non-Gaussian likelihoods. On several datasets we demonstrate that our method can reduce bias when learning the hyperparameters and can lead to better predictive performance.
Authors: Zeyi Ren, Qingfeng Lin, Jingreng Lei, Yang Li, Yik-Chung Wu
Abstract: In the realm of activity detection for massive machine-type communications, intelligent reflecting surfaces (IRS) have shown significant potential in enhancing coverage for devices lacking direct connections to the base station (BS). However, traditional activity detection methods are typically designed for a single type of channel model, which does not reflect the complexities of real-world scenarios, particularly in systems incorporating IRS. To address this challenge, this paper introduces a novel approach that combines model-driven deep unfolding with a mixture of experts (MoE) framework. By automatically selecting one of three expert designs and applying it to the unfolded projected gradient method, our approach eliminates the need for prior knowledge of channel types between devices and the BS. Simulation results demonstrate that the proposed MoE-augmented deep unfolding method surpasses the traditional covariance-based method and black-box neural network design, delivering superior detection performance under mixed channel fading conditions.
Authors: Vicente Balmaseda, Bokun Wang, Ching-Long Lin, Tianbao Yang
Abstract: In self-supervised contrastive learning, negative pairs are typically constructed using an anchor image and a sample drawn from the entire dataset, excluding the anchor. However, this approach can result in the creation of negative pairs with similar semantics, referred to as "false negatives", leading to their embeddings being falsely pushed apart. To address this issue, we introduce GloFND, an optimization-based approach that automatically learns on the fly the threshold for each anchor data to identify its false negatives during training. In contrast to previous methods for false negative discovery, our approach globally detects false negatives across the entire dataset rather than locally within the mini-batch. Moreover, its per-iteration computation cost remains independent of the dataset size. Experimental results on image and image-text data demonstrate the effectiveness of the proposed method. Our implementation is available at https://github.com/vibalcam/GloFND.
Authors: Ivan Saboli\'c, Matej Grci\'c, Sini\v{s}a \v{S}egvi\'c
Abstract: We propose VIBE, a model-agnostic framework that trains classifiers resilient to backdoor attacks. The key concept behind our approach is to treat malicious inputs and corrupted labels from the training dataset as observed random variables, while the actual clean labels are latent. VIBE then recovers the corresponding latent clean label posterior through variational inference. The resulting training procedure follows the expectation-maximization (EM) algorithm. The E-step infers the clean pseudolabels by solving an entropy-regularized optimal transport problem, while the M-step updates the classifier parameters via gradient descent. Being modular, VIBE can seamlessly integrate with recent advancements in self-supervised representation learning, which enhance its ability to resist backdoor attacks. We experimentally validate the method effectiveness against contemporary backdoor attacks on standard datasets, a large-scale setup with 1$k$ classes, and a dataset poisoned with multiple attacks. VIBE consistently outperforms previous defenses across all tested scenarios.
Authors: Hojjat Azimi Asrari, Megan A. K. Peters
Abstract: Studies often aim to reveal ``first-order" representations (FORs), which encode aspects of an observer's environment, such as contents or structure. A less-common target is ``higher-order" representations (HORs), which are ``about" FORs -- e.g., their strength or uncertainty -- and which may contribute to learning. HORs about uncertainty are unlikely to be direct ``read-outs" of FOR characteristics, instead reflecting noisy estimation processes incorporating prior expectations about uncertainty, but how the brain represents such expected uncertainty distributions remains largely unexplored. Here, we study ``noise expectation" HORs using neural data from a task which may require the brain to learn about its own noise: decoded neurofeedback, wherein human subjects learn to volitionally produce target neural patterns. We develop and apply a Noise Estimation through Reinforcement-based Diffusion (NERD) model to characterize how brains may undertake this process, and show that NERD offers high explanatory power for human behavior.
Authors: Alexander Ryabchenko, Idan Attias, Daniel M. Roy
Abstract: We study online learning with oblivious losses and delays under a novel ``capacity constraint'' that limits how many past rounds can be tracked simultaneously for delayed feedback. Under ``clairvoyance'' (i.e., delay durations are revealed upfront each round) and/or ``preemptibility'' (i.e., we can stop tracking previously chosen round feedback), we establish matching upper and lower bounds (up to logarithmic terms) on achievable regret, characterizing the ``optimal capacity'' needed to match the minimax rates of classical delayed online learning, which implicitly assume unlimited capacity. Our algorithms achieve minimax-optimal regret across all capacity levels, with performance gracefully degrading under suboptimal capacity. For $K$ actions and total delay $D$ over $T$ rounds, under clairvoyance and assuming capacity $C = \Omega(\log(T))$, we achieve regret $\widetilde{\Theta}(\sqrt{TK + DK/C + D\log(K)})$ for bandits and $\widetilde{\Theta}(\sqrt{(D+T)\log(K)})$ for full-information feedback. When replacing clairvoyance with preemptibility, we require a known maximum delay bound $d_{\max}$, adding ${\widetilde{O}(d_{\max})}$ to the regret. For fixed delays $d$ (i.e., $D=Td$), the minimax regret is $\Theta(\sqrt{TK(1+d/C)+Td\log(K)})$ and the optimal capacity is $\Theta(\min\{K/\log(K),d\})$ in the bandit setting, while in the full-information feedback setting, the minimax regret is $\Theta(\sqrt{T(d+1)\log(K)})$ and the optimal capacity is $\Theta(1)$. For round-dependent and fixed delays, our upper bounds are achieved using novel preemptive and non-preemptive scheduling policies, based on Pareto-distributed proxy delays, and batching techniques, respectively. Crucially, our work unifies delayed bandits, label-efficient learning, and online scheduling frameworks, demonstrating that robust online learning under delayed feedback is possible with surprisingly modest tracking capacity.
Authors: Chongjie Si, Zhiyi Shi, Xuehui Wang, Yichen Xiao, Xiaokang Yang, Wei Shen
Abstract: Adapting pre-trained foundation models for diverse downstream tasks is a core practice in artificial intelligence. However, the wide range of tasks and high computational costs make full fine-tuning impractical. To overcome this, parameter-efficient fine-tuning (PEFT) methods like LoRA have emerged and are becoming a growing research focus. Despite the success of these methods, they are primarily designed for linear layers, focusing on two-dimensional matrices while largely ignoring higher-dimensional parameter spaces like convolutional kernels. Moreover, directly applying these methods to higher-dimensional parameter spaces often disrupts their structural relationships. Given the rapid advancements in matrix-based PEFT methods, rather than designing a specialized strategy, we propose a generalization that extends matrix-based PEFT methods to higher-dimensional parameter spaces without compromising their structural properties. Specifically, we treat parameters as elements of a Lie group, with updates modeled as perturbations in the corresponding Lie algebra. These perturbations are mapped back to the Lie group through the exponential map, ensuring smooth, consistent updates that preserve the inherent structure of the parameter space. Extensive experiments on computer vision and natural language processing validate the effectiveness and versatility of our approach, demonstrating clear improvements over existing methods.
Authors: Alexander Appel, Vanessa Kosoy
Abstract: We propose a framework which generalizes "decision making with structured observations" by allowing robust (i.e. multivalued) models. In this framework, each model associates each decision with a convex set of probability distributions over outcomes. Nature can choose distributions out of this set in an arbitrary (adversarial) manner, that can be nonoblivious and depend on past history. The resulting framework offers much greater generality than classical bandits and reinforcement learning, since the realizability assumption becomes much weaker and more realistic. We then derive a theory of regret bounds for this framework. Although our lower and upper bounds are not tight, they are sufficient to fully characterize power-law learnability. We demonstrate this theory in two special cases: robust linear bandits and tabular robust online reinforcement learning. In both cases, we derive regret bounds that improve state-of-the-art (except that we do not address computational efficiency).
Authors: Michal Balcerak, Tamaz Amiranashvili, Antonio Terpin, Suprosanna Shit, Lea Bogensperger, Sebastian Kaltenbach, Petros Koumoutsakos, Bjoern Menze
Abstract: The most widely used generative models map noise and data distributions by matching flows or scores. However, they struggle to incorporate partial observations and additional priors--something energy-based models (EBMs) handle elegantly by simply adding corresponding scalar energy terms. We address this issue by proposing Energy Matching, a framework that endows flow-based approaches with the flexibility of EBMs. Far from the data manifold, samples move along curl-free, optimal transport paths from noise to data. As they approach the data manifold, an entropic energy term guides the system into a Boltzmann equilibrium distribution, explicitly capturing the underlying likelihood structure of the data. We parameterize this dynamic with a single time-independent scalar field, which serves as both a powerful generator and a flexible prior for effective regularization of inverse problems. Our method substantially outperforms existing EBMs on CIFAR-10 and ImageNet generation in terms of fidelity, while retaining simulation-free training of transport-based approaches away from the data manifold. Furthermore, we leverage the method's flexibility to introduce an interaction energy that supports diverse mode exploration, which we demonstrate in a controlled protein-generation setting. Our approach focuses on learning a scalar potential energy--without time-conditioning, auxiliary generators, or additional networks--which marks a significant departure from recent EBM methods. We believe that this simplified framework significantly advances EBMs capabilities and paves the way for their wider adoption in generative modeling across diverse domains.
