Authors: Mingyu Lu, Ethan Weinberger, Chanwoo Kim, Su-In Lee
Abstract: High-content screening (HCS) assays based on high-throughput microscopy techniques such as Cell Painting have enabled the interrogation of cells' morphological responses to perturbations at an unprecedented scale. The collection of such data promises to facilitate a better understanding of the relationships between different perturbations and their effects on cellular state. Towards achieving this goal, recent advances in cross-modal contrastive learning could, in theory, be leveraged to learn a unified latent space that aligns perturbations with their corresponding morphological effects. However, the application of such methods to HCS data is not straightforward due to substantial differences in the semantics of Cell Painting images compared to natural images, and the difficulty of representing different classes of perturbations (e.g., small molecule vs CRISPR gene knockout) in a single latent space. In response to these challenges, here we introduce CellCLIP, a cross-modal contrastive learning framework for HCS data. CellCLIP leverages pre-trained image encoders coupled with a novel channel encoding scheme to better capture relationships between different microscopy channels in image embeddings, along with natural language encoders for representing perturbations. Our framework outperforms current open-source models, demonstrating the best performance in both cross-modal retrieval and biologically meaningful downstream tasks while also achieving significant reductions in computation time.
Authors: Xiaoke Wang, Batuhan Altundas, Zhaoxin Li, Aaron Zhao, Matthew Gombolay
Abstract: Mixed Integer Linear Programs (MILPs) are essential tools for solving planning and scheduling problems across critical industries such as construction, manufacturing, and logistics. However, their widespread adoption is limited by long computational times, especially in large-scale, real-time scenarios. To address this, we present a learning-based framework that leverages Behavior Cloning (BC) and Reinforcement Learning (RL) to train Graph Neural Networks (GNNs), producing high-quality initial solutions for warm-starting MILP solvers in Multi-Agent Task Allocation and Scheduling Problems. Experimental results demonstrate that our method reduces optimization time and variance compared to traditional techniques while maintaining solution quality and feasibility.
Authors: Tianyuan Shi, Canbin Huang, Fanqi Wan, Longguang Zhong, Ziyi Yang, Weizhou Shen, Xiaojun Quan, Ming Yan
Abstract: During the preference optimization of large language models (LLMs), distribution shifts may arise between newly generated model samples and the data used to train the reward model (RM). This shift reduces the efficacy of the RM, which in turn negatively impacts the performance of the policy model (PM). To address this challenge, we propose Mutual-Taught, a self-training method that iteratively improves both the PM and RM without requiring additional human annotation. Our approach mirrors the expectation-maximization (EM) algorithm. In the E-step, the PM is updated using feedback from the current RM, guiding the PM toward a better approximation of the latent optimal preference distribution. In the M-step, we update the RM by constructing training data from the outputs of the PM before and after the E-step update. This process ensures that the RM adapts to the evolving policy distribution. Experimental results demonstrate that this iterative approach leads to consistent improvements in both models. Specifically, our 8B policy model, LLaMA-3-8B-Instruct-MT, achieves a length-controlled win rate of 54.1\% on AlpacaEval-2, while our 8B reward model, FsfairX-LLaMA3-RM-MT, performs on par with GPT-4o-2024-08-06 on RewardBench.
Authors: Junyi Liu, Stanley Kok
Abstract: Agencies such as Standard & Poor's and Moody's provide bank credit ratings that influence economic stability and decision-making by stakeholders. Accurate and timely predictions support informed decision-making, regulatory actions, and investor protection. However, a complete interbank connection graph is often unavailable due to privacy concerns, complicating the direct application of Graph Neural Networks (GNNs) for rating prediction. our research utilizes persistent homology to construct a network that captures relationships among banks and combines this with a traditional lending network to create a heterogeneous network that integrates information from both sources, leading to improved predictions. Experiments on a global, real-world dataset validate the effectiveness of HTGNN. This research has implications for investors and regulatory bodies in enhancing proactive risk mitigation and the implementation of effective market interventions.The code can be find at https://github.com/Liu-Jun-Yi/HTGNN.
Authors: Yunqing Liu, Wenqi Fan, Xiaoyong Wei, Qing Li
Abstract: Proteins are central to biological systems, participating as building blocks across all forms of life. Despite advancements in understanding protein functions through protein sequence analysis, there remains potential for further exploration in integrating protein structural information. We argue that the structural information of proteins is not only limited to their 3D information but also encompasses information from amino acid molecules (local information) to protein-protein structure similarity (global information). To address this, we propose \textbf{GLProtein}, the first framework in protein pre-training that incorporates both global structural similarity and local amino acid details to enhance prediction accuracy and functional insights. GLProtein innovatively combines protein-masked modelling with triplet structure similarity scoring, protein 3D distance encoding and substructure-based amino acid molecule encoding. Experimental results demonstrate that GLProtein outperforms previous methods in several bioinformatics tasks, including predicting protein-protein interaction, contact prediction, and so on.
Authors: Zhiyuan Liu, Yicun Yang, Yaojie Zhang, Junjie Chen, Chang Zou, Qingyuan Wei, Shaobo Wang, Linfeng Zhang
Abstract: Autoregressive Models (ARMs) have long dominated the landscape of Large Language Models. Recently, a new paradigm has emerged in the form of diffusion-based Large Language Models (dLLMs), which generate text by iteratively denoising masked segments. This approach has shown significant advantages and potential. However, dLLMs suffer from high inference latency. Traditional ARM acceleration techniques, such as Key-Value caching, are incompatible with dLLMs due to their bidirectional attention mechanism. To address this specific challenge, our work begins with a key observation that dLLM inference involves a static prompt and a partially dynamic response, where most tokens remain stable across adjacent denoising steps. Based on this, we propose dLLM-Cache, a training-free adaptive caching framework that combines long-interval prompt caching with partial response updates guided by feature similarity. This design enables efficient reuse of intermediate computations without compromising model performance. Extensive experiments on representative dLLMs, including LLaDA 8B and Dream 7B, show that dLLM-Cache achieves up to 9.1 x speedup over standard inference without compromising output quality. Notably, our method brings dLLM inference latency close to that of ARMs under many settings. Codes are provided in the supplementary material and will be released publicly on GitHub.
Authors: Hanaa El Afia, Said Ohamouddou, Raddouane Chiheb, Abdellatif El Afia
Abstract: We introduce Jacobi-KAN-DGCNN, a framework that integrates Dynamic Graph Convolutional Neural Network (DGCNN) with Jacobi Kolmogorov-Arnold Networks (KAN) for the classification of three-dimensional point clouds. This method replaces Multi-Layer Perceptron (MLP) layers with adaptable univariate polynomial expansions within a streamlined DGCNN architecture, circumventing deep levels for both MLP and KAN to facilitate a layer-by-layer comparison. In comparative experiments on the ModelNet40 dataset, KAN layers employing Jacobi polynomials outperform the traditional linear layer-based DGCNN baseline in terms of accuracy and convergence speed, while maintaining parameter efficiency. Our results demonstrate that higher polynomial degrees do not automatically improve performance, highlighting the need for further theoretical and empirical investigation to fully understand the interactions between polynomial bases, degrees, and the mechanisms of graph-based learning.
Authors: Bozhi Sun, Seda Tierney, Jeffrey A. Feinstein, Frederick Damen, Alison L. Marsden, Daniele E. Schiavazzi
Abstract: Scheduling echocardiographic exams in a hospital presents significant challenges due to non-deterministic factors (e.g., patient no-shows, patient arrival times, diverse exam durations, etc.) and asymmetric resource constraints between fetal and non-fetal patient streams. To address these challenges, we first conducted extensive pre-processing on one week of operational data from the Echo Laboratory at Stanford University's Lucile Packard Children's Hospital, to estimate patient no-show probabilities and derive empirical distributions of arrival times and exam durations. Based on these inputs, we developed a discrete-event stochastic simulation model using SimPy, and integrate it with the open source Gymnasium Python library. As a baseline for policy optimization, we developed a comparative framework to evaluate on-the-fly versus reservation-based allocation strategies, in which different proportions of resources are reserved in advance. Considering a hospital configuration with a 1:6 ratio of fetal to non-fetal rooms and a 4:2 ratio of fetal to non-fetal sonographers, we show that on-the-fly allocation generally yields better performance, more effectively adapting to patient variability and resource constraints. Building on this foundation, we apply reinforcement learning (RL) to derive an approximated optimal dynamic allocation policy. This RL-based policy is benchmarked against the best-performing rule-based strategies, allowing us to quantify their differences and provide actionable insights for improving echo lab efficiency through intelligent, data-driven resource management.
Authors: Daniel Halpern, Evi Micha, Ariel D. Procaccia, Itai Shapira
Abstract: Current alignment pipelines presume a single, universal notion of desirable behavior. However, human preferences often diverge across users, contexts, and cultures. As a result, disagreement collapses into the majority signal and minority perspectives are discounted. To address this, we propose reflecting diverse human preferences through a distribution over multiple reward functions, each inducing a distinct aligned policy. The distribution is learned directly from pairwise preference without annotator identifiers or predefined groups. Instead, annotator disagreements are treated as informative soft labels. Our central criterion is pairwise calibration: for every pair of candidate responses, the proportion of reward functions preferring one response matches the fraction of annotators with that preference. We prove that even a small outlier-free ensemble can accurately represent diverse preference distributions. Empirically, we introduce and validate a practical training heuristic to learn such ensembles, and demonstrate its effectiveness through improved calibration, implying a more faithful representation of pluralistic values.
Authors: Yuanye Zhou, Zhaokun Wang, Kai Zhou, Hui Tang, Xiaofan Li
Abstract: Physics-informed neural networks (PINNs) have emerged as a powerful meshless tool for topology optimization, capable of simultaneously determining optimal topologies and physical solutions. However, conventional PINNs rely on density-based topology descriptions, which necessitate manual interpolation and limit their applicability to complex geometries. To address this, we propose Lagrangian topology-conscious PINNs (LT-PINNs), a novel framework for boundary-focused engineering optimization. By parameterizing the control variables of topology boundary curves as learnable parameters, LT-PINNs eliminate the need for manual interpolation and enable precise boundary determination. We further introduce specialized boundary condition loss function and topology loss function to ensure sharp and accurate boundary representations, even for intricate topologies. The accuracy and robustness of LT-PINNs are validated via two types of partial differential equations (PDEs), including elastic equation with Dirichlet boundary conditions and Laplace's equation with Neumann boundary conditions. Furthermore, we demonstrate effectiveness of LT-PINNs on more complex time-dependent and time-independent flow problems without relying on measurement data, and showcase their engineering application potential in flow velocity rearrangement, transforming a uniform upstream velocity into a sine-shaped downstream profile. The results demonstrate (1) LT-PINNs achieve substantial reductions in relative L2 errors compared with the state-of-art density topology-oriented PINNs (DT-PINNs), (2) LT-PINNs can handle arbitrary boundary conditions, making them suitable for a wide range of PDEs, and (3) LT-PINNs can infer clear topology boundaries without manual interpolation, especially for complex topologies.