Authors: Xiang Li, Jianpeng Qi, Haobing Liu, Yuan Cao, Guoqing Chao, Zhongying Zhao, Junyu Dong, Yanwei Yu
Abstract: Graph Neural Networks (GNNs) have demonstrated impressive performance across diverse graph-based tasks by leveraging message passing to capture complex node relationships. However, when applied to large-scale real-world graphs, GNNs face two major challenges: First, it becomes increasingly difficult to ensure both scalability and efficiency, as the repeated aggregation of large neighborhoods leads to significant computational overhead; Second, the over-smoothing problem arises, where excessive or deep propagation makes node representations indistinguishable, severely hindering model expressiveness. To tackle these issues, we propose ScaleGNN, a novel framework that adaptively fuses multi-hop node features for both scalable and effective graph learning. First, we construct per-hop pure neighbor matrices that capture only the exclusive structural information at each hop, avoiding the redundancy of conventional aggregation. Then, an enhanced feature fusion strategy significantly balances low-order and high-order information, preserving both local detail and global correlations without incurring excessive complexity. To further reduce redundancy and over-smoothing, we introduce a Local Contribution Score (LCS)-based masking mechanism to filter out less relevant high-order neighbors, ensuring that only the most meaningful information is aggregated. In addition, learnable sparse constraints selectively integrate multi-hop valuable features, emphasizing the most informative high-order neighbors. Extensive experiments on real-world datasets demonstrate that ScaleGNN consistently outperforms state-of-the-art GNNs in both predictive accuracy and computational efficiency, highlighting its practical value for large-scale graph learning.
Authors: Sanjeev Raja, Martin \v{S}\'ipka, Michael Psenka, Tobias Kreiman, Michal Pavelka, Aditi S. Krishnapriyan
Abstract: Transition path sampling (TPS), which involves finding probable paths connecting two points on an energy landscape, remains a challenge due to the complexity of real-world atomistic systems. Current machine learning approaches use expensive, task-specific, and data-free training procedures, limiting their ability to benefit from high-quality datasets and large-scale pre-trained models. In this work, we address TPS by interpreting candidate paths as trajectories sampled from stochastic dynamics induced by the learned score function of pre-trained generative models, specifically denoising diffusion and flow matching. Under these dynamics, finding high-likelihood transition paths becomes equivalent to minimizing the Onsager-Machlup (OM) action functional. This enables us to repurpose pre-trained generative models for TPS in a zero-shot manner, in contrast with bespoke, task-specific approaches in previous work. We demonstrate our approach on varied molecular systems, obtaining diverse, physically realistic transition pathways and generalizing beyond the pre-trained model's original training dataset. Our method can be easily incorporated into new generative models, making it practically relevant as models continue to scale and improve with increased data availability. Code is available at github.com/ASK-Berkeley/OM-TPS.
Authors: Yiping Ji, Hemanth Saratchandran, Peyman Moghadam, Simon Lucey
Abstract: We highlight a curious empirical result within modern Vision Transformers (ViTs). Specifically, self-attention catastrophically fails to train unless it is used in conjunction with a skip connection. This is in contrast to other elements of a ViT that continue to exhibit good performance (albeit suboptimal) when skip connections are removed. Further, we show that this critical dependence on skip connections is a relatively new phenomenon, with previous deep architectures (\eg, CNNs) exhibiting good performance in their absence. In this paper, we theoretically characterize that the self-attention mechanism is fundamentally ill-conditioned and is, therefore, uniquely dependent on skip connections for regularization. Additionally, we propose Token Graying -- a simple yet effective complement (to skip connections) that further improves the condition of input tokens. We validate our approach in both supervised and self-supervised training methods.
Authors: Lei Lei (Sherman), Kan Zheng (Sherman), Xuemin (Sherman), Shen
Abstract: As Cellular Vehicle-to-Everything (C-V2X) evolves towards future sixth-generation (6G) networks, Connected Autonomous Vehicles (CAVs) are emerging to become a key application. Leveraging data-driven Machine Learning (ML), especially Deep Reinforcement Learning (DRL), is expected to significantly enhance CAV decision-making in both vehicle control and V2X communication under uncertainty. These two decision-making processes are closely intertwined, with the value of information (VoI) acting as a crucial bridge between them. In this paper, we introduce Sequential Stochastic Decision Process (SSDP) models to define and assess VoI, demonstrating their application in optimizing communication systems for CAVs. Specifically, we formally define the SSDP model and demonstrate that the MDP model is a special case of it. The SSDP model offers a key advantage by explicitly representing the set of information that can enhance decision-making when available. Furthermore, as current research on VoI remains fragmented, we propose a systematic VoI modeling framework grounded in the MDP, Reinforcement Learning (RL) and Optimal Control theories. We define different categories of VoI and discuss their corresponding estimation methods. Finally, we present a structured approach to leverage the various VoI metrics for optimizing the ``When", ``What", and ``How" to communicate problems. For this purpose, SSDP models are formulated with VoI-associated reward functions derived from VoI-based optimization objectives. While we use a simple vehicle-following control problem to illustrate the proposed methodology, it holds significant potential to facilitate the joint optimization of stochastic, sequential control and communication decisions in a wide range of networked control systems.
Authors: Rishi Jha, Collin Zhang, Vitaly Shmatikov, John X. Morris
Abstract: We introduce the first method for translating text embeddings from one vector space to another without any paired data, encoders, or predefined sets of matches. Our unsupervised approach translates any embedding to and from a universal latent representation (i.e., a universal semantic structure conjectured by the Platonic Representation Hypothesis). Our translations achieve high cosine similarity across model pairs with different architectures, parameter counts, and training datasets. The ability to translate unknown embeddings into a different space while preserving their geometry has serious implications for the security of vector databases. An adversary with access only to embedding vectors can extract sensitive information about the underlying documents, sufficient for classification and attribute inference.
Authors: Lingkai Kong, Haichuan Wang, Tonghan Wang, Guojun Xiong, Milind Tambe
Abstract: Incorporating pre-collected offline data from a source environment can significantly improve the sample efficiency of reinforcement learning (RL), but this benefit is often challenged by discrepancies between the transition dynamics of the source and target environments. Existing methods typically address this issue by penalizing or filtering out source transitions in high dynamics-gap regions. However, their estimation of the dynamics gap often relies on KL divergence or mutual information, which can be ill-defined when the source and target dynamics have disjoint support. To overcome these limitations, we propose CompFlow, a method grounded in the theoretical connection between flow matching and optimal transport. Specifically, we model the target dynamics as a conditional flow built upon the output distribution of the source-domain flow, rather than learning it directly from a Gaussian prior. This composite structure offers two key advantages: (1) improved generalization for learning target dynamics, and (2) a principled estimation of the dynamics gap via the Wasserstein distance between source and target transitions. Leveraging our principled estimation of the dynamics gap, we further introduce an optimistic active data collection strategy that prioritizes exploration in regions of high dynamics gap, and theoretically prove that it reduces the performance disparity with the optimal policy. Empirically, CompFlow outperforms strong baselines across several RL benchmarks with shifted dynamics.
Authors: Yuxuan Yue, Zukang Xu, Zhihang Yuan, Dawei Yang, Jianlong Wu, Liqiang Nie
Abstract: Large Language Models (LLMs) face significant challenges in edge deployment due to their massive parameter scale. Vector Quantization (VQ), a clustering-based quantization method, serves as a prevalent solution to this issue for its extremely low-bit (even at 2-bit) and considerable accuracy. Since a vector is a quantity in mathematics and physics that has both direction and magnitude, existing VQ works typically quantize them in a coupled manner. However, we find that direction exhibits significantly greater sensitivity to quantization compared to the magnitude. For instance, when separately clustering the directions and magnitudes of weight vectors in LLaMA-2-7B, the accuracy drop of zero-shot tasks are 46.5\% and 2.3\%, respectively. This gap even increases with the reduction of clustering centers. Further, Euclidean distance, a common metric to access vector similarities in current VQ works, places greater emphasis on reducing the magnitude error. This property is contrary to the above finding, unavoidably leading to larger quantization errors. To these ends, this paper proposes Polar Coordinate Decoupled Vector Quantization (PCDVQ), an effective and efficient VQ framework consisting of two key modules: 1) Polar Coordinate Decoupling (PCD), which transforms vectors into their polar coordinate representations and perform independent quantization of the direction and magnitude parameters.2) Distribution Aligned Codebook Construction (DACC), which optimizes the direction and magnitude codebooks in accordance with the source distribution. Experimental results show that PCDVQ outperforms baseline methods at 2-bit level by at least 1.5\% zero-shot accuracy, establishing a novel paradigm for accurate and highly compressed LLMs.
Authors: Xiucheng Wang, Honggang Jia, Nan Cheng, Dusit Niyato
Abstract: In this paper, a novel semantic communication framework empowered by generative artificial intelligence (GAI) is proposed, to enhance the robustness against both channel noise and transmission data distribution shifts. A theoretical foundation is established using stochastic differential equations (SDEs), from which a closed-form mapping between any signal-to-noise ratio (SNR) and the optimal denoising timestep is derived. Moreover, to address distribution mismatch, a mathematical scaling method is introduced to align received semantic features with the training distribution of the GAI. Built on this theoretical foundation, a latent diffusion model (LDM)-based semantic communication framework is proposed that combines a variational autoencoder for semantic features extraction, where a pretrained diffusion model is used for denoising. The proposed system is a training-free framework that supports zero-shot generalization, and achieves superior performance under low-SNR and out-of-distribution conditions, offering a scalable and robust solution for future 6G semantic communication systems. Experimental results demonstrate that the proposed semantic communication framework achieves state-of-the-art performance in both pixel-level accuracy and semantic perceptual quality, consistently outperforming baselines across a wide range of SNRs and data distributions without any fine-tuning or post-training.
Authors: Ziwen Wang, Jiajun Fan, Thao Nguyen, Heng Ji, Ge Liu
Abstract: Contrastive learning has proven to be highly efficient and adaptable in shaping representation spaces across diverse modalities by pulling similar samples together and pushing dissimilar ones apart. However, two key limitations persist: (1) Without explicit regulation of the embedding distribution, semantically related instances can inadvertently be pushed apart unless complementary signals guide pair selection, and (2) excessive reliance on large in-batch negatives and tailored augmentations hinders generalization. To address these limitations, we propose Variational Supervised Contrastive Learning (VarCon), which reformulates supervised contrastive learning as variational inference over latent class variables and maximizes a posterior-weighted evidence lower bound (ELBO) that replaces exhaustive pair-wise comparisons for efficient class-aware matching and grants fine-grained control over intra-class dispersion in the embedding space. Trained exclusively on image data, our experiments on CIFAR-10, CIFAR-100, ImageNet-100, and ImageNet-1K show that VarCon (1) achieves state-of-the-art performance for contrastive learning frameworks, reaching 79.36% Top-1 accuracy on ImageNet-1K and 78.29% on CIFAR-100 with a ResNet-50 encoder while converging in just 200 epochs; (2) yields substantially clearer decision boundaries and semantic organization in the embedding space, as evidenced by KNN classification, hierarchical clustering results, and transfer-learning assessments; and (3) demonstrates superior performance in few-shot learning than supervised baseline and superior robustness across various augmentation strategies.
Authors: Longzhu He, Chaozhuo Li, Peng Tang, Li Sun, Sen Su, Philip S. Yu
Abstract: Graph neural networks (GNNs) have achieved significant success in graph representation learning and have been applied to various domains. However, many real-world graphs contain sensitive personal information, such as user profiles in social networks, raising serious privacy concerns when graph learning is performed using GNNs. To address this issue, locally private graph learning protocols have gained considerable attention. These protocols leverage the privacy advantages of local differential privacy (LDP) and the effectiveness of GNN's message-passing in calibrating noisy data, offering strict privacy guarantees for users' local data while maintaining high utility (e.g., node classification accuracy) for graph learning. Despite these advantages, such protocols may be vulnerable to data poisoning attacks, a threat that has not been considered in previous research. Identifying and addressing these threats is crucial for ensuring the robustness and security of privacy-preserving graph learning frameworks. This work introduces the first data poisoning attack targeting locally private graph learning protocols. The attacker injects fake users into the protocol, manipulates these fake users to establish links with genuine users, and sends carefully crafted data to the server, ultimately compromising the utility of private graph learning. The effectiveness of the attack is demonstrated both theoretically and empirically. In addition, several defense strategies have also been explored, but their limited effectiveness highlights the need for more robust defenses.
Authors: Sonia Mazelet, R\'emi Flamary, Bertrand Thirion
Abstract: Optimal transport between graphs, based on Gromov-Wasserstein and other extensions, is a powerful tool for comparing and aligning graph structures. However, solving the associated non-convex optimization problems is computationally expensive, which limits the scalability of these methods to large graphs. In this work, we present Unbalanced Learning of Optimal Transport (ULOT), a deep learning method that predicts optimal transport plans between two graphs. Our method is trained by minimizing the fused unbalanced Gromov-Wasserstein (FUGW) loss. We propose a novel neural architecture with cross-attention that is conditioned on the FUGW tradeoff hyperparameters. We evaluate ULOT on synthetic stochastic block model (SBM) graphs and on real cortical surface data obtained from fMRI. ULOT predicts transport plans with competitive loss up to two orders of magnitude faster than classical solvers. Furthermore, the predicted plan can be used as a warm start for classical solvers to accelerate their convergence. Finally, the predicted transport plan is fully differentiable with respect to the graph inputs and FUGW hyperparameters, enabling the optimization of functionals of the ULOT plan.
Authors: Yaozhong Shi, Zachary E. Ross, Domniki Asimaki, Kamyar Azizzadenesheli
Abstract: Generative models in function spaces, situated at the intersection of generative modeling and operator learning, are attracting increasing attention due to their immense potential in diverse scientific and engineering applications. While functional generative models are theoretically domain- and discretization-agnostic, current implementations heavily rely on the Fourier Neural Operator (FNO), limiting their applicability to regular grids and rectangular domains. To overcome these critical limitations, we introduce the Mesh-Informed Neural Operator (MINO). By leveraging graph neural operators and cross-attention mechanisms, MINO offers a principled, domain- and discretization-agnostic backbone for generative modeling in function spaces. This advancement significantly expands the scope of such models to more diverse applications in generative, inverse, and regression tasks. Furthermore, MINO provides a unified perspective on integrating neural operators with general advanced deep learning architectures. Finally, we introduce a suite of standardized evaluation metrics that enable objective comparison of functional generative models, addressing another critical gap in the field.
Authors: Daniel Wurgaft, Ekdeep Singh Lubana, Core Francisco Park, Hidenori Tanaka, Gautam Reddy, Noah D. Goodman
Abstract: Recent work analyzing in-context learning (ICL) has identified a broad set of strategies that describe model behavior in different experimental conditions. We aim to unify these findings by asking why a model learns these disparate strategies in the first place. Specifically, we start with the observation that when trained to learn a mixture of tasks, as is popular in the literature, the strategies learned by a model for performing ICL can be captured by a family of Bayesian predictors: a memorizing predictor, which assumes a discrete prior on the set of seen tasks, and a generalizing predictor, where the prior matches the underlying task distribution. Adopting the normative lens of rational analysis, where a learner's behavior is explained as an optimal adaptation to data given computational constraints, we develop a hierarchical Bayesian framework that almost perfectly predicts Transformer next-token predictions throughout training -- without assuming access to its weights. Under this framework, pretraining is viewed as a process of updating the posterior probability of different strategies, and inference-time behavior as a posterior-weighted average over these strategies' predictions. Our framework draws on common assumptions about neural network learning dynamics, which make explicit a tradeoff between loss and complexity among candidate strategies: beyond how well it explains the data, a model's preference towards implementing a strategy is dictated by its complexity. This helps explain well-known ICL phenomena, while offering novel predictions: e.g., we show a superlinear trend in the timescale for transitioning from generalization to memorization as task diversity increases. Overall, our work advances an explanatory and predictive account of ICL grounded in tradeoffs between strategy loss and complexity.
Authors: Xingyu Alice Yang, Jianyu Zhang, L\'eon Bottou
Abstract: Transfer learning is a cornerstone of modern machine learning, promising a way to adapt models pretrained on a broad mix of data to new tasks with minimal new data. However, a significant challenge remains in ensuring that transferred features are sufficient to handle unseen datasets, amplified by the difficulty of quantifying whether two tasks are "related". To address these challenges, we evaluate model transfer from a pretraining mixture to each of its component tasks, assessing whether pretrained features can match the performance of task-specific direct training. We identify a fundamental limitation in deep learning models -- an "information saturation bottleneck" -- where networks fail to learn new features once they encode similar competing features during training. When restricted to learning only a subset of key features during pretraining, models will permanently lose critical features for transfer and perform inconsistently on data distributions, even components of the training mixture. Empirical evidence from published studies suggests that this phenomenon is pervasive in deep learning architectures -- factors such as data distribution or ordering affect the features that current representation learning methods can learn over time. This study suggests that relying solely on large-scale networks may not be as effective as focusing on task-specific training, when available. We propose richer feature representations as a potential solution to better generalize across new datasets and, specifically, present existing methods alongside a novel approach, the initial steps towards addressing this challenge.
Authors: Steven Kolawole, Keshav Santhanam, Virginia Smith, Pratiksha Thaker
Abstract: LLM serving systems typically treat user prompts as monolithic inputs, optimizing inference through decoding tricks or inter-query batching. However, many real-world prompts contain latent semantic parallelism--decomposable structures where subtasks can be executed independently to reduce latency while preserving meaning. We introduce PARALLELPROMPT, the first benchmark for measuring intra-query parallelism in natural user prompts. Our dataset comprises over 37,000 real-world prompts from public LLM chat logs, each annotated with a structured schema capturing task templates, shared context, and iteration inputs. These schemas are extracted using LLM-assisted prompting with rule-based multilingual validation. To evaluate the benefits of decomposition, we provide an execution suite that benchmarks serial vs. parallel strategies, measuring latency, structural adherence, and semantic fidelity. Our results show that intra-query parallelism can be successfully parsed in over 75% of curated datasets, unlocking up to 5x speedups on tasks like translation, comprehension, and comparative analysis, with minimal quality degradation. By releasing this benchmark, curation pipeline, and evaluation suite, we provide the first standardized testbed for studying structure-aware execution in LLM serving pipelines.
Authors: Shriyank Somvanshi, Md Monzurul Islam, Gaurab Chhetri, Rohit Chakraborty, Mahmuda Sultana Mimi, Sawgat Ahmed Shuvo, Kazi Sifatul Islam, Syed Aaqib Javed, Sharif Ahmed Rafat, Anandi Dutta, Subasish Das
Abstract: The rapid growth of edge devices has driven the demand for deploying artificial intelligence (AI) at the edge, giving rise to Tiny Machine Learning (TinyML) and its evolving counterpart, Tiny Deep Learning (TinyDL). While TinyML initially focused on enabling simple inference tasks on microcontrollers, the emergence of TinyDL marks a paradigm shift toward deploying deep learning models on severely resource-constrained hardware. This survey presents a comprehensive overview of the transition from TinyML to TinyDL, encompassing architectural innovations, hardware platforms, model optimization techniques, and software toolchains. We analyze state-of-the-art methods in quantization, pruning, and neural architecture search (NAS), and examine hardware trends from MCUs to dedicated neural accelerators. Furthermore, we categorize software deployment frameworks, compilers, and AutoML tools enabling practical on-device learning. Applications across domains such as computer vision, audio recognition, healthcare, and industrial monitoring are reviewed to illustrate the real-world impact of TinyDL. Finally, we identify emerging directions including neuromorphic computing, federated TinyDL, edge-native foundation models, and domain-specific co-design approaches. This survey aims to serve as a foundational resource for researchers and practitioners, offering a holistic view of the ecosystem and laying the groundwork for future advancements in edge AI.