Authors: Kefan Song, Amir Moeini, Peng Wang, Lei Gong, Rohan Chandra, Yanjun Qi, Shangtong Zhang
Abstract: Reinforcement learning (RL) is a human-designed framework for solving sequential decision making problems. In this work, we demonstrate that, surprisingly, RL emerges in LLM's (Large Language Model) inference time -- a phenomenon known as in-context RL (ICRL). Specifically, we propose a novel multi-round prompting framework called ICRL prompting. The goal is to prompt the LLM to complete a task. After the LLM generates a response at the current round, we give numerical scalar feedbacks for the response, called the rewards. At the next round, we prompt the LLM again with the same task and a context consisting of all previous responses and rewards. We observe that the quality of the LLM's response increases as the context grows. In other words, the LLM is able to maximize the scalar reward signal in the inference time, just like an RL algorithm. We evaluate ICRL prompting in three benchmarks (Game of 24, creative writing, and ScienceWorld) and demonstrate significant performance improvements over baseline methods such as Self-Refine and Reflexion. Surprisingly, in some experiments the reward signals are generated by the LLM itself, yet performance improvements are still observed from ICRL prompting, offering a promising paradigm for scaling test-time compute.
Authors: Zilang Chen
Abstract: Accurate and reproducible wine-quality assessment is critical for production control yet remains dominated by subjective, labour-intensive tasting panels. We present the first unified benchmark of five ensemble learners (Random Forest, Gradient Boosting, XGBoost, LightGBM, CatBoost) on the canonical Vinho Verde red- and white-wine datasets (1,599 and 4,898 instances, 11 physicochemical attributes). Our leakage-free workflow employs an 80:20 stratified train-test split, five-fold StratifiedGroupKFold within the training set, per-fold standardisation, SMOTE-Tomek resampling, inverse-frequency cost weighting, Optuna hyper-parameter search (120-200 trials per model) and a two-stage feature-selection refit. Final scores on untouched test sets are reported with weighted F1 as the headline metric. Gradient Boosting achieves the highest accuracy (weighted F1 0.693 +/- 0.028 for red and 0.664 +/- 0.016 for white), followed within three percentage points by Random Forest and XGBoost. Limiting each model to its five top-ranked variables lowers dimensionality by 55 percent while reducing weighted F1 by only 2.6 percentage points for red and 3.0 percentage points for white, indicating that alcohol, volatile acidity, sulphates, free SO2 and chlorides capture most predictive signal. Runtime profiling on an EPYC 9K84/H20 node reveals a steep efficiency gradient: Gradient Boosting averages 12 h per five-fold study, XGBoost and LightGBM require 2-3 h, CatBoost 1 h, and Random Forest under 50 min. We therefore recommend Random Forest as the most cost-effective production model, XGBoost and LightGBM as GPU-efficient alternatives, and Gradient Boosting as the accuracy ceiling for offline benchmarking. The fully documented pipeline and metric set provide a reproducible baseline for future work on imbalanced multi-class wine-quality prediction.
Authors: James Afful
Abstract: As machine learning systems are increasingly deployed in high-stakes domains such as criminal justice, finance, and healthcare, the demand for interpretable and trustworthy models has intensified. Despite the proliferation of local explanation techniques, including SHAP, LIME, and counterfactual methods, there exists no standardized, reproducible framework for their comparative evaluation, particularly in fairness-sensitive settings. We introduce ExplainBench, an open-source benchmarking suite for systematic evaluation of local model explanations across ethically consequential datasets. ExplainBench provides unified wrappers for popular explanation algorithms, integrates end-to-end pipelines for model training and explanation generation, and supports evaluation via fidelity, sparsity, and robustness metrics. The framework includes a Streamlit-based graphical interface for interactive exploration and is packaged as a Python module for seamless integration into research workflows. We demonstrate ExplainBench on datasets commonly used in fairness research, such as COMPAS, UCI Adult Income, and LendingClub, and showcase how different explanation methods behave under a shared experimental protocol. By enabling reproducible, comparative analysis of local explanations, ExplainBench advances the methodological foundations of interpretable machine learning and facilitates accountability in real-world AI systems.
Authors: Benjamin von Berg, Bernhard K. Aichernig
Abstract: AALpy is a well-established open-source automata learning library written in Python with a focus on active learning of systems with IO behavior. It provides a wide range of state-of-the-art algorithms for different automaton types ranging from fully deterministic to probabilistic automata. In this work, we present the recent addition of a generalized implementation of an important method from the domain of passive automata learning: state-merging in the red-blue framework. Using a common internal representation for different automaton types allows for a general and highly configurable implementation of the red-blue framework. We describe how to define and execute state-merging algorithms using AALpy, which reduces the implementation effort for state-merging algorithms mainly to the definition of compatibility criteria and scoring. This aids the implementation of both existing and novel algorithms. In particular, defining some existing state-merging algorithms from the literature with AALpy only takes a few lines of code.
Authors: Ali Murad, Bo Hui, Wei-Shinn Ku
Abstract: Federated Learning (FL) is a distributed framework for collaborative model training over large-scale distributed data, enabling higher performance while maintaining client data privacy. However, the nature of model aggregation at the centralized server can result in a performance drop in the presence of non-IID data across different clients. We remark that training a client locally on more data than necessary does not benefit the overall performance of all clients. In this paper, we devise a novel framework that leverages a Deep Reinforcement Learning (DRL) agent to select an optimized amount of data necessary to train a client model without oversharing information with the server. Starting without awareness of the client's performance, the DRL agent utilizes the change in training loss as a reward signal and learns to optimize the amount of training data necessary for improving the client's performance. Specifically, after each aggregation round, the DRL algorithm considers the local performance as the current state and outputs the optimized weights for each class, in the training data, to be used during the next round of local training. In doing so, the agent learns a policy that creates an optimized partition of the local training dataset during the FL rounds. After FL, the client utilizes the entire local training dataset to further enhance its performance on its own data distribution, mitigating the non-IID effects of aggregation. Through extensive experiments, we demonstrate that training FL clients through our algorithm results in superior performance on multiple benchmark datasets and FL frameworks. Our code is available at https://github.com/amuraddd/optimized_client_training.git.
URLs: https://github.com/amuraddd/optimized_client_training.git.
Authors: Gabriel Antonesi, Tudor Cioara, Ionut Anghel, Vasilis Michalakopoulos, Elissaios Sarmas, Liana Toderean
Abstract: Artificial intelligence (AI) has long promised to improve energy management in smart grids by enhancing situational awareness and supporting more effective decision-making. While traditional machine learning has demonstrated notable results in forecasting and optimization, it often struggles with generalization, situational awareness, and heterogeneous data integration. Recent advances in foundation models such as Transformer architecture and Large Language Models (LLMs) have demonstrated improved capabilities in modelling complex temporal and contextual relationships, as well as in multi-modal data fusion which is essential for most AI applications in the energy sector. In this review we synthesize the rapid expanding field of AI applications in the energy domain focusing on Transformers and LLMs. We examine the architectural foundations, domain-specific adaptations and practical implementations of transformer models across various forecasting and grid management tasks. We then explore the emerging role of LLMs in the field: adaptation and fine tuning for the energy sector, the type of tasks they are suited for, and the new challenges they introduce. Along the way, we highlight practical implementations, innovations, and areas where the research frontier is rapidly expanding. These recent developments reviewed underscore a broader trend: Generative AI (GenAI) is beginning to augment decision-making not only in high-level planning but also in day-to-day operations, from forecasting and grid balancing to workforce training and asset onboarding. Building on these developments, we introduce the concept of the Agentic Digital Twin, a next-generation model that integrates LLMs to bring autonomy, proactivity, and social interaction into digital twin-based energy management systems.
Authors: Jingyi Gu, Xuan Zhang, Guiling Wang
Abstract: Extreme events, such as market crashes, natural disasters, and pandemics, are rare but catastrophic, often triggering cascading failures across interconnected systems. Accurate prediction and early warning can help minimize losses and improve preparedness. While data-driven methods offer powerful capabilities for extreme event modeling, they require abundant training data, yet extreme event data is inherently scarce, creating a fundamental challenge. Synthetic data generation has emerged as a powerful solution. However, existing surveys focus on general data with privacy preservation emphasis, rather than extreme events' unique performance requirements. This survey provides the first overview of synthetic data generation for extreme events. We systematically review generative modeling techniques and large language models, particularly those enhanced by statistical theory as well as specialized training and sampling mechanisms to capture heavy-tailed distributions. We summarize benchmark datasets and introduce a tailored evaluation framework covering statistical, dependence, visual, and task-oriented metrics. A central contribution is our in-depth analysis of each metric's applicability in extremeness and domain-specific adaptations, providing actionable guidance for model evaluation in extreme settings. We categorize key application domains and identify underexplored areas like behavioral finance, wildfires, earthquakes, windstorms, and infectious outbreaks. Finally, we outline open challenges, providing a structured foundation for advancing synthetic rare-event research.
Authors: Yin Li
Abstract: Positional encodings are a core part of transformer-based models, enabling processing of sequential data without recurrence. This paper presents a theoretical framework to analyze how various positional encoding methods, including sinusoidal, learned, relative, and bias-based methods like Attention with Linear Biases (ALiBi), impact a transformer's expressiveness, generalization ability, and extrapolation to longer sequences. Expressiveness is defined via function approximation, generalization bounds are established using Rademacher complexity, and new encoding methods based on orthogonal functions, such as wavelets and Legendre polynomials, are proposed. The extrapolation capacity of existing and proposed encodings is analyzed, extending ALiBi's biasing approach to a unified theoretical context. Experimental evaluation on synthetic sequence-to-sequence tasks shows that orthogonal transform-based encodings outperform traditional sinusoidal encodings in generalization and extrapolation. This work addresses a critical gap in transformer theory, providing insights for design choices in natural language processing, computer vision, and other transformer applications.