Authors: Yuhui Sun (University of Alberta), Xiyao Wang (University of Toronto), Zixi Li (Zhejiang University), Jinman Zhao (University of Toronto)
Abstract: While large-scale unsupervised language models (LMs) capture broad world knowledge and reasoning capabilities, steering their behavior toward desired objectives remains challenging due to the lack of explicit supervision. Existing alignment techniques, such as reinforcement learning from human feedback (RLHF), rely on training a reward model and performing reinforcement learning to align with human preferences. However, RLHF is often computationally intensive, unstable, and sensitive to hyperparameters. To address these limitations, Direct Preference Optimization (DPO) was introduced as a lightweight and stable alternative, enabling direct alignment of language models with pairwise preference data via classification loss. However, DPO and its extensions generally assume a single static preference distribution, limiting flexibility in multi-objective or dynamic alignment settings. In this paper, we propose a novel framework: Multi-Preference Lambda-weighted Listwise DPO, which extends DPO to incorporate multiple human preference dimensions (e.g., helpfulness, harmlessness, informativeness) and enables dynamic interpolation through a controllable simplex-weighted formulation. Our method supports both listwise preference feedback and flexible alignment across varying user intents without re-training. Empirical and theoretical analysis demonstrates that our method is as effective as traditional DPO on static objectives while offering greater generality and adaptability for real-world deployment.
Authors: Jaesung Yoo, Sunghyuk Choi, Ye Seul Yang, Suhyeon Kim, Jieun Choi, Dongkyeong Lim, Yaeji Lim, Hyung Joon Joo, Dae Jung Kim, Rae Woong Park, Hyeong-Jin Yoon, Kwangsoo Kim
Abstract: When a deep learning model is trained sequentially on different datasets, it often forgets the knowledge learned from previous data, a problem known as catastrophic forgetting. This damages the model's performance on diverse datasets, which is critical in privacy-preserving deep learning (PPDL) applications based on transfer learning (TL). To overcome this, we introduce "review learning" (RevL), a low cost continual learning algorithm for diagnosis prediction using electronic health records (EHR) within a PPDL framework. RevL generates data samples from the model which are used to review knowledge from previous datasets. Six simulated institutional experiments and one real-world experiment involving three medical institutions were conducted to validate RevL, using three binary classification EHR data. In the real-world experiment with data from 106,508 patients, the mean global area under the receiver operating curve was 0.710 for RevL and 0.655 for TL. These results demonstrate RevL's ability to retain previously learned knowledge and its effectiveness in real-world PPDL scenarios. Our work establishes a realistic pipeline for PPDL research based on model transfers across institutions and highlights the practicality of continual learning in real-world medical settings using private EHR data.
Authors: Steven Walton, Ali Hassani, Xingqian Xu, Zhangyang Wang, Humphrey Shi
Abstract: While the integration of transformers in vision models have yielded significant improvements on vision tasks they still require significant amounts of computation for both training and inference. Restricted attention mechanisms significantly reduce these computational burdens but come at the cost of losing either global or local coherence. We propose a simple, yet powerful method to reduce these trade-offs: allow the attention heads of a single transformer to attend to multiple receptive fields. We demonstrate our method utilizing Neighborhood Attention (NA) and integrate it into a StyleGAN based architecture for image generation. With this work, dubbed StyleNAT, we are able to achieve a FID of 2.05 on FFHQ, a 6% improvement over StyleGAN-XL, while utilizing 28% fewer parameters and with 4$\times$ the throughput capacity. StyleNAT achieves the Pareto Frontier on FFHQ-256 and demonstrates powerful and efficient image generation on other datasets. Our code and model checkpoints are publicly available at: https://github.com/SHI-Labs/StyleNAT
Authors: Junpei Komiyama, Masaaki Imaizumi
Abstract: In this research, we investigate the high-dimensional linear contextual bandit problem where the number of features $p$ is greater than the budget $T$, or it may even be infinite. Differing from the majority of previous works in this field, we do not impose sparsity on the regression coefficients. Instead, we rely on recent findings on overparameterized models, which enables us to analyze the performance of the minimum-norm interpolating estimator when data distributions have small effective ranks. We propose an explore-then-commit (EtC) algorithm to address this problem and examine its performance. Through our analysis, we derive the optimal rate of the ETC algorithm in terms of $T$ and show that this rate can be achieved by balancing exploration and exploitation. Moreover, we introduce an adaptive explore-then-commit (AEtC) algorithm that adaptively finds the optimal balance. We assess the performance of the proposed algorithms through a series of simulations.
Authors: C. -C. Joseph Wang, F. Perkkola, I. Salmenper\"a, A. Meijer-van de Griend, J. K. Nurminen, R. S. Bennink
Abstract: Hybrid variational quantum algorithms (VQAs) are promising for solving practical problems such as combinatorial optimization, quantum chemistry simulation, quantum machine learning, and quantum error correction on noisy quantum computers. However, with typical random ansatz or quantum alternating operator ansatz, derived variational quantum algorithms become a black box that cannot be trusted for model interpretation, not to mention deploying as applications in informing critical decisions: the results of these variational parameters are just rotational angles for the quantum gates and have nothing to do with interpretable values that a model can provide directly. In this paper, we construct the first interpretable quantum regression algorithm, in which the quantum state exactly encodes the classical data table and the variational parameters correspond directly to the regression coefficients, which are real numbers by construction, providing a high degree of model interpretability and minimal cost to optimize due to the right expressiveness. We also take advantage of the encoded data structure to reduce the time complexity of computing the regression map. To shorten the circuit depth for nonlinear regression, our algorithm can be extended by building nonlinear features by classical preprocessing as the independent encoded column vectors. Even though the realization of compressed encoding in superconducting qubits has been achieved by the less noisy compressed encoding recently by the authors, we envision potential quantum utilities with multi-qubit gates implemented in neutral cold atoms and ions.
Authors: Ozlem Ceviz (WISE Lab., Deparment of Computer Engineering, Hacettepe University, Ankara, Turkey), Pinar Sadioglu (WISE Lab., Deparment of Computer Engineering, Hacettepe University, Ankara, Turkey), Sevil Sen (WISE Lab., Deparment of Computer Engineering, Hacettepe University, Ankara, Turkey), Vassilios G. Vassilakis (Department of Computer Science, University of York, York, United Kingdom)
Abstract: Unmanned aerial vehicles (UAVs) operating within Flying Ad-hoc Networks (FANETs) encounter security challenges due to the dynamic and distributed nature of these networks. Previous studies focused predominantly on centralized intrusion detection, assuming a central entity responsible for storing and analyzing data from all devices. However, these approaches face challenges including computation and storage costs, along with a single point of failure risk, threatening data privacy and availability. The widespread dispersion of data across interconnected devices underscores the need for decentralized approaches. This paper introduces the Federated Learning-based Intrusion Detection System (FL-IDS), addressing challenges encountered by centralized systems in FANETs. FL-IDS reduces computation and storage costs for both clients and the central server, which is crucial for resource-constrained UAVs. Operating in a decentralized manner, FL-IDS enables UAVs to collaboratively train a global intrusion detection model without sharing raw data, thus avoiding delay in decisions based on collected data, as is often the case with traditional methods. Experimental results demonstrate FL-IDS's competitive performance with Central IDS (C-IDS) while mitigating privacy concerns, with the Bias Towards Specific Clients (BTSC) method further enhancing FL-IDS performance even at lower attacker ratios. Comparative analysis with traditional intrusion detection methods, including Local IDS (L-IDS), sheds light on the strengths of FL-IDS. This study significantly contributes to UAV security by introducing a privacy-aware, decentralized intrusion detection approach tailored to UAV networks. Moreover, by introducing a realistic dataset for FANETs and federated learning, our approach differs from others lacking high dynamism and 3D node movements or accurate federated data federations.
Authors: Ping Guo, Xiang Li, Zhiyuan Yang, Xi Lin, Qingchuan Zhao, Qingfu Zhang
Abstract: Black-box query-based attacks constitute significant threats to Machine Learning as a Service (MLaaS) systems since they can generate adversarial examples without accessing the target model's architecture and parameters. Traditional defense mechanisms, such as adversarial training, gradient masking, and input transformations, either impose substantial computational costs or compromise the test accuracy of non-adversarial inputs. To address these challenges, we propose an efficient defense mechanism, PuriDefense, that employs random patch-wise purifications with an ensemble of lightweight purification models at a low level of inference cost. These models leverage the local implicit function and rebuild the natural image manifold. Our theoretical analysis suggests that this approach slows down the convergence of query-based attacks by incorporating randomness into purifications. Extensive experiments on CIFAR-10 and ImageNet validate the effectiveness of our proposed purifier-based defense mechanism, demonstrating significant improvements in robustness against query-based attacks.
Authors: V Hewes, Adam Aurisano, Giuseppe Cerati, Jim Kowalkowski, Claire Lee, Wei-keng Liao, Daniel Grzenda, Kaushal Gumpula, Xiaohe Zhang
Abstract: Liquid Argon Time Projection Chamber (LArTPC) detector technology offers a wealth of high-resolution information on particle interactions, and leveraging that information to its full potential requires sophisticated automated reconstruction techniques. This article describes NuGraph2, a Graph Neural Network (GNN) for low-level reconstruction of simulated neutrino interactions in a LArTPC detector. Simulated neutrino interactions in the MicroBooNE detector geometry are described as heterogeneous graphs, with energy depositions on each detector plane forming nodes on planar subgraphs. The network utilizes a multi-head attention message-passing mechanism to perform background filtering and semantic labelling on these graph nodes, identifying those associated with the primary physics interaction with 98.0\% efficiency and labelling them according to particle type with 94.9\% efficiency. The network operates directly on detector observables across multiple 2D representations, but utilizes a 3D-context-aware mechanism to encourage consistency between these representations. Model inference takes 0.12~s/event on a CPU, and 0.005s/event batched on a GPU. This architecture is designed to be a general-purpose solution for particle reconstruction in neutrino physics, with the potential for deployment across a broad range of detector technologies, and offers a core convolution engine that can be leveraged for a variety of tasks beyond the two described in this article.
Authors: Murat Onur Yildirim, Elif Ceren Gok Yildirim, Decebal Constantin Mocanu, Joaquin Vanschoren
Abstract: Neural networks often struggle with catastrophic forgetting when learning sequences of tasks or data streams, unlike humans who can continuously learn and consolidate new concepts even in the absence of explicit cues. Online data-incremental learning seeks to emulate this capability by processing each sample only once, without having access to task or stream cues at any point in time since this is more realistic compared to offline setups, where all data from novel class(es) is assumed to be readily available. However, existing methods typically rely on storing the subsets of data in memory or expanding the initial model architecture, resulting in significant computational overhead. Drawing inspiration from 'self-regulated neurogenesis'-brain's mechanism for creating specialized regions or circuits for distinct functions-we propose a novel approach SERENA which encodes each concept in a specialized network path called 'concept cell', integrated into a single over-parameterized network. Once a concept is learned, its corresponding concept cell is frozen, effectively preventing the forgetting of previously acquired information. Furthermore, we introduce two new continual learning scenarios that more closely reflect real-world conditions, characterized by gradually changing sample sizes. Experimental results show that our method not only establishes new state-of-the-art results across ten benchmarks but also remarkably surpasses offline supervised batch learning performance. The code is available at https://github.com/muratonuryildirim/serena.
Authors: Enzo Terreau, Antoine Gourru, Julien Velcin
Abstract: A wide range of Deep Natural Language Processing (NLP) models integrates continuous and low dimensional representations of words and documents. Surprisingly, very few models study representation learning for authors. These representations can be used for many NLP tasks, such as author identification and classification, or in recommendation systems. A strong limitation of existing works is that they do not explicitly capture writing style, making them hardly applicable to literary data. We therefore propose a new architecture based on Variational Information Bottleneck (VIB) that learns embeddings for both authors and documents with a stylistic constraint. Our model fine-tunes a pre-trained document encoder. We stimulate the detection of writing style by adding predefined stylistic features making the representation axis interpretable with respect to writing style indicators. We evaluate our method on three datasets: a literary corpus extracted from the Gutenberg Project, the Blog Authorship Corpus and IMDb62, for which we show that it matches or outperforms strong/recent baselines in authorship attribution while capturing much more accurately the authors stylistic aspects.
Authors: El Mahdi Chayti, Nikita Doikov, Martin Jaggi
Abstract: We study stochastic second-order methods for solving general non-convex optimization problems. We propose using a special version of momentum to stabilize the stochastic gradient and Hessian estimates in Newton's method. We show that momentum provably improves the variance of stochastic estimates and allows the method to converge for any noise level. Using the cubic regularization technique, we prove a global convergence rate for our method on general non-convex problems to a second-order stationary point, even when using only a single stochastic data sample per iteration. This starkly contrasts with all existing stochastic second-order methods for non-convex problems, which typically require large batches. Therefore, we are the first to demonstrate global convergence for batches of arbitrary size in the non-convex case for the Stochastic Cubic Newton. Additionally, we show improved speed on convex stochastic problems for our regularized Newton methods with momentum.
Authors: Xiao Xia, Dan Zhang, Zibo Liao, Zhenyu Hou, Tianrui Sun, Jing Li, Ling Fu, Yuxiao Dong
Abstract: The modeling of industrial scenes is essential for simulations in industrial manufacturing. While large language models (LLMs) have shown significant progress in generating general 3D scenes from textual descriptions, generating industrial scenes with LLMs poses a unique challenge due to their demand for precise measurements and positioning, requiring complex planning over spatial arrangement. To address this challenge, we introduce SceneGenAgent, an LLM-based agent for generating industrial scenes through C# code. SceneGenAgent ensures precise layout planning through a structured and calculable format, layout verification, and iterative refinement to meet the quantitative requirements of industrial scenarios. Experiment results demonstrate that LLMs powered by SceneGenAgent exceed their original performance, reaching up to 81.0% success rate in real-world industrial scene generation tasks and effectively meeting most scene generation requirements. To further enhance accessibility, we construct SceneInstruct, a dataset designed for fine-tuning open-source LLMs to integrate into SceneGenAgent. Experiments show that fine-tuning open-source LLMs on SceneInstruct yields significant performance improvements, with Llama3.1-70B approaching the capabilities of GPT-4o. Our code and data are available at https://github.com/THUDM/SceneGenAgent .
Authors: Robert Fonod, Haechan Cho, Hwasoo Yeo, Nikolas Geroliminis
Abstract: This paper presents a framework for extracting georeferenced vehicle trajectories from high-altitude drone imagery, addressing key challenges in urban traffic monitoring and the limitations of traditional ground-based systems. Our approach integrates several novel contributions, including a tailored object detector optimized for high-altitude bird's-eye view perspectives, a unique track stabilization method that uses detected vehicle bounding boxes as exclusion masks during image registration, and an orthophoto and master frame-based georeferencing strategy that enhances consistent alignment across multiple drone viewpoints. Additionally, our framework features robust vehicle dimension estimation and detailed road segmentation, enabling comprehensive traffic analysis. Conducted in the Songdo International Business District, South Korea, the study utilized a multi-drone experiment covering 20 intersections, capturing approximately 12TB of 4K video data over four days. The framework produced two high-quality datasets: the Songdo Traffic dataset, comprising approximately 700,000 unique vehicle trajectories, and the Songdo Vision dataset, containing over 5,000 human-annotated images with about 300,000 vehicle instances in four classes. Comparisons with high-precision sensor data from an instrumented probe vehicle highlight the accuracy and consistency of our extraction pipeline in dense urban environments. The public release of Songdo Traffic and Songdo Vision, and the complete source code for the extraction pipeline, establishes new benchmarks in data quality, reproducibility, and scalability in traffic research. Results demonstrate the potential of integrating drone technology with advanced computer vision for precise and cost-effective urban traffic monitoring, providing valuable resources for developing intelligent transportation systems and enhancing traffic management strategies.
Authors: Hoang H Nguyen, Khyati Mahajan, Vikas Yadav, Julian Salazar, Philip S. Yu, Masoud Hashemi, Rishabh Maheshwary
Abstract: Although multilingual LLMs have achieved remarkable performance across benchmarks, we find they continue to underperform on non-Latin script languages across contemporary LLM families. This discrepancy arises from the fact that LLMs are pretrained with orthographic scripts, which are dominated by Latin characters that obscure their shared phonology with non-Latin scripts. We propose leveraging phonemic transcriptions as complementary signals to induce script-invariant representations. Our study demonstrates that integrating phonemic signals improves performance across both non-Latin and Latin script languages, with a particularly significant impact on closing the performance gap between the two. Through detailed experiments, we show that phonemic and orthographic scripts retrieve distinct examples for in-context learning (ICL). This motivates our proposed Mixed-ICL retrieval strategy, where further aggregation from both leads to our significant performance improvements for both Latin script languages (up to 12.6%) and non-Latin script languages (up to 15.1%) compared to randomized ICL retrieval.
Authors: Zhichen Zeng, Xiaolong Liu, Mengyue Hang, Xiaoyi Liu, Qinghai Zhou, Chaofei Yang, Yiqun Liu, Yichen Ruan, Laming Chen, Yuxin Chen, Yujia Hao, Jiaqi Xu, Jade Nie, Xi Liu, Buyun Zhang, Wei Wen, Siyang Yuan, Hang Yin, Xin Zhang, Kai Wang, Wen-Yen Chen, Yiping Han, Huayu Li, Chunzhi Yang, Bo Long, Philip S. Yu, Hanghang Tong, Jiyan Yang
Abstract: Click-through rate (CTR) prediction, which predicts the probability of a user clicking an ad, is a fundamental task in recommender systems. The emergence of heterogeneous information, such as user profile and behavior sequences, depicts user interests from different aspects. A mutually beneficial integration of heterogeneous information is the cornerstone towards the success of CTR prediction. However, most of the existing methods suffer from two fundamental limitations, including (1) insufficient inter-mode interaction due to the unidirectional information flow between modes, and (2) aggressive information aggregation caused by early summarization, resulting in excessive information loss. To address the above limitations, we propose a novel module named InterFormer to learn heterogeneous information interaction in an interleaving style. To achieve better interaction learning, InterFormer enables bidirectional information flow for mutually beneficial learning across different modes. To avoid aggressive information aggregation, we retain complete information in each data mode and use a separate bridging arch for effective information selection and summarization. Our proposed InterFormer achieves state-of-the-art performance on three public datasets and a large-scale industrial dataset.
Authors: Rongkun Xue, Jinouwen Zhang, Yazhe Niu, Dazhong Shen, Bingqi Ma, Yu Liu, Jing Yang
Abstract: Recent generative models based on score matching and flow matching have significantly advanced generation tasks, but their potential in discriminative tasks remains underexplored. Previous approaches, such as generative classifiers, have not fully leveraged the capabilities of these models for discriminative tasks due to their intricate designs. We propose Pretrained Reversible Generation (PRG), which extracts unsupervised representations by reversing the generative process of a pretrained continuous generation model. PRG effectively reuses unsupervised generative models, leveraging their high capacity to serve as robust and generalizable feature extractors for downstream tasks. This framework enables the flexible selection of feature hierarchies tailored to specific downstream tasks. Our method consistently outperforms prior approaches across multiple benchmarks, achieving state-of-the-art performance among generative model based methods, including 78% top-1 accuracy on ImageNet at a resolution of 64*64. Extensive ablation studies, including out-of-distribution evaluations, further validate the effectiveness of our approach. Code is available at https://github.com/opendilab/PRG.