Authors: Paul Fogel (Data Services, ForvisMazars, Courbevoie, France), Christophe Geissler (Data Services, ForvisMazars, Courbevoie, France), George Luta (Department of Biostatistics, Bioinformatics and Biomathematics, Georgetown University Medical Center, Washington, DC, USA)
Abstract: The interpretation of the results of survival analysis often benefits from latent factor representations of baseline covariates. However, existing methods, such as Nonnegative Matrix Factorization (NMF), do not incorporate survival information, limiting their predictive power. We present CoxNTF, a novel approach that uses non-negative tensor factorization (NTF) to derive meaningful latent representations that are closely associated with survival outcomes. CoxNTF constructs a weighted covariate tensor in which survival probabilities derived from the Coxnet model are used to guide the tensorization process. Our results show that CoxNTF achieves survival prediction performance comparable to using Coxnet with the original covariates, while providing a structured and interpretable clustering framework. In addition, the new approach effectively handles feature redundancy, making it a powerful tool for joint clustering and prediction in survival analysis.
Authors: Junming Wang, Yi Shi
Abstract: Discovering novel classes in open-world settings is crucial for real-world applications. Traditional explicit representations, such as object descriptors or 3D segmentation maps, are constrained by their discrete, hole-prone, and noisy nature, which hinders accurate novel class discovery. To address these challenges, we introduce NeurNCD, the first versatile and data-efficient framework for novel class discovery that employs the meticulously designed Embedding-NeRF model combined with KL divergence as a substitute for traditional explicit 3D segmentation maps to aggregate semantic embedding and entropy in visual embedding space. NeurNCD also integrates several key components, including feature query, feature modulation and clustering, facilitating efficient feature augmentation and information exchange between the pre-trained semantic segmentation network and implicit neural representations. As a result, our framework achieves superior segmentation performance in both open and closed-world settings without relying on densely labelled datasets for supervised training or human interaction to generate sparse label supervision. Extensive experiments demonstrate that our method significantly outperforms state-of-the-art approaches on the NYUv2 and Replica datasets.
Authors: Luis Pinto
Abstract: Pretrained molecular encoders have become indispensable in computational chemistry for tasks such as property prediction and molecular generation. However, the standard practice of relying solely on final-layer embeddings for downstream tasks may discard valuable information. In this work, we challenge this convention by conducting a comprehensive layer-wise analysis of five diverse molecular encoders across 22 ADMET property prediction tasks. Our results demonstrate that embeddings from intermediate layers consistently outperform final-layer representations. Specifically, using fixed embeddings from the optimal intermediate layers improved downstream performance by an average of 5.4%, reaching gains up to 28.6%. Furthermore, finetuning up to these intermediate layers yielded even greater average improvements of 8.5%, with performance increases as high as 40.8%, achieving new state-of-the-art results on several benchmarks. Additionally, a strong positive correlation between fixed embedding performance and finetuning outcomes supports an efficient evaluate-then-finetune approach, enabling identification of optimal layers with reduced computational cost. These findings highlight the importance of exploring the full representational depth of molecular encoders to achieve substantial performance improvements and computational efficiency. The code is made publicly available at https://github.com/luispintoc/Unlocking-Chemical-Insights.
URLs: https://github.com/luispintoc/Unlocking-Chemical-Insights.
Authors: Ruizhong Qiu, Gaotang Li, Tianxin Wei, Jingrui He, Hanghang Tong
Abstract: Existing safety assurance research has primarily focused on training-phase alignment to instill safe behaviors into LLMs. However, recent studies have exposed these methods' susceptibility to diverse jailbreak attacks. Concurrently, inference scaling has significantly advanced LLM reasoning capabilities but remains unexplored in the context of safety assurance. Addressing this gap, our work pioneers inference scaling for robust and effective LLM safety against emerging threats. We reveal that conventional inference scaling techniques, despite their success in reasoning tasks, perform poorly in safety contexts, even falling short of basic approaches like Best-of-N Sampling. We attribute this inefficiency to a newly identified challenge, the exploration--efficiency dilemma, arising from the high computational overhead associated with frequent process reward model (PRM) evaluations. To overcome this dilemma, we propose SAFFRON, a novel inference scaling paradigm tailored explicitly for safety assurance. Central to our approach is the introduction of a multifurcation reward model (MRM) that significantly reduces the required number of reward model evaluations. To operationalize this paradigm, we further propose: (i) a partial supervision training objective for MRM, (ii) a conservative exploration constraint to prevent out-of-distribution explorations, and (iii) a Trie-based key--value caching strategy that facilitates cache sharing across sequences during tree search. Extensive experiments validate the effectiveness of our method. Additionally, we publicly release our trained multifurcation reward model (Saffron-1) and the accompanying token-level safety reward dataset (Safety4M) to accelerate future research in LLM safety. Our code, model, and data are publicly available at https://github.com/q-rz/saffron , and our project homepage is at https://q-rz.github.io/p/saffron .
URLs: https://github.com/q-rz/saffron, https://q-rz.github.io/p/saffron
Authors: Abrar Majeedi, Viswanatha Reddy Gajjala, Satya Sai Srinath Namburi GNVV, Nada Magdi Elkordi, Yin Li
Abstract: Real-world time series are often governed by complex nonlinear dynamics. Understanding these underlying dynamics is crucial for precise future prediction. While deep learning has achieved major success in time series forecasting, many existing approaches do not explicitly model the dynamics. To bridge this gap, we introduce DeepEDM, a framework that integrates nonlinear dynamical systems modeling with deep neural networks. Inspired by empirical dynamic modeling (EDM) and rooted in Takens' theorem, DeepEDM presents a novel deep model that learns a latent space from time-delayed embeddings, and employs kernel regression to approximate the underlying dynamics, while leveraging efficient implementation of softmax attention and allowing for accurate prediction of future time steps. To evaluate our method, we conduct comprehensive experiments on synthetic data of nonlinear dynamical systems as well as real-world time series across domains. Our results show that DeepEDM is robust to input noise, and outperforms state-of-the-art methods in forecasting accuracy. Our code is available at: https://abrarmajeedi.github.io/deep_edm.
Authors: Antonio Jes\'us Banegas-Luna, Horacio P\'erez-S\'anchez, Carlos Mart\'inez-Cort\'es
Abstract: While predictive accuracy is often prioritized in machine learning (ML) models, interpretability remains essential in scientific and high-stakes domains. However, diverse interpretability algorithms frequently yield conflicting explanations, highlighting the need for consensus to harmonize results. In this study, six ML models were trained on six synthetic datasets with known ground truths, utilizing various model-agnostic interpretability techniques. Consensus explanations were generated using established methods and a novel approach: WISCA (Weighted Scaled Consensus Attributions), which integrates class probability and normalized attributions. WISCA consistently aligned with the most reliable individual method, underscoring the value of robust consensus strategies in improving explanation reliability.
Authors: Ruitao Chen, Mozhang Guo, Jinge Li
Abstract: Automated driving (AD) has substantially improved vehicle safety and driving comfort, but their impact on passenger well-being, particularly infant sleep, is not sufficiently studied. Sudden acceleration, abrupt braking, and sharp maneuvers can disrupt infant sleep, compromising both passenger comfort and parental convenience. To solve this problem, this paper explores the integration of reinforcement learning (RL) within AD to personalize driving behavior and optimally balance occupant comfort and travel efficiency. In particular, we propose an intelligent cruise control framework that adapts to varying driving conditions to enhance infant sleep quality by effectively synergizing wearable sensing and vehicle data. Long short-term memory (LSTM) and transformer-based neural networks are integrated with RL to model the relationship between driving behavior and infant sleep quality under diverse traffic and road conditions. Based on the sleep quality indicators from the wearable sensors, driving action data from vehicle controllers, and map data from map applications, the model dynamically computes the optimal driving aggressiveness level, which is subsequently translated into specific AD control strategies, e.g., the magnitude and frequency of acceleration, lane change, and overtaking. Simulation results demonstrate that the proposed solution significantly improves infant sleep quality compared to baseline methods, while preserving desirable travel efficiency.
Authors: Zhiyuan Zhao, Juntong Ni, Shangqing Xu, Haoxin Liu, Wei Jin, B. Aditya Prakash
Abstract: Time-series forecasting is an essential task with wide real-world applications across domains. While recent advances in deep learning have enabled time-series forecasting models with accurate predictions, there remains considerable debate over which architectures and design components, such as series decomposition or normalization, are most effective under varying conditions. Existing benchmarks primarily evaluate models at a high level, offering limited insight into why certain designs work better. To mitigate this gap, we propose TimeRecipe, a unified benchmarking framework that systematically evaluates time-series forecasting methods at the module level. TimeRecipe conducts over 10,000 experiments to assess the effectiveness of individual components across a diverse range of datasets, forecasting horizons, and task settings. Our results reveal that exhaustive exploration of the design space can yield models that outperform existing state-of-the-art methods and uncover meaningful intuitions linking specific design choices to forecasting scenarios. Furthermore, we release a practical toolkit within TimeRecipe that recommends suitable model architectures based on these empirical insights. The benchmark is available at: https://github.com/AdityaLab/TimeRecipe.
Authors: Umit Yigit Basaran, Sk Miraj Ahmed, Amit Roy-Chowdhury, Basak Guler
Abstract: With the growing adoption of data privacy regulations, the ability to erase private or copyrighted information from trained models has become a crucial requirement. Traditional unlearning methods often assume access to the complete training dataset, which is unrealistic in scenarios where the source data is no longer available. To address this challenge, we propose a certified unlearning framework that enables effective data removal \final{without access to the original training data samples}. Our approach utilizes a surrogate dataset that approximates the statistical properties of the source data, allowing for controlled noise scaling based on the statistical distance between the two. \updated{While our theoretical guarantees assume knowledge of the exact statistical distance, practical implementations typically approximate this distance, resulting in potentially weaker but still meaningful privacy guarantees.} This ensures strong guarantees on the model's behavior post-unlearning while maintaining its overall utility. We establish theoretical bounds, introduce practical noise calibration techniques, and validate our method through extensive experiments on both synthetic and real-world datasets. The results demonstrate the effectiveness and reliability of our approach in privacy-sensitive settings.