Authors: Giacomo Belli, Marco Mordacci, Michele Amoretti
Abstract: In this work, a scalable quantum neural network is introduced as a means to approximate any unitary evolution through the Standard Recursive Block Basis (SRBB) and, subsequently, redesigned with a number of CNOTs asymptotically reduced by an exponential contribution. This algebraic approach to the problem of unitary synthesis exploits Lie algebras and their topological features to obtain scalable parameterizations of unitary operators. First, the original SRBB-based scalability scheme, already known in the literature only from a theoretical point of view, is reformulated for efficient algorithm implementation and complexity management. Remarkably, 2-qubit operators emerge as a special case outside the original scaling scheme. Furthermore, an algorithm is proposed to reduce the number of CNOTs, thus deriving a new implementable scaling scheme that requires only one layer of approximation. The scalable CNOT-reduced quantum neural network is implemented and its performance is assessed with a variety of different unitary matrices, both sparse and dense, up to 6 qubits via the PennyLane library. The effectiveness of the approximation is measured with different metrics in relation to two optimizers: a gradient-based method and the Nelder-Mead method. The approximate CNOT-reduced SRBB-based synthesis algorithm is also tested on real hardware and compared with other valid approximation and decomposition methods available in the literature.
Authors: Le Xu, Lei Cheng, Junting Chen, Wenqiang Pu, Xiao Fu
Abstract: Radio map estimation (RME), also known as spectrum cartography, aims to reconstruct the strength of radio interference across different domains (e.g., space and frequency) from sparsely sampled measurements. To tackle this typical inverse problem, state-of-the-art RME methods rely on handcrafted or data-driven structural information of radio maps. However, the former often struggles to model complex radio frequency (RF) environments and the latter requires excessive training -- making it hard to quickly adapt to in situ sensing tasks. This work presents a spatio-spectral RME approach based on plug-and-play (PnP) denoising, a technique from computational imaging. The idea is to leverage the observation that the denoising operations of signals like natural images and radio maps are similar -- despite the nontrivial differences of the signals themselves. Hence, sophisticated denoisers designed for or learned from natural images can be directly employed to assist RME, avoiding using radio map data for training. Unlike conventional PnP methods that operate directly in the data domain, the proposed method exploits the underlying physical structure of radio maps and proposes an ADMM algorithm that denoises in a latent domain. This design significantly improves computational efficiency and enhances noise robustness. Theoretical aspects, e.g., recoverability of the complete radio map and convergence of the ADMM algorithm are analyzed. Synthetic and real data experiments are conducted to demonstrate the effectiveness of our approach.
Authors: Xiaozhi Liu, Yong Xia
Abstract: The computation of the dominant eigenvector of symmetric positive semidefinite matrices is a cornerstone operation in numerous optimization-driven applications. Traditional methods, typically based on the \textit{Quotient} formulation, often suffer from challenges related to computational efficiency and reliance on prior spectral knowledge. In this work, we leverage the alternative \textit{Difference} formulation to reinterpret the classical power method as a first-order optimization algorithm. This perspective allows for a novel convergence analysis and facilitates the development of accelerated variants with larger step-sizes, achieving faster convergence without additional computational cost. Building on this insight, we introduce a generalized family of Difference-based methods, with the power method as a special case. Within this family, we propose Split-Merge, an algorithm that attains accelerated convergence without requiring spectral knowledge and operates solely via matrix-vector products. Extensive experiments on both synthetic and real-world datasets demonstrate that Split-Merge consistently outperforms state-of-the-art methods in both efficiency and scalability. In particular, it achieves more than a $\boldsymbol{10\times}$ speedup over the classical power method, underscoring its practical effectiveness for large-scale problems.
Authors: Sheila E. Whitman, Marat I. Latypov
Abstract: Machine learning of microstructure--property relationships from data is an emerging approach in computational materials science. Most existing machine learning efforts focus on the development of task-specific models for each microstructure--property relationship. We propose utilizing pre-trained foundational vision transformers for the extraction of task-agnostic microstructure features and subsequent light-weight machine learning of a microstructure-dependent property. We demonstrate our approach with pre-trained state-of-the-art vision transformers (CLIP, DINOv2, SAM) in two case studies on machine-learning: (i) elastic modulus of two-phase microstructures based on simulations data; and (ii) Vicker's hardness of Ni-base and Co-base superalloys based on experimental data published in literature. Our results show the potential of foundational vision transformers for robust microstructure representation and efficient machine learning of microstructure--property relationships without the need for expensive task-specific training or fine-tuning of bespoke deep learning models.
Authors: Hao Zhu, Joschka Boedecker
Abstract: We consider the inverse problem of multi-armed bandits (IMAB) that are widely used in neuroscience and psychology research for behavior modelling. We first show that the IMAB problem is not convex in general, but can be relaxed to a convex problem via variable transformation. Based on this result, we propose a two-step sequential heuristic for (approximately) solving the IMAB problem. We discuss a condition where our method provides global solution to the IMAB problem with certificate, as well as approximations to further save computing time. Numerical experiments indicate that our heuristic method is more robust than directly solving the IMAB problem via repeated local optimization, and can achieve the performance of Monte Carlo methods within a significantly decreased running time. We provide the implementation of our method based on CVXPY, which allows straightforward application by users not well versed in convex optimization.
Authors: Grigory Bartosh, Dmitry Vetrov, Christian A. Naesseth
Abstract: The Latent Stochastic Differential Equation (SDE) is a powerful tool for time series and sequence modeling. However, training Latent SDEs typically relies on adjoint sensitivity methods, which depend on simulation and backpropagation through approximate SDE solutions, which limit scalability. In this work, we propose SDE Matching, a new simulation-free method for training Latent SDEs. Inspired by modern Score- and Flow Matching algorithms for learning generative dynamics, we extend these ideas to the domain of stochastic dynamics for time series and sequence modeling, eliminating the need for costly numerical simulations. Our results demonstrate that SDE Matching achieves performance comparable to adjoint sensitivity methods while drastically reducing computational complexity.
Authors: Clement Ruah, Houssem Sifaou, Osvaldo Simeone, Bashir Al-Hashimi
Abstract: The widespread adoption of artificial intelligence (AI) in next-generation communication systems is challenged by the heterogeneity of traffic and network conditions, which call for the use of highly contextual, site-specific, data. A promising solution is to rely not only on real-world data, but also on synthetic pseudo-data generated by a network digital twin (NDT). However, the effectiveness of this approach hinges on the accuracy of the NDT, which can vary widely across different contexts. To address this problem, this paper introduces context-aware doubly-robust (CDR) learning, a novel semi-supervised scheme that adapts its reliance on the pseudo-data to the different levels of fidelity of the NDT across contexts. CDR is evaluated on the task of downlink beamforming where it outperforms previous state-of-the-art approaches, providing a 24% loss decrease when compared to doubly-robust (DR) semi-supervised learning in regimes with low labeled data availability.
Authors: Daniel Yiming Cao, August Y. Chen, Karthik Sridharan, Benjamin Tang
Abstract: We study the optimization of non-convex functions that are not necessarily smooth (gradient and/or Hessian are Lipschitz) using first order methods. Smoothness is a restrictive assumption in machine learning in both theory and practice, motivating significant recent work on finding first order stationary points of functions satisfying generalizations of smoothness with first order methods. We develop a novel framework that lets us systematically study the convergence of a large class of first-order optimization algorithms (which we call decrease procedures) under generalizations of smoothness. We instantiate our framework to analyze the convergence of first order optimization algorithms to first and \textit{second} order stationary points under generalizations of smoothness. As a consequence, we establish the first convergence guarantees for first order methods to second order stationary points under generalizations of smoothness. We demonstrate that several canonical examples fall under our framework, and highlight practical implications.
Authors: Hao Zhu, Shengchao Yan, Jasper Hoffmann, Joschka Boedecker
Abstract: Discrete latent factor models (DLFMs) are widely used in various domains such as machine learning, economics, neuroscience, psychology, etc. Currently, fitting a DLFM to some dataset relies on a customized solver for individual models, which requires lots of effort to implement and is limited to the targeted specific instance of DLFMs. In this paper, we propose a generic framework based on CVXPY, which allows users to specify and solve the fitting problem of a wide range of DLFMs, including both regression and classification models, within a very short script. Our framework is flexible and inherently supports the integration of regularization terms and constraints on the DLFM parameters and latent factors, such that the users can easily prototype the DLFM structure according to their dataset and application scenario. We introduce our open-source Python implementation and illustrate the framework in several examples.
Authors: Yaofeng Cheng, Fusheng Zha, Wei Guo, Pengfei Wang, Chao Zeng, Lining Sun, Chenguang Yang
Abstract: The 6-Degree of Freedom (DoF) grasp method based on point clouds has shown significant potential in enabling robots to grasp target objects. However, most existing methods are based on the point clouds (2.5D points) generated from single-view depth images. These point clouds only have one surface side of the object providing incomplete geometry information, which mislead the grasping algorithm to judge the shape of the target object, resulting in low grasping accuracy. Humans can accurately grasp objects from a single view by leveraging their geometry experience to estimate object shapes. Inspired by humans, we propose a novel 6-DoF grasping framework that converts the point completion results as object shape features to train the 6-DoF grasp network. Here, point completion can generate approximate complete points from the 2.5D points similar to the human geometry experience, and converting it as shape features is the way to utilize it to improve grasp efficiency. Furthermore, due to the gap between the network generation and actual execution, we integrate a score filter into our framework to select more executable grasp proposals for the real robot. This enables our method to maintain a high grasp quality in any camera viewpoint. Extensive experiments demonstrate that utilizing complete point features enables the generation of significantly more accurate grasp proposals and the inclusion of a score filter greatly enhances the credibility of real-world robot grasping. Our method achieves a 17.8\% success rate higher than the state-of-the-art method in real-world experiments.
Authors: Harshit Kapadia, Peter Benner, Lihong Feng
Abstract: In situations where the solution of a high-fidelity dynamical system needs to be evaluated repeatedly, over a vast pool of parametric configurations and in absence of access to the underlying governing equations, data-driven model reduction techniques are preferable. We propose a novel active learning approach to build a parametric data-driven reduced-order model (ROM) by greedily picking the most important parameter samples from the parameter domain. As a result, during the ROM construction phase, the number of high-fidelity solutions dynamically grow in a principled fashion. The high-fidelity solution snapshots are expressed in several parameter-specific linear subspaces, with the help of proper orthogonal decomposition (POD), and the relative distance between these subspaces is used as a guiding mechanism to perform active learning. For successfully achieving this, we provide a distance measure to evaluate the similarity between pairs of linear subspaces with different dimensions, and also show that this distance measure is a metric. The usability of the proposed subspace-distance-enabled active learning (SDE-AL) framework is demonstrated by augmenting two existing non-intrusive reduced-order modeling approaches, and providing their active-learning-driven (ActLearn) extensions, namely, SDE-ActLearn-POD-KSNN, and SDE-ActLearn-POD-NN. Furthermore, we report positive results for two parametric physical models, highlighting the efficiency of the proposed SDE-AL approach.
Authors: Henry Peng Zou, Wei-Chieh Huang, Yaozu Wu, Yankai Chen, Chunyu Miao, Hoang Nguyen, Yue Zhou, Weizhi Zhang, Liancheng Fang, Langzhou He, Yangning Li, Dongyuan Li, Renhe Jiang, Xue Liu, Philip S. Yu
Abstract: Recent advances in large language models (LLMs) have sparked growing interest in building fully autonomous agents. However, fully autonomous LLM-based agents still face significant challenges, including limited reliability due to hallucinations, difficulty in handling complex tasks, and substantial safety and ethical risks, all of which limit their feasibility and trustworthiness in real-world applications. To overcome these limitations, LLM-based human-agent systems (LLM-HAS) incorporate human-provided information, feedback, or control into the agent system to enhance system performance, reliability and safety. These human-agent collaboration systems enable humans and LLM-based agents to collaborate effectively by leveraging their complementary strengths. This paper provides the first comprehensive and structured survey of LLM-HAS. It clarifies fundamental concepts, systematically presents core components shaping these systems, including environment & profiling, human feedback, interaction types, orchestration and communication, explores emerging applications, and discusses unique challenges and opportunities arising from human-AI collaboration. By consolidating current knowledge and offering a structured overview, we aim to foster further research and innovation in this rapidly evolving interdisciplinary field. Paper lists and resources are available at https://github.com/HenryPengZou/Awesome-Human-Agent-Collaboration-Interaction-Systems.
URLs: https://github.com/HenryPengZou/Awesome-Human-Agent-Collaboration-Interaction-Systems.
Authors: Olawale Salaudeen, Anka Reuel, Ahmed Ahmed, Suhana Bedi, Zachary Robertson, Sudharsan Sundar, Ben Domingue, Angelina Wang, Sanmi Koyejo
Abstract: While the capabilities and utility of AI systems have advanced, rigorous norms for evaluating these systems have lagged. Grand claims, such as models achieving general reasoning capabilities, are supported with model performance on narrow benchmarks, like performance on graduate-level exam questions, which provide a limited and potentially misleading assessment. We provide a structured approach for reasoning about the types of evaluative claims that can be made given the available evidence. For instance, our framework helps determine whether performance on a mathematical benchmark is an indication of the ability to solve problems on math tests or instead indicates a broader ability to reason. Our framework is well-suited for the contemporary paradigm in machine learning, where various stakeholders provide measurements and evaluations that downstream users use to validate their claims and decisions. At the same time, our framework also informs the construction of evaluations designed to speak to the validity of the relevant claims. By leveraging psychometrics' breakdown of validity, evaluations can prioritize the most critical facets for a given claim, improving empirical utility and decision-making efficacy. We illustrate our framework through detailed case studies of vision and language model evaluations, highlighting how explicitly considering validity strengthens the connection between evaluation evidence and the claims being made.
Authors: Jeffrey T. H. Wong, Cheng Zhang, Xinye Cao, Pedro Gimenes, George A. Constantinides, Wayne Luk, Yiren Zhao
Abstract: Large language models have demonstrated remarkable performance; however, their massive parameter counts make deployment highly expensive. Low-rank approximation offers a promising compression solution, yet existing approaches have two main limitations: (1) They focus on minimizing the output error of individual linear layers, without considering the architectural characteristics of Transformers, and (2) they decompose a large weight matrix into two small low-rank matrices. Consequently, these methods often fall short compared to other compression techniques like pruning and quantization, and introduce runtime overhead such as the extra GEMM kernel launches for decomposed small matrices. To address these limitations, we propose $\tt A^\tt 3$, a post-training low-rank approximation framework. $\tt A^\tt 3$ splits a Transformer layer into three functional components, namely $\tt QK$, $\tt OV$, and $\tt MLP$. For each component, $\tt A^\tt 3$ provides an analytical solution that reduces the hidden dimension size inside each component while minimizing the component's functional loss ($\it i.e.$, error in attention scores, attention outputs, and MLP outputs). This approach directly reduces model sizes, KV cache sizes, and FLOPs without introducing any runtime overheads. In addition, it provides a new narrative in advancing the optimization problem from singular linear layer loss optimization toward improved end-to-end performance. Through extensive experiments, we show that $\tt A^\tt 3$ maintains superior performance compared to SoTAs. For example, under the same reduction budget in computation and memory, our low-rank approximated LLaMA 3.1-70B achieves a perplexity of 4.69 on WikiText-2, outperforming the previous SoTA's 7.87 by 3.18. We also demonstrate the versatility of $\tt A^\tt 3$, including KV cache compression, quantization, and mixed-rank assignments for enhanced performance.
Authors: Shogo Nakakita
Abstract: We examine the concentration of uniform generalization errors around their expectation in binary linear classification problems via an isoperimetric argument. In particular, we establish Poincar\'{e} and log-Sobolev inequalities for the joint distribution of the output labels and the label-weighted input vectors, which we apply to derive concentration bounds. The derived concentration bounds are sharp up to moderate multiplicative constants by those under well-balanced labels. In asymptotic analysis, we also show that almost sure convergence of uniform generalization errors to their expectation occurs in very broad settings, such as proportionally high-dimensional regimes. Using this convergence, we establish uniform laws of large numbers under dimension-free conditions.
Authors: Sanjana Chalavadi, Andrei Pastor, Terry Leitch
Abstract: This study analyzes tract-level real estate ownership patterns in New York State (NYS) and New York City (NYC) to uncover racial disparities. We use an advanced race/ethnicity imputation model (LSTM+Geo with XGBoost filtering, validated at 89.2% accuracy) to compare the predicted racial composition of property owners to the resident population from census data. We examine both a Full Model (statewide) and a Name-Only LSTM Model (NYC) to assess how incorporating geospatial context affects our predictions and disparity estimates. The results reveal significant inequities: White individuals hold a disproportionate share of properties and property value relative to their population, while Black, Hispanic, and Asian communities are underrepresented as property owners. These disparities are most pronounced in minority-majority neighborhoods, where ownership is predominantly White despite a predominantly non-White population. Corporate ownership (LLCs, trusts, etc.) exacerbates these gaps by reducing owner-occupied opportunities in urban minority communities. We provide a breakdown of ownership vs. population by race for majority-White, -Black, -Hispanic, and -Asian tracts, identify those with extreme ownership disparities, and compare patterns in urban, suburban, and rural contexts. The findings underscore persistent racial inequity in property ownership, reflecting broader historical and socio-economic forces, and highlight the importance of data-driven approaches to address these issues.
Authors: Zeinab Dehghani, Mohammed Naveed Akram, Koorosh Aslansefat, Adil Khan
Abstract: Large language models like GPT, LLAMA, and Claude have become incredibly powerful at generating text, but they are still black boxes, so it is hard to understand how they decide what to say. That lack of transparency can be problematic, especially in fields where trust and accountability matter. To help with this, we introduce SMILE, a new method that explains how these models respond to different parts of a prompt. SMILE is model-agnostic and works by slightly changing the input, measuring how the output changes, and then highlighting which words had the most impact. Create simple visual heat maps showing which parts of a prompt matter the most. We tested SMILE on several leading LLMs and used metrics such as accuracy, consistency, stability, and fidelity to show that it gives clear and reliable explanations. By making these models easier to understand, SMILE brings us one step closer to making AI more transparent and trustworthy.