Authors: Pratiksha Thaker, Neil Kale, Zhiwei Steven Wu, Virginia Smith
Abstract: Shadow model attacks are the state-of-the-art approach for membership inference attacks on machine learning models. However, these attacks typically assume an adversary has access to a background (nonmember) data distribution that matches the distribution the target model was trained on. We initiate a study of membership inference attacks where the adversary or auditor cannot access an entire subclass from the distribution -- a more extreme but realistic version of distribution shift than has been studied previously. In this setting, we first show that the performance of shadow model attacks degrades catastrophically, and then demonstrate the promise of another approach, quantile regression, that does not have the same limitations. We show that quantile regression attacks consistently outperform shadow model attacks in the class dropout setting -- for example, quantile regression attacks achieve up to 11$\times$ the TPR of shadow models on the unseen class on CIFAR-100, and achieve nontrivial TPR on ImageNet even with 90% of training classes removed. We also provide a theoretical model that illustrates the potential and limitations of this approach.
Authors: Daniel Kunin, Giovanni Luca Marchetti, Feng Chen, Dhruva Karkada, James B. Simon, Michael R. DeWeese, Surya Ganguli, Nina Miolane
Abstract: What features neural networks learn, and how, remains an open question. In this paper, we introduce Alternating Gradient Flows (AGF), an algorithmic framework that describes the dynamics of feature learning in two-layer networks trained from small initialization. Prior works have shown that gradient flow in this regime exhibits a staircase-like loss curve, alternating between plateaus where neurons slowly align to useful directions and sharp drops where neurons rapidly grow in norm. AGF approximates this behavior as an alternating two-step process: maximizing a utility function over dormant neurons and minimizing a cost function over active ones. AGF begins with all neurons dormant. At each round, a dormant neuron activates, triggering the acquisition of a feature and a drop in the loss. AGF quantifies the order, timing, and magnitude of these drops, matching experiments across architectures. We show that AGF unifies and extends existing saddle-to-saddle analyses in fully connected linear networks and attention-only linear transformers, where the learned features are singular modes and principal components, respectively. In diagonal linear networks, we prove AGF converges to gradient flow in the limit of vanishing initialization. Applying AGF to quadratic networks trained to perform modular addition, we give the first complete characterization of the training dynamics, revealing that networks learn Fourier features in decreasing order of coefficient magnitude. Altogether, AGF offers a promising step towards understanding feature learning in neural networks.
Authors: Alex Havrilla, Edward Hughes, Mikayel Samvelyan, Jacob Abernethy
Abstract: Large language model (LLM) driven synthetic data generation has emerged as a powerful method for improving model reasoning capabilities. However, most methods either distill large state-of-the-art models into small students or use natural ground-truth problem statements to guarantee problem statement quality. This limits the scalability of these approaches to more complex and diverse problem domains. To address this, we present SPARQ: Synthetic Problem Generation for Reasoning via Quality-Diversity Algorithms, a novel approach for generating high-quality and diverse synthetic math problem and solution pairs using only a single model by measuring a problem's solve-rate: a proxy for problem difficulty. Starting from a seed dataset of 7.5K samples, we generate over 20 million new problem-solution pairs. We show that filtering the generated data by difficulty and then fine-tuning the same model on the resulting data improves relative model performance by up to 24\%. Additionally, we conduct ablations studying the impact of synthetic data quantity, quality and diversity on model generalization. We find that higher quality, as measured by problem difficulty, facilitates better in-distribution performance. Further, while generating diverse synthetic data does not as strongly benefit in-distribution performance, filtering for more diverse data facilitates more robust OOD generalization. We also confirm the existence of model and data scaling laws for synthetically generated problems, which positively benefit downstream model generalization.
Authors: Ran Levinstein, Amit Attia, Matan Schliserman, Uri Sherman, Tomer Koren, Daniel Soudry, Itay Evron
Abstract: We study realizable continual linear regression under random task orderings, a common setting for developing continual learning theory. In this setup, the worst-case expected loss after $k$ learning iterations admits a lower bound of $\Omega(1/k)$. However, prior work using an unregularized scheme has only established an upper bound of $O(1/k^{1/4})$, leaving a significant gap. Our paper proves that this gap can be narrowed, or even closed, using two frequently used regularization schemes: (1) explicit isotropic $\ell_2$ regularization, and (2) implicit regularization via finite step budgets. We show that these approaches, which are used in practice to mitigate forgetting, reduce to stochastic gradient descent (SGD) on carefully defined surrogate losses. Through this lens, we identify a fixed regularization strength that yields a near-optimal rate of $O(\log k / k)$. Moreover, formalizing and analyzing a generalized variant of SGD for time-varying functions, we derive an increasing regularization strength schedule that provably achieves an optimal rate of $O(1/k)$. This suggests that schedules that increase the regularization coefficient or decrease the number of steps per task are beneficial, at least in the worst case.
Authors: Keisuke Sugiura, Hiroki Matsutani
Abstract: Training deep neural networks (DNNs) requires significantly more computation and memory than inference, making runtime adaptation of DNNs challenging on resource-limited IoT platforms. We propose InstantFT, an FPGA-based method for ultra-fast CNN fine-tuning on IoT devices, by optimizing the forward and backward computations in parameter-efficient fine-tuning (PEFT). Experiments on datasets with concept drift demonstrate that InstantFT fine-tunes a pre-trained CNN 17.4x faster than existing Low-Rank Adaptation (LoRA)-based approaches, while achieving comparable accuracy. Our FPGA-based InstantFT reduces the fine-tuning time to just 0.36s and improves energy-efficiency by 16.3x, enabling on-the-fly adaptation of CNNs to non-stationary data distributions.
Authors: Shulun Chen, Runlong Zhou, Zihan Zhang, Maryam Fazel, Simon S. Du
Abstract: We consider the gap-dependent regret bounds for episodic MDPs. We show that the Monotonic Value Propagation (MVP) algorithm achieves a variance-aware gap-dependent regret bound of $$\tilde{O}\left(\left(\sum_{\Delta_h(s,a)>0} \frac{H^2 \log K \land \mathtt{Var}_{\max}^{\text{c}}}{\Delta_h(s,a)} +\sum_{\Delta_h(s,a)=0}\frac{ H^2 \land \mathtt{Var}_{\max}^{\text{c}}}{\Delta_{\mathrm{min}}} + SAH^4 (S \lor H) \right) \log K\right),$$ where $H$ is the planning horizon, $S$ is the number of states, $A$ is the number of actions, and $K$ is the number of episodes. Here, $\Delta_h(s,a) =V_h^* (a) - Q_h^* (s, a)$ represents the suboptimality gap and $\Delta_{\mathrm{min}} := \min_{\Delta_h (s,a) > 0} \Delta_h(s,a)$. The term $\mathtt{Var}_{\max}^{\text{c}}$ denotes the maximum conditional total variance, calculated as the maximum over all $(\pi, h, s)$ tuples of the expected total variance under policy $\pi$ conditioned on trajectories visiting state $s$ at step $h$. $\mathtt{Var}_{\max}^{\text{c}}$ characterizes the maximum randomness encountered when learning any $(h, s)$ pair. Our result stems from a novel analysis of the weighted sum of the suboptimality gap and can be potentially adapted for other algorithms. To complement the study, we establish a lower bound of $$\Omega \left( \sum_{\Delta_h(s,a)>0} \frac{H^2 \land \mathtt{Var}_{\max}^{\text{c}}}{\Delta_h(s,a)}\cdot \log K\right),$$ demonstrating the necessity of dependence on $\mathtt{Var}_{\max}^{\text{c}}$ even when the maximum unconditional total variance (without conditioning on $(h, s)$) approaches zero.
Authors: Zijiang Yan, Hao Zhou, Jianhua Pei, Hina Tabassum
Abstract: Unmanned aerial vehicles (UAVs) have been widely adopted in various real-world applications. However, the control and optimization of multi-UAV systems remain a significant challenge, particularly in dynamic and constrained environments. This work explores the joint motion and communication control of multiple UAVs operating within integrated terrestrial and non-terrestrial networks that include high-altitude platform stations (HAPS). Specifically, we consider an aerial highway scenario in which UAVs must accelerate, decelerate, and change lanes to avoid collisions and maintain overall traffic flow. Different from existing studies, we propose a novel hierarchical and collaborative method based on large language models (LLMs). In our approach, an LLM deployed on the HAPS performs UAV access control, while another LLM onboard each UAV handles motion planning and control. This LLM-based framework leverages the rich knowledge embedded in pre-trained models to enable both high-level strategic planning and low-level tactical decisions. This knowledge-driven paradigm holds great potential for the development of next-generation 3D aerial highway systems. Experimental results demonstrate that our proposed collaborative LLM-based method achieves higher system rewards, lower operational costs, and significantly reduced UAV collision rates compared to baseline approaches.
Authors: Atefeh Gilani, Naima Tasnim, Lalitha Sankar, Oliver Kosut
Abstract: Differentially private stochastic gradient descent (DP-SGD) is the most widely used method for training machine learning models with provable privacy guarantees. A key challenge in DP-SGD is setting the per-sample gradient clipping threshold, which significantly affects the trade-off between privacy and utility. While recent adaptive methods improve performance by adjusting this threshold during training, they operate in the standard coordinate system and fail to account for correlations across the coordinates of the gradient. We propose GeoClip, a geometry-aware framework that clips and perturbs gradients in a transformed basis aligned with the geometry of the gradient distribution. GeoClip adaptively estimates this transformation using only previously released noisy gradients, incurring no additional privacy cost. We provide convergence guarantees for GeoClip and derive a closed-form solution for the optimal transformation that minimizes the amount of noise added while keeping the probability of gradient clipping under control. Experiments on both tabular and image datasets demonstrate that GeoClip consistently outperforms existing adaptive clipping methods under the same privacy budget.
Authors: Long Dang, Thushari Hapuarachchi, Kaiqi Xiong, Yi Li
Abstract: As the development of autonomous and connected vehicles advances, the complexity of modern vehicles increases, with numerous Electronic Control Units (ECUs) integrated into the system. In an in-vehicle network, these ECUs communicate with one another using an standard protocol called Controller Area Network (CAN). Securing communication among ECUs plays a vital role in maintaining the safety and security of the vehicle. This paper proposes a robust SDN-based False Data Detection and Mitigation System (FDDMS) for in-vehicle networks. Leveraging the unique capabilities of Software-Defined Networking (SDN), FDDMS is designed to monitor and detect false data injection attacks in real-time. Specifically, we focus on brake-related ECUs within an SDN-enabled in-vehicle network. First, we decode raw CAN data to create an attack model that illustrates how false data can be injected into the system. Then, FDDMS, incorporating a Long Short Term Memory (LSTM)-based detection model, is used to identify false data injection attacks. We further propose an effective variant of DeepFool attack to evaluate the model's robustness. To countermeasure the impacts of four adversarial attacks including Fast gradient descent method, Basic iterative method, DeepFool, and the DeepFool variant, we further enhance a re-training technique method with a threshold based selection strategy. Finally, a mitigation scheme is implemented to redirect attack traffic by dynamically updating flow rules through SDN. Our experimental results show that the proposed FDDMS is robust against adversarial attacks and effectively detects and mitigates false data injection attacks in real-time.
Authors: Atamert Rahma, Chinmay Datar, Ana Cukarska, Felix Dietrich
Abstract: Learning dynamical systems that respect physical symmetries and constraints remains a fundamental challenge in data-driven modeling. Integrating physical laws with graph neural networks facilitates principled modeling of complex N-body dynamics and yields accurate and permutation-invariant models. However, training graph neural networks with iterative, gradient-based optimization algorithms (e.g., Adam, RMSProp, LBFGS) often leads to slow training, especially for large, complex systems. In comparison to 15 different optimizers, we demonstrate that Hamiltonian Graph Networks (HGN) can be trained up to 600x faster--but with comparable accuracy--by replacing iterative optimization with random feature-based parameter construction. We show robust performance in diverse simulations, including N-body mass-spring systems in up to 3 dimensions with different geometries, while retaining essential physical invariances with respect to permutation, rotation, and translation. We reveal that even when trained on minimal 8-node systems, the model can generalize in a zero-shot manner to systems as large as 4096 nodes without retraining. Our work challenges the dominance of iterative gradient-descent-based optimization algorithms for training neural network models for physical systems.
Authors: Mattie Ji, Amauri H. Souza, Vikas Garg
Abstract: Including intricate topological information (e.g., cycles) provably enhances the expressivity of message-passing graph neural networks (GNNs) beyond the Weisfeiler-Leman (WL) hierarchy. Consequently, Persistent Homology (PH) methods are increasingly employed for graph representation learning. In this context, recent works have proposed decorating classical PH diagrams with vertex and edge features for improved expressivity. However, due to their dependence on features, these methods still fail to capture basic graph structural information. In this paper, we propose SpectRe -- a new topological descriptor for graphs that integrates spectral information into PH diagrams. Notably, SpectRe is strictly more expressive than existing descriptors on graphs. We also introduce notions of global and local stability to analyze existing descriptors and establish that SpectRe is locally stable. Finally, experiments on synthetic and real-world datasets demonstrate the effectiveness of SpectRe and its potential to enhance the capabilities of graph models in relevant learning tasks.
Authors: Adarsh Prasad Behera, Jaya Prakash Champati, Roberto Morabito, Sasu Tarkoma, James Gross
Abstract: Recent progress in Language Models (LMs) has dramatically advanced the field of natural language processing (NLP), excelling at tasks like text generation, summarization, and question answering. However, their inference remains computationally expensive and energy intensive, especially in settings with limited hardware, power, or bandwidth. This makes it difficult to deploy LMs in mobile, edge, or cost sensitive environments. To address these challenges, recent approaches have introduced multi LLM intelligent model selection strategies that dynamically allocate computational resources based on query complexity -- using lightweight models for simpler queries and escalating to larger models only when necessary. This survey explores two complementary strategies for efficient LLM inference: (i) routing, which selects the most suitable model based on the query, and (ii) cascading or hierarchical inference (HI), which escalates queries through a sequence of models until a confident response is found. Both approaches aim to reduce computation by using lightweight models for simpler tasks while offloading only when needed. We provide a comparative analysis of these techniques across key performance metrics, discuss benchmarking efforts, and outline open challenges. Finally, we outline future research directions to enable faster response times, adaptive model selection based on task complexity, and scalable deployment across heterogeneous environments, making LLM based systems more efficient and accessible for real world applications.
Authors: Diaaeldin Taha, James Chapman, Marzieh Eidi, Karel Devriendt, Guido Mont\'ufar
Abstract: Topological deep learning (TDL) has emerged as a powerful tool for modeling higher-order interactions in relational data. However, phenomena such as oversquashing in topological message-passing remain understudied and lack theoretical analysis. We propose a unifying axiomatic framework that bridges graph and topological message-passing by viewing simplicial and cellular complexes and their message-passing schemes through the lens of relational structures. This approach extends graph-theoretic results and algorithms to higher-order structures, facilitating the analysis and mitigation of oversquashing in topological message-passing networks. Through theoretical analysis and empirical studies on simplicial networks, we demonstrate the potential of this framework to advance TDL.
Authors: Mo Zhou, Weihang Xu, Maryam Fazel, Simon S. Du
Abstract: Learning Gaussian Mixture Models (GMMs) is a fundamental problem in machine learning, with the Expectation-Maximization (EM) algorithm and its popular variant gradient EM being arguably the most widely used algorithms in practice. In the exact-parameterized setting, where both the ground truth GMM and the learning model have the same number of components $m$, a vast line of work has aimed to establish rigorous recovery guarantees for EM. However, global convergence has only been proven for the case of $m=2$, and EM is known to fail to recover the ground truth when $m\geq 3$. In this paper, we consider the $\textit{over-parameterized}$ setting, where the learning model uses $n>m$ components to fit an $m$-component ground truth GMM. In contrast to the exact-parameterized case, we provide a rigorous global convergence guarantee for gradient EM. Specifically, for any well separated GMMs in general position, we prove that with only mild over-parameterization $n = \Omega(m\log m)$, randomly initialized gradient EM converges globally to the ground truth at a polynomial rate with polynomial samples. Our analysis proceeds in two stages and introduces a suite of novel tools for Gaussian Mixture analysis. We use Hermite polynomials to study the dynamics of gradient EM and employ tensor decomposition to characterize the geometric landscape of the likelihood loss. This is the first global convergence and recovery result for EM or Gradient EM beyond the special case of $m=2$.
Authors: Yuanjie Shi, Hooman Shahrokhi, Xuesong Jia, Xiongzhi Chen, Janardhan Rao Doppa, Yan Yan
Abstract: Conformal prediction (CP) is a promising uncertainty quantification framework which works as a wrapper around a black-box classifier to construct prediction sets (i.e., subset of candidate classes) with provable guarantees. However, standard calibration methods for CP tend to produce large prediction sets which makes them less useful in practice. This paper considers the problem of integrating conformal principles into the training process of deep classifiers to directly minimize the size of prediction sets. We formulate conformal training as a bilevel optimization problem and propose the {\em Direct Prediction Set Minimization (DPSM)} algorithm to solve it. The key insight behind DPSM is to minimize a measure of the prediction set size (upper level) that is conditioned on the learned quantile of conformity scores (lower level). We analyze that DPSM has a learning bound of $O(1/\sqrt{n})$ (with $n$ training samples), while prior conformal training methods based on stochastic approximation for the quantile has a bound of $\Omega(1/s)$ (with batch size $s$ and typically $s \ll \sqrt{n}$). Experiments on various benchmark datasets and deep models show that DPSM significantly outperforms the best prior conformal training baseline with $20.46\%\downarrow$ in the prediction set size and validates our theory.
Authors: Joseph T Colonel, Carolyn Hagler, Guiselle Wismer, Laura Curtis, Jacqueline Becker, Juan Wisnivesky, Alex Federman, Gaurav Pandey
Abstract: Several machine learning algorithms have been developed for the prediction of Alzheimer's disease and related dementia (ADRD) from spontaneous speech. However, none of these algorithms have been translated for the prediction of broader cognitive impairment (CI), which in some cases is a precursor and risk factor of ADRD. In this paper, we evaluated several speech-based open-source methods originally proposed for the prediction of ADRD, as well as methods from multimodal sentiment analysis for the task of predicting CI from patient audio recordings. Results demonstrated that multimodal methods outperformed unimodal ones for CI prediction, and that acoustics-based approaches performed better than linguistics-based ones. Specifically, interpretable acoustic features relating to affect and prosody were found to significantly outperform BERT-based linguistic features and interpretable linguistic features, respectively. All the code developed for this study is available at https://github.com/JTColonel/catch.
Authors: Xinyu Luo, Cedar Site Bai, Bolian Li, Petros Drineas, Ruqi Zhang, Brian Bullins
Abstract: While popular optimization methods such as SGD, AdamW, and Lion depend on steepest descent updates in either $\ell_2$ or $\ell_\infty$ norms, there remains a critical gap in handling the non-Euclidean structure observed in modern deep networks training. In this work, we address this need by introducing a new accelerated $\ell_p$ steepest descent algorithm, called Stacey, which uses interpolated primal-dual iterate sequences to effectively navigate non-Euclidean smooth optimization tasks. In addition to providing novel theoretical guarantees for the foundations of our algorithm, we empirically compare our approach against these popular methods on tasks including image classification and language model (LLM) pretraining, demonstrating both faster convergence and higher final accuracy. We further evaluate different values of $p$ across various models and datasets, underscoring the importance and efficiency of non-Euclidean approaches over standard Euclidean methods. Code can be found at https://github.com/xinyuluo8561/Stacey .
Authors: Shubham Parashar, Shurui Gui, Xiner Li, Hongyi Ling, Sushil Vemuri, Blake Olson, Eric Li, Yu Zhang, James Caverlee, Dileep Kalathil, Shuiwang Ji
Abstract: We aim to improve the reasoning capabilities of language models via reinforcement learning (RL). Recent RL post-trained models like DeepSeek-R1 have demonstrated reasoning abilities on mathematical and coding tasks. However, prior studies suggest that using RL alone to improve reasoning on inherently difficult tasks is less effective. Here, we draw inspiration from curriculum learning and propose to schedule tasks from easy to hard (E2H), allowing LLMs to build reasoning skills gradually. Our method is termed E2H Reasoner. Empirically, we observe that, although easy tasks are important initially, fading them out through appropriate scheduling is essential in preventing overfitting. Theoretically, we establish convergence guarantees for E2H Reasoner within an approximate policy iteration framework. We derive finite-sample complexity bounds and show that when tasks are appropriately decomposed and conditioned, learning through curriculum stages requires fewer total samples than direct learning. Experiments across multiple domains show that E2H Reasoner significantly improves the reasoning ability of small LLMs (1.5B to 3B), which otherwise struggle when trained with vanilla RL alone, highlighting the effectiveness of our method.
Authors: Chi-Sheng Chen
Abstract: Recent advancements in quantum machine learning have shown promise in enhancing classical neural network architectures, particularly in domains involving complex, high-dimensional data. Building upon prior work in temporal sequence modeling, this paper introduces Vision-QRWKV, a hybrid quantum-classical extension of the Receptance Weighted Key Value (RWKV) architecture, applied for the first time to image classification tasks. By integrating a variational quantum circuit (VQC) into the channel mixing component of RWKV, our model aims to improve nonlinear feature transformation and enhance the expressive capacity of visual representations. We evaluate both classical and quantum RWKV models on a diverse collection of 14 medical and standard image classification benchmarks, including MedMNIST datasets, MNIST, and FashionMNIST. Our results demonstrate that the quantum-enhanced model outperforms its classical counterpart on a majority of datasets, particularly those with subtle or noisy class distinctions (e.g., ChestMNIST, RetinaMNIST, BloodMNIST). This study represents the first systematic application of quantum-enhanced RWKV in the visual domain, offering insights into the architectural trade-offs and future potential of quantum models for lightweight and efficient vision tasks.
Authors: Olimjon Toirov, Wei Yu
Abstract: Non-Intrusive Load Monitoring (NILM) identifies the operating status and energy consumption of each electrical device in the circuit by analyzing the electrical signals at the bus, which is of great significance for smart power management. However, the complex and changeable load combinations and application environments lead to the challenges of poor feature robustness and insufficient model generalization of traditional NILM methods. To this end, this paper proposes a new non-intrusive load monitoring method that integrates "image load signature" and continual learning. This method converts multi-dimensional power signals such as current, voltage, and power factor into visual image load feature signatures, and combines deep convolutional neural networks to realize the identification and classification of multiple devices; at the same time, self-supervised pre-training is introduced to improve feature generalization, and continual online learning strategies are used to overcome model forgetting to adapt to the emergence of new loads. This paper conducts a large number of experiments on high-sampling rate load datasets, and compares a variety of existing methods and model variants. The results show that the proposed method has achieved significant improvements in recognition accuracy.
Authors: Chong You, Kan Wu, Zhipeng Jia, Lin Chen, Srinadh Bhojanapalli, Jiaxian Guo, Utku Evci, Jan Wassenberg, Praneeth Netrapalli, Jeremiah J. Willcock, Suvinay Subramanian, Felix Chern, Alek Andreev, Shreya Pathak, Felix Yu, Prateek Jain, David E. Culler, Henry M. Levy, Sanjiv Kumar
Abstract: The discovery of the lazy neuron phenomenon in trained Transformers, where the vast majority of neurons in their feed-forward networks (FFN) are inactive for each token, has spurred tremendous interests in activation sparsity for enhancing large model efficiency. While notable progress has been made in translating such sparsity to wall-time benefits, modern Transformers have moved away from the ReLU activation function crucial to this phenomenon. Existing efforts on re-introducing activation sparsity often degrade model quality, increase parameter count, complicate or slow down training. Sparse attention, the application of sparse activation to the attention mechanism, often faces similar challenges. This paper introduces the Spark Transformer, a novel architecture that achieves a high level of activation sparsity in both FFN and the attention mechanism while maintaining model quality, parameter count, and standard training procedures. Our method realizes sparsity via top-k masking for explicit control over sparsity level. Crucially, we introduce statistical top-k, a hardware-accelerator-friendly, linear-time approximate algorithm that avoids costly sorting and mitigates significant training slowdown from standard top-$k$ operators. Furthermore, Spark Transformer reallocates existing FFN parameters and attention key embeddings to form a low-cost predictor for identifying activated entries. This design not only mitigates quality loss from enforced sparsity, but also enhances wall-time benefit. Pretrained with the Gemma-2 recipe, Spark Transformer demonstrates competitive performance on standard benchmarks while exhibiting significant sparsity: only 8% of FFN neurons are activated, and each token attends to a maximum of 256 tokens. This sparsity translates to a 2.5x reduction in FLOPs, leading to decoding wall-time speedups of up to 1.79x on CPU and 1.40x on GPU.
Authors: Yishan Shen, Yuyang Ye, Hui Xiong, Yong Chen
Abstract: Dynamic treatment regimes (DTRs) are critical to precision medicine, optimizing long-term outcomes through personalized, real-time decision-making in evolving clinical contexts, but require careful supervision for unsafe treatment risks. Existing efforts rely primarily on clinician-prescribed gold standards despite the absence of a known optimal strategy, and predominantly using structured EHR data without extracting valuable insights from clinical notes, limiting their reliability for treatment recommendations. In this work, we introduce SAFER, a calibrated risk-aware tabular-language recommendation framework for DTR that integrates both structured EHR and clinical notes, enabling them to learn from each other, and addresses inherent label uncertainty by assuming ambiguous optimal treatment solution for deceased patients. Moreover, SAFER employs conformal prediction to provide statistical guarantees, ensuring safe treatment recommendations while filtering out uncertain predictions. Experiments on two publicly available sepsis datasets demonstrate that SAFER outperforms state-of-the-art baselines across multiple recommendation metrics and counterfactual mortality rate, while offering robust formal assurances. These findings underscore SAFER potential as a trustworthy and theoretically grounded solution for high-stakes DTR applications.
Authors: Ittai Rubinstein, Samuel B. Hopkins
Abstract: How does the training data affect a model's behavior? This is the question we seek to answer with data attribution. The leading practical approaches to data attribution are based on influence functions (IF). IFs utilize a first-order Taylor approximation to efficiently predict the effect of removing a set of samples from the training set without retraining the model, and are used in a wide variety of machine learning applications. However, especially in the high-dimensional regime (# params $\geq \Omega($# samples$)$), they are often imprecise and tend to underestimate the effect of sample removals, even for simple models such as logistic regression. We present rescaled influence functions (RIF), a new tool for data attribution which can be used as a drop-in replacement for influence functions, with little computational overhead but significant improvement in accuracy. We compare IF and RIF on a range of real-world datasets, showing that RIFs offer significantly better predictions in practice, and present a theoretical analysis explaining this improvement. Finally, we present a simple class of data poisoning attacks that would fool IF-based detections but would be detected by RIF.
Authors: Hong-Ming Chiu, Hao Chen, Huan Zhang, Richard Y. Zhang
Abstract: Neural network verifiers based on linear bound propagation scale impressively to massive models but can be surprisingly loose when neuron coupling is crucial. Conversely, semidefinite programming (SDP) verifiers capture inter-neuron coupling naturally, but their cubic complexity restricts them to only small models. In this paper, we propose SDP-CROWN, a novel hybrid verification framework that combines the tightness of SDP relaxations with the scalability of bound-propagation verifiers. At the core of SDP-CROWN is a new linear bound, derived via SDP principles, that explicitly captures $\ell_{2}$-norm-based inter-neuron coupling while adding only one extra parameter per layer. This bound can be integrated seamlessly into any linear bound-propagation pipeline, preserving the inherent scalability of such methods yet significantly improving tightness. In theory, we prove that our inter-neuron bound can be up to a factor of $\sqrt{n}$ tighter than traditional per-neuron bounds. In practice, when incorporated into the state-of-the-art $\alpha$-CROWN verifier, we observe markedly improved verification performance on large models with up to 65 thousand neurons and 2.47 million parameters, achieving tightness that approaches that of costly SDP-based methods.
Authors: Oktay Karaku\c{s}, Hasan Arkada\c{s}
Abstract: Line-breaking passes (LBPs) are crucial tactical actions in football, allowing teams to penetrate defensive lines and access high-value spaces. In this study, we present an unsupervised, clustering-based framework for detecting and analysing LBPs using synchronised event and tracking data from elite matches. Our approach models opponent team shape through vertical spatial segmentation and identifies passes that disrupt defensive lines within open play. Beyond detection, we introduce several tactical metrics, including the space build-up ratio (SBR) and two chain-based variants, LBPCh$^1$ and LBPCh$^2$, which quantify the effectiveness of LBPs in generating immediate or sustained attacking threats. We evaluate these metrics across teams and players in the 2022 FIFA World Cup, revealing stylistic differences in vertical progression and structural disruption. The proposed methodology is explainable, scalable, and directly applicable to modern performance analysis and scouting workflows.
Authors: Di Lin, Wanjing Ren, Xuanbin Li, Rui Zhang
Abstract: In graph self-supervised learning, masked autoencoders (MAE) and contrastive learning (CL) are two prominent paradigms. MAE focuses on reconstructing masked elements, while CL maximizes similarity between augmented graph views. Recent studies highlight their complementarity: MAE excels at local feature capture, and CL at global information extraction. Hybrid frameworks for homogeneous graphs have been proposed, but face challenges in designing shared encoders to meet the semantic requirements of both tasks. In semantically sparse scenarios, CL struggles with view construction, and gradient imbalance between positive and negative samples persists. This paper introduces HetCRF, a novel dual-channel self-supervised learning framework for heterogeneous graphs. HetCRF uses a two-stage aggregation strategy to adapt embedding semantics, making it compatible with both MAE and CL. To address semantic sparsity, it enhances encoder output for view construction instead of relying on raw features, improving efficiency. Two positive sample augmentation strategies are also proposed to balance gradient contributions. Node classification experiments on four real-world heterogeneous graph datasets demonstrate that HetCRF outperforms state-of-the-art baselines. On datasets with missing node features, such as Aminer and Freebase, at a 40% label rate in node classification, HetCRF improves the Macro-F1 score by 2.75% and 2.2% respectively compared to the second-best baseline, validating its effectiveness and superiority.
Authors: Yuan Yuan, Yukun Liu, Chonghua Han, Jie Feng, Yong Li
Abstract: Foundation models have revolutionized fields such as natural language processing and computer vision by enabling general-purpose learning across diverse tasks and datasets. However, building analogous models for human mobility remains challenging due to the privacy-sensitive nature of mobility data and the resulting data silos across institutions. To bridge this gap, we propose MoveGCL, a scalable and privacy-preserving framework for training mobility foundation models via generative continual learning. Without sharing raw data, MoveGCL enables decentralized and progressive model evolution by replaying synthetic trajectories generated from a frozen teacher model, and reinforces knowledge retention through a tailored distillation strategy that mitigates catastrophic forgetting. To address the heterogeneity of mobility patterns, MoveGCL incorporates a Mixture-of-Experts Transformer with a mobility-aware expert routing mechanism, and employs a layer-wise progressive adaptation strategy to stabilize continual updates. Experiments on six real-world urban datasets demonstrate that MoveGCL achieves performance comparable to joint training and significantly outperforms federated learning baselines, while offering strong privacy protection. MoveGCL marks a crucial step toward unlocking foundation models for mobility, offering a practical blueprint for open, scalable, and privacy-preserving model development in the era of foundation models.
Authors: Rajeev Bhatt Ambati, James Lester, Shashank Srivastava, Snigdha Chaturvedi
Abstract: Large Language Models (LLMs) excel at few-shot learning via in-context learning (ICL). However, the effectiveness of ICL is often sensitive to the selection and ordering of demonstration examples. To address this, we present MarginSel: Max-Margin Demonstration Selection for LLMs, a two-step method that selects hard demonstration examples for the ICL prompt, adapting to each test instance. Our approach achieves 2-7% absolute improvement in F1-score across classification tasks, compared to a random selection of examples. We also provide theoretical insights and empirical evidence showing that MarginSel induces max-margin behavior in LLMs by effectively increasing the margin for hard examples, analogous to support vectors, thereby shifting the decision boundary in a beneficial direction.
Authors: Fudong Lin, Wanrou Du, Jinchan Liu, Tarikul Milon, Shelby Meche, Wu Xu, Xiaoqi Qin, Xu Yuan
Abstract: Deep neural networks, particularly Transformers, have been widely adopted for predicting the functional properties of proteins. In this work, we focus on exploring whether Protein Transformers can capture biological intelligence among protein sequences. To achieve our goal, we first introduce a protein function dataset, namely Protein-FN, providing over 9000 protein data with meaningful labels. Second, we devise a new Transformer architecture, namely Sequence Protein Transformers (SPT), for computationally efficient protein function predictions. Third, we develop a novel Explainable Artificial Intelligence (XAI) technique called Sequence Score, which can efficiently interpret the decision-making processes of protein models, thereby overcoming the difficulty of deciphering biological intelligence bided in Protein Transformers. Remarkably, even our smallest SPT-Tiny model, which contains only 5.4M parameters, demonstrates impressive predictive accuracy, achieving 94.3% on the Antibiotic Resistance (AR) dataset and 99.6% on the Protein-FN dataset, all accomplished by training from scratch. Besides, our Sequence Score technique helps reveal that our SPT models can discover several meaningful patterns underlying the sequence structures of protein data, with these patterns aligning closely with the domain knowledge in the biology community. We have officially released our Protein-FN dataset on Hugging Face Datasets https://huggingface.co/datasets/Protein-FN/Protein-FN. Our code is available at https://github.com/fudong03/BioIntelligence.
URLs: https://huggingface.co/datasets/Protein-FN/Protein-FN., https://github.com/fudong03/BioIntelligence.
Authors: Minh-Duc Nguyen, Dung D. Le
Abstract: Pareto Set Learning (PSL) is popular as an efficient approach to obtaining the complete optimal solution in Multi-objective Learning (MOL). A set of optimal solutions approximates the Pareto set, and its mapping is a set of dense points in the Pareto front in objective space. However, some current methods face a challenge: how to make the Pareto solution is diverse while maximizing the hypervolume value. In this paper, we propose a novel method to address this challenge, which employs Stein Variational Gradient Descent (SVGD) to approximate the entire Pareto set. SVGD pushes a set of particles towards the Pareto set by applying a form of functional gradient descent, which helps to converge and diversify optimal solutions. Additionally, we employ diverse gradient direction strategies to thoroughly investigate a unified framework for SVGD in multi-objective optimization and adapt this framework with an annealing schedule to promote stability. We introduce our method, SVH-MOL, and validate its effectiveness through extensive experiments on multi-objective problems and multi-task learning, demonstrating its superior performance.
Authors: Adri\`a Molina Rodr\'iguez, Oriol Ramos Terrades, Josep Llad\'os
Abstract: Achieving robustness in recognition systems across diverse domains is crucial for their practical utility. While ample data availability is usually assumed, low-resource languages, such as ancient manuscripts and non-western languages, tend to be kept out of the equations of massive pretraining and foundational techniques due to an under representation. In this work, we aim for building models which can generalize to new distributions of data, such as alphabets, faster than centralized fine-tune strategies. For doing so, we take advantage of the recent advancements in model editing to enhance the incorporation of unseen scripts (low-resource learning). In contrast to state-of-the-art meta-learning, we showcase the effectiveness of domain merging in sparse distributions of data, with agnosticity of its relation to the overall distribution or any other prototyping necessity. Even when using the same exact training data, our experiments showcase significant performance boosts in \textbf{transfer learning} to new alphabets and \textbf{out-of-domain evaluation} in challenging domain shifts, including historical ciphered texts and non-Latin scripts. This research contributes a novel approach into building models that can easily adopt under-represented alphabets and, therefore, enable document recognition to a wider set of contexts and cultures.
Authors: Qinting Jiang, Chuyang Ye, Dongyan Wei, Bingli Wang, Yuan Xue, Jingyan Jiang, Zhi Wang
Abstract: Despite progress, deep neural networks still suffer performance declines under distribution shifts between training and test domains, leading to a substantial decrease in Quality of Experience (QoE) for applications. Existing test-time adaptation (TTA) methods are challenged by dynamic, multiple test distributions within batches. We observe that feature distributions across different domains inherently cluster into distinct groups with varying means and variances. This divergence reveals a critical limitation of previous global normalization strategies in TTA, which inevitably distort the original data characteristics. Based on this insight, we propose Feature-based Instance Neighbor Discovery (FIND), which comprises three key components: Layer-wise Feature Disentanglement (LFD), Feature Aware Batch Normalization (FABN) and Selective FABN (S-FABN). LFD stably captures features with similar distributions at each layer by constructing graph structures. While FABN optimally combines source statistics with test-time distribution specific statistics for robust feature representation. Finally, S-FABN determines which layers require feature partitioning and which can remain unified, thereby enhancing inference efficiency. Extensive experiments demonstrate that FIND significantly outperforms existing methods, achieving a 30\% accuracy improvement in dynamic scenarios while maintaining computational efficiency.
Authors: Marek \v{C}ern\'y
Abstract: Message-passing graph neural networks (MPGNNs) dominate modern graph learning, typically prioritizing maximal expressive power. In contrast, we introduce an \emph{efficient aggregation} mechanism, deliberately trading off some expressivity for stronger and more structured aggregation capabilities. Our approach allows seamless scaling between classical message-passing and simpler methods based on colored or plain walks. We rigorously characterize the expressive power at each intermediate step using homomorphism counts from a hierarchy of generalized \emph{caterpillar graphs}. Based on this foundation, we propose the \emph{Caterpillar GNN}, whose robust graph-level aggregation enables it to successfully tackle synthetic graph-level task specifically designed to be challenging for classical MPGNNs. Moreover, we demonstrate that, on real-world datasets, the Caterpillar GNN achieves comparable predictive performance while significantly reducing the number of nodes in the hidden layers of the computational graph.
Authors: Qiyun Zhao
Abstract: As integrated circuit scale grows and design complexity rises, effective circuit representation helps support logic synthesis, formal verification, and other automated processes in electronic design automation. And-Inverter Graphs (AIGs), as a compact and canonical structure, are widely adopted for representing Boolean logic in these workflows. However, the increasing complexity and integration density of modern circuits introduce structural heterogeneity and global logic information loss in AIGs, posing significant challenges to accurate circuit modeling. To address these issues, we propose FuncGNN, which integrates hybrid feature aggregation to extract multi-granularity topological patterns, thereby mitigating structural heterogeneity and enhancing logic circuit representations. FuncGNN further introduces gate-aware normalization that adapts to circuit-specific gate distributions, improving robustness to structural heterogeneity. Finally, FuncGNN employs multi-layer integration to merge intermediate features across layers, effectively synthesizing local and global semantic information for comprehensive logic representations. Experimental results on two logic-level analysis tasks (i.e., signal probability prediction and truth-table distance prediction) demonstrate that FuncGNN outperforms existing state-of-the-art methods, achieving improvements of 2.06% and 18.71%, respectively, while reducing training time by approximately 50.6% and GPU memory usage by about 32.8%.
Authors: Zixuan Dong, Yumi Omori, Keith Ross
Abstract: Inverse Reinforcement Learning (IRL) aims to recover a reward function from expert demonstrations. Recently, Optimal Transport (OT) methods have been successfully deployed to align trajectories and infer rewards. While OT-based methods have shown strong empirical results, they introduce algorithmic complexity, hyperparameter sensitivity, and require solving the OT optimization problems. In this work, we challenge the necessity of OT in IRL by proposing two simple, heuristic alternatives: (1) Minimum-Distance Reward, which assigns rewards based on the nearest expert state regardless of temporal order; and (2) Segment-Matching Reward, which incorporates lightweight temporal alignment by matching agent states to corresponding segments in the expert trajectory. These methods avoid optimization, exhibit linear-time complexity, and are easy to implement. Through extensive evaluations across 32 online and offline benchmarks with three reinforcement learning algorithms, we show that our simple rewards match or outperform recent OT-based approaches. Our findings suggest that the core benefits of OT may arise from basic proximity alignment rather than its optimal coupling formulation, advocating for reevaluation of complexity in future IRL design.
Authors: Di Lin, Wanjing Ren, Xuanbin Li, Rui Zhang
Abstract: Self-supervised learning (SSL) methods have been increasingly applied to diverse downstream tasks due to their superior generalization capabilities and low annotation costs. However, most existing heterogeneous graph SSL models convert heterogeneous graphs into homogeneous ones via meta-paths for training, which only leverage information from nodes at both ends of meta-paths while underutilizing the heterogeneous node information along the meta-paths. To address this limitation, this paper proposes a novel framework named IMPA-HGAE to enhance target node embeddings by fully exploiting internal node information along meta-paths. Experimental results validate that IMPA-HGAE achieves superior performance on heterogeneous datasets. Furthermore, this paper introduce innovative masking strategies to strengthen the representational capacity of generative SSL models on heterogeneous graph data. Additionally, this paper discuss the interpretability of the proposed method and potential future directions for generative self-supervised learning in heterogeneous graphs. This work provides insights into leveraging meta-path-guided structural semantics for robust representation learning in complex graph scenarios.
Authors: Max McGuinness, Eirik Fladmark, Francisco Vargas
Abstract: We present an early investigation into the use of neural diffusion processes for global optimisation, focusing on Zhang et al.'s Path Integral Sampler. One can use the Boltzmann distribution to formulate optimization as solving a Schr\"odinger bridge sampling problem, then apply Girsanov's theorem with a simple (single-point) prior to frame it in stochastic control terms, and compute the solution's integral terms via a neural approximation (a Fourier MLP). We provide theoretical bounds for this optimiser, results on toy optimisation tasks, and a summary of the stochastic theory motivating the model. Ultimately, we found the optimiser to display promising per-step performance at optimisation tasks between 2 and 1,247 dimensions, but struggle to explore higher-dimensional spaces when faced with a 15.9k parameter model, indicating a need for work on adaptation in such environments.
Authors: Ilya Kaufman Sirot, Omri Azencot
Abstract: Deep learning models with a large number of parameters, often referred to as over-parameterized models, have achieved exceptional performance across various tasks. Despite concerns about overfitting, these models frequently generalize well to unseen data, thanks to effective regularization techniques, with data augmentation being among the most widely used. While data augmentation has shown great success in classification tasks using label-preserving transformations, its application in regression problems has received less attention. Recently, a novel \emph{manifold learning} approach for generating synthetic data was proposed, utilizing a first-order approximation of the data manifold. Building on this foundation, we present a theoretical framework and practical tools for approximating and sampling general data manifolds. Furthermore, we introduce the Curvature-Enhanced Manifold Sampling (CEMS) method for regression tasks. CEMS leverages a second-order representation of the data manifold to enable efficient sampling and reconstruction of new data points. Extensive evaluations across multiple datasets and comparisons with state-of-the-art methods demonstrate that CEMS delivers superior performance in both in-distribution and out-of-distribution scenarios, while introducing only minimal computational overhead. Code is available at https://github.com/azencot-group/CEMS.
Authors: Ziwei Li, Yuhan Duan, Tianyu Xiong, Yi-Tang Chen, Wei-Lun Chao, Han-Wei Shen
Abstract: Effective surrogate models are critical for accelerating scientific simulations. Implicit neural representations (INRs) offer a compact and continuous framework for modeling spatially structured data, but they often struggle with complex scientific fields exhibiting localized, high-frequency variations. Recent approaches address this by introducing additional features along rigid geometric structures (e.g., grids), but at the cost of flexibility and increased model size. In this paper, we propose a simple yet effective alternative: Feature-Adaptive INR (FA-INR). FA-INR leverages cross-attention to an augmented memory bank to learn flexible feature representations, enabling adaptive allocation of model capacity based on data characteristics, rather than rigid structural assumptions. To further improve scalability, we introduce a coordinate-guided mixture of experts (MoE) that enhances the specialization and efficiency of feature representations. Experiments on three large-scale ensemble simulation datasets show that FA-INR achieves state-of-the-art fidelity while significantly reducing model size, establishing a new trade-off frontier between accuracy and compactness for INR-based surrogates.
Authors: Xizhi Tian, Meng Ding, Touming Tao, Zihang Xiang, Di Wang
Abstract: As a fundamental problem in machine learning and differential privacy (DP), DP linear regression has been extensively studied. However, most existing methods focus primarily on either regular data distributions or low-dimensional cases with irregular data. To address these limitations, this paper provides a comprehensive study of DP sparse linear regression with heavy-tailed responses in high-dimensional settings. In the first part, we introduce the DP-IHT-H method, which leverages the Huber loss and private iterative hard thresholding to achieve an estimation error bound of \( \tilde{O}\biggl( s^{* \frac{1 }{2}} \cdot \biggl(\frac{\log d}{n}\biggr)^{\frac{\zeta}{1 + \zeta}} + s^{* \frac{1 + 2\zeta}{2 + 2\zeta}} \cdot \biggl(\frac{\log^2 d}{n \varepsilon}\biggr)^{\frac{\zeta}{1 + \zeta}} \biggr) \) under the $(\varepsilon, \delta)$-DP model, where $n$ is the sample size, $d$ is the dimensionality, $s^*$ is the sparsity of the parameter, and $\zeta \in (0, 1]$ characterizes the tail heaviness of the data. In the second part, we propose DP-IHT-L, which further improves the error bound under additional assumptions on the response and achieves \( \tilde{O}\Bigl(\frac{(s^*)^{3/2} \log d}{n \varepsilon}\Bigr). \) Compared to the first result, this bound is independent of the tail parameter $\zeta$. Finally, through experiments on synthetic and real-world datasets, we demonstrate that our methods outperform standard DP algorithms designed for ``regular'' data.
Authors: Dongyeop Lee, Kwanhee Lee, Jinseok Chung, Namhoon Lee
Abstract: Sparsifying neural networks often suffers from seemingly inevitable performance degradation, and it remains challenging to restore the original performance despite much recent progress. Motivated by recent studies in robust optimization, we aim to tackle this problem by finding subnetworks that are both sparse and flat at the same time. Specifically, we formulate pruning as a sparsity-constrained optimization problem where flatness is encouraged as an objective. We solve it explicitly via an augmented Lagrange dual approach and extend it further by proposing a generalized projection operation, resulting in novel pruning methods called SAFE and its extension, SAFE$^+$. Extensive evaluations on standard image classification and language modeling tasks reveal that SAFE consistently yields sparse networks with improved generalization performance, which compares competitively to well-established baselines. In addition, SAFE demonstrates resilience to noisy data, making it well-suited for real-world conditions.
Authors: Armin Behnamnia, Gholamali Aminian, Alireza Aghaei, Chengchun Shi, Vincent Y. F. Tan, Hamid R. Rabiee
Abstract: Off-policy learning and evaluation leverage logged bandit feedback datasets, which contain context, action, propensity score, and feedback for each data point. These scenarios face significant challenges due to high variance and poor performance with low-quality propensity scores and heavy-tailed reward distributions. We address these issues by introducing a novel estimator based on the log-sum-exponential (LSE) operator, which outperforms traditional inverse propensity score estimators. Our LSE estimator demonstrates variance reduction and robustness under heavy-tailed conditions. For off-policy evaluation, we derive upper bounds on the estimator's bias and variance. In the off-policy learning scenario, we establish bounds on the regret -- the performance gap between our LSE estimator and the optimal policy -- assuming bounded $(1+\epsilon)$-th moment of weighted reward. Notably, we achieve a convergence rate of $O(n^{-\epsilon/(1+ \epsilon)})$ for the regret bounds, where $\epsilon \in [0,1]$ and $n$ is the size of logged bandit feedback dataset. Theoretical analysis is complemented by comprehensive empirical evaluations in both off-policy learning and evaluation scenarios, confirming the practical advantages of our approach. The code for our estimator is available at the following link: https://github.com/armin-behnamnia/lse-offpolicy-learning.
URLs: https://github.com/armin-behnamnia/lse-offpolicy-learning.
Authors: Divya Jyoti Bajpai, Manjesh Kumar Hanawal
Abstract: In recent years, Vision-Language Models (VLMs) have shown remarkable performance improvements in Vision-Language tasks. However, their large size poses challenges for real-world applications where inference latency is a concern. To tackle this issue, we propose employing Early Exit (EE) strategies in VLMs. However, training exit classifiers in VLMs is challenging, particularly with limited labeled training data. To address this, we introduce FREE, an adversarial training approach within a GAN-based framework. Here, each exit consists of a transformer layer and a classifier. The transformer layer is adversarially trained to produce feature representations similar to the final layer, while a feature classifier serves as the discriminator. Our method focuses on performing input-adaptive inference that increases inference speed with minimal drop in performance. Experimental results demonstrate the effectiveness of our approach in enhancing accuracy and model robustness by mitigating overthinking and the phenomenon of mid-crisis that we highlight. We experimentally validate that our method speeds up the inference process by more than 1.51x while retaining comparable performance. The source code is available at https://github.com/Div290/FREE.
Authors: Paulius Sasnauskas, Yi\u{g}it Yal{\i}n, Goran Radanovi\'c
Abstract: We study the corruption-robustness of in-context reinforcement learning (ICRL), focusing on the Decision-Pretrained Transformer (DPT, Lee et al., 2023). To address the challenge of reward poisoning attacks targeting the DPT, we propose a novel adversarial training framework, called Adversarially Trained Decision-Pretrained Transformer (AT-DPT). Our method simultaneously trains an attacker to minimize the true reward of the DPT by poisoning environment rewards, and a DPT model to infer optimal actions from the poisoned data. We evaluate the effectiveness of our approach against standard bandit algorithms, including robust baselines designed to handle reward contamination. Our results show that the proposed method significantly outperforms these baselines in bandit settings, under a learned attacker. We additionally evaluate AT-DPT on an adaptive attacker, and observe similar results. Furthermore, we extend our evaluation to the MDP setting, confirming that the robustness observed in bandit scenarios generalizes to more complex environments.
Authors: Jihao Andreas Lin, Sebastian Ament, Maximilian Balandat, David Eriksson, Jos\'e Miguel Hern\'andez-Lobato, Eytan Bakshy
Abstract: Applying Gaussian processes (GPs) to very large datasets remains a challenge due to limited computational scalability. Matrix structures, such as the Kronecker product, can accelerate operations significantly, but their application commonly entails approximations or unrealistic assumptions. In particular, the most common path to creating a Kronecker-structured kernel matrix is by evaluating a product kernel on gridded inputs that can be expressed as a Cartesian product. However, this structure is lost if any observation is missing, breaking the Cartesian product structure, which frequently occurs in real-world data such as time series. To address this limitation, we propose leveraging latent Kronecker structure, by expressing the kernel matrix of observed values as the projection of a latent Kronecker product. In combination with iterative linear system solvers and pathwise conditioning, our method facilitates inference of exact GPs while requiring substantially fewer computational resources than standard iterative methods. We demonstrate that our method outperforms state-of-the-art sparse and variational GPs on real-world datasets with up to five million examples, including robotics, automated machine learning, and climate applications.
Authors: Fred Xu, Thomas Markovich
Abstract: Graph Neural Networks have achieved impressive results across diverse network modeling tasks, but accurately estimating uncertainty on graphs remains difficult, especially under distributional shifts. Unlike traditional uncertainty estimation, graph-based uncertainty must account for randomness arising from both the graph's structure and its label distribution, which adds complexity. In this paper, making an analogy between the evolution of a stochastic partial differential equation (SPDE) driven by Matern Gaussian Process and message passing using GNN layers, we present a principled way to design a novel message passing scheme that incorporates spatial-temporal noises motivated by the Gaussian Process approach to SPDE. Our method simultaneously captures uncertainty across space and time and allows explicit control over the covariance kernel smoothness, thereby enhancing uncertainty estimates on graphs with both low and high label informativeness. Our extensive experiments on Out-of-Distribution (OOD) detection on graph datasets with varying label informativeness demonstrate the soundness and superiority of our model to existing approaches.
Authors: Shangjie Du