Authors: Amin Karimi Monsefi, Mridul Khurana, Rajiv Ramnath, Anuj Karpatne, Wei-Lun Chao, Cheng Zhang
Abstract: We propose TaxaDiffusion, a taxonomy-informed training framework for diffusion models to generate fine-grained animal images with high morphological and identity accuracy. Unlike standard approaches that treat each species as an independent category, TaxaDiffusion incorporates domain knowledge that many species exhibit strong visual similarities, with distinctions often residing in subtle variations of shape, pattern, and color. To exploit these relationships, TaxaDiffusion progressively trains conditioned diffusion models across different taxonomic levels -- starting from broad classifications such as Class and Order, refining through Family and Genus, and ultimately distinguishing at the Species level. This hierarchical learning strategy first captures coarse-grained morphological traits shared by species with common ancestors, facilitating knowledge transfer before refining fine-grained differences for species-level distinction. As a result, TaxaDiffusion enables accurate generation even with limited training samples per species. Extensive experiments on three fine-grained animal datasets demonstrate that outperforms existing approaches, achieving superior fidelity in fine-grained animal image generation. Project page: https://amink8.github.io/TaxaDiffusion/
Authors: Yanting Wang, Wei Zou, Runpeng Geng, Jinyuan Jia
Abstract: Long context large language models (LLMs) are deployed in many real-world applications such as RAG, agent, and broad LLM-integrated applications. Given an instruction and a long context (e.g., documents, PDF files, webpages), a long context LLM can generate an output grounded in the provided context, aiming to provide more accurate, up-to-date, and verifiable outputs while reducing hallucinations and unsupported claims. This raises a research question: how to pinpoint the texts (e.g., sentences, passages, or paragraphs) in the context that contribute most to or are responsible for the generated output by an LLM? This process, which we call context traceback, has various real-world applications, such as 1) debugging LLM-based systems, 2) conducting post-attack forensic analysis for attacks (e.g., prompt injection attack, knowledge corruption attacks) to an LLM, and 3) highlighting knowledge sources to enhance the trust of users towards outputs generated by LLMs. When applied to context traceback for long context LLMs, existing feature attribution methods such as Shapley have sub-optimal performance and/or incur a large computational cost. In this work, we develop TracLLM, the first generic context traceback framework tailored to long context LLMs. Our framework can improve the effectiveness and efficiency of existing feature attribution methods. To improve the efficiency, we develop an informed search based algorithm in TracLLM. We also develop contribution score ensemble/denoising techniques to improve the accuracy of TracLLM. Our evaluation results show TracLLM can effectively identify texts in a long context that lead to the output of an LLM. Our code and data are at: https://github.com/Wang-Yanting/TracLLM.
Authors: Haibo Qiu, Xiaohan Lan, Fanfan Liu, Xiaohu Sun, Delian Ruan, Peng Shi, Lin Ma
Abstract: Recent advancements in large language models (LLMs) have witnessed a surge in the development of advanced reasoning paradigms, which are now being integrated into multimodal large language models (MLLMs). However, existing approaches often fall short: methods solely employing reinforcement learning (RL) can struggle with sample inefficiency and activating entirely absent reasoning capabilities, while conventional pipelines that initiate with a cold-start supervised fine-tuning (SFT) phase before RL may restrict the model's exploratory capacity and face suboptimal convergence. In this work, we introduce \textbf{Metis-RISE} (\textbf{R}L \textbf{I}ncentivizes and \textbf{S}FT \textbf{E}nhances) for multimodal reasoning model learning. Unlike conventional approaches, Metis-RISE distinctively omits an initial SFT stage, beginning instead with an RL phase (e.g., using a Group Relative Policy Optimization variant) to incentivize and activate the model's latent reasoning capacity. Subsequently, the targeted SFT stage addresses two key challenges identified during RL: (1) \textit{inefficient trajectory sampling} for tasks where the model possesses but inconsistently applies correct reasoning, which we tackle using self-distilled reasoning trajectories from the RL model itself; and (2) \textit{fundamental capability absence}, which we address by injecting expert-augmented knowledge for prompts where the model entirely fails. This strategic application of RL for incentivization followed by SFT for enhancement forms the core of Metis-RISE, leading to two versions of our MLLMs (7B and 72B parameters). Evaluations on the OpenCompass Multimodal Reasoning Leaderboard demonstrate that both models achieve state-of-the-art performance among similar-sized models, with the 72B version ranking fourth overall. Please refer to our project page for open-source information.
Authors: Runtao Liu, Jiahao Zhan, Yingqing He, Chen Wei, Alan Yuille, Qifeng Chen
Abstract: An effective reward model plays a pivotal role in reinforcement learning for post-training enhancement of visual generative models. However, current approaches of reward modeling suffer from implementation complexity due to their reliance on extensive human-annotated preference data or meticulously engineered quality dimensions that are often incomplete and engineering-intensive. Inspired by adversarial training in generative adversarial networks (GANs), this paper proposes GAN-RM, an efficient reward modeling framework that eliminates manual preference annotation and explicit quality dimension engineering. Our method trains the reward model through discrimination between a small set of representative, unpaired target samples(denoted as Preference Proxy Data) and model-generated ordinary outputs, requiring only a few hundred target samples. Comprehensive experiments demonstrate our GAN-RM's effectiveness across multiple key applications including test-time scaling implemented as Best-of-N sample filtering, post-training approaches like Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO). Code and data will be released at https://github.com/Visualignment/GAN-RM.
Authors: Andrew Wagenmaker, Mitsuhiko Nakamoto, Yunchu Zhang, Seohong Park, Waleed Yagoub, Anusha Nagabandi, Abhishek Gupta, Sergey Levine
Abstract: Robotic control policies learned from human demonstrations have achieved impressive results in many real-world applications. However, in scenarios where initial performance is not satisfactory, as is often the case in novel open-world settings, such behavioral cloning (BC)-learned policies typically require collecting additional human demonstrations to further improve their behavior -- an expensive and time-consuming process. In contrast, reinforcement learning (RL) holds the promise of enabling autonomous online policy improvement, but often falls short of achieving this due to the large number of samples it typically requires. In this work we take steps towards enabling fast autonomous adaptation of BC-trained policies via efficient real-world RL. Focusing in particular on diffusion policies -- a state-of-the-art BC methodology -- we propose diffusion steering via reinforcement learning (DSRL): adapting the BC policy by running RL over its latent-noise space. We show that DSRL is highly sample efficient, requires only black-box access to the BC policy, and enables effective real-world autonomous policy improvement. Furthermore, DSRL avoids many of the challenges associated with finetuning diffusion policies, obviating the need to modify the weights of the base policy at all. We demonstrate DSRL on simulated benchmarks, real-world robotic tasks, and for adapting pretrained generalist policies, illustrating its sample efficiency and effective performance at real-world policy improvement.
Authors: Weizhi Zhang, Yangning Li, Yuanchen Bei, Junyu Luo, Guancheng Wan, Liangwei Yang, Chenxuan Xie, Yuyao Yang, Wei-Chieh Huang, Chunyu Miao, Henry Peng Zou, Xiao Luo, Yusheng Zhao, Yankai Chen, Chunkit Chan, Peilin Zhou, Xinyang Zhang, Chenwei Zhang, Jingbo Shang, Ming Zhang, Yangqiu Song, Irwin King, Philip S. Yu
Abstract: Information retrieval is a cornerstone of modern knowledge acquisition, enabling billions of queries each day across diverse domains. However, traditional keyword-based search engines are increasingly inadequate for handling complex, multi-step information needs. Our position is that Large Language Models (LLMs), endowed with reasoning and agentic capabilities, are ushering in a new paradigm termed Agentic Deep Research. These systems transcend conventional information search techniques by tightly integrating autonomous reasoning, iterative retrieval, and information synthesis into a dynamic feedback loop. We trace the evolution from static web search to interactive, agent-based systems that plan, explore, and learn. We also introduce a test-time scaling law to formalize the impact of computational depth on reasoning and search. Supported by benchmark results and the rise of open-source implementations, we demonstrate that Agentic Deep Research not only significantly outperforms existing approaches, but is also poised to become the dominant paradigm for future information seeking. All the related resources, including industry products, research papers, benchmark datasets, and open-source implementations, are collected for the community in https://github.com/DavidZWZ/Awesome-Deep-Research.
Authors: Timoteo Kelly, Abdulkadir Korkmaz, Samuel Mallet, Connor Souders, Sadra Aliakbarpour, Praveen Rao
Abstract: We present HARPT, a large-scale annotated corpus of mobile health app store reviews aimed at advancing research in user privacy and trust. The dataset comprises over 480,000 user reviews labeled into seven categories that capture critical aspects of trust in applications, trust in providers and privacy concerns. Creating HARPT required addressing multiple complexities, such as defining a nuanced label schema, isolating relevant content from large volumes of noisy data, and designing an annotation strategy that balanced scalability with accuracy. This strategy integrated rule-based filtering, iterative manual labeling with review, targeted data augmentation, and weak supervision using transformer-based classifiers to accelerate coverage. In parallel, a carefully curated subset of 7,000 reviews was manually annotated to support model development and evaluation. We benchmark a broad range of classification models, demonstrating that strong performance is achievable and providing a baseline for future research. HARPT is released as a public resource to support work in health informatics, cybersecurity, and natural language processing.
Authors: Xuesong Li, Dianye Huang, Yameng Zhang, Nassir Navab, Zhongliang Jiang
Abstract: Understanding medical ultrasound imaging remains a long-standing challenge due to significant visual variability caused by differences in imaging and acquisition parameters. Recent advancements in large language models (LLMs) have been used to automatically generate terminology-rich summaries orientated to clinicians with sufficient physiological knowledge. Nevertheless, the increasing demand for improved ultrasound interpretability and basic scanning guidance among non-expert users, e.g., in point-of-care settings, has not yet been explored. In this study, we first introduce the scene graph (SG) for ultrasound images to explain image content to ordinary and provide guidance for ultrasound scanning. The ultrasound SG is first computed using a transformer-based one-stage method, eliminating the need for explicit object detection. To generate a graspable image explanation for ordinary, the user query is then used to further refine the abstract SG representation through LLMs. Additionally, the predicted SG is explored for its potential in guiding ultrasound scanning toward missing anatomies within the current imaging view, assisting ordinary users in achieving more standardized and complete anatomical exploration. The effectiveness of this SG-based image explanation and scanning guidance has been validated on images from the left and right neck regions, including the carotid and thyroid, across five volunteers. The results demonstrate the potential of the method to maximally democratize ultrasound by enhancing its interpretability and usability for ordinaries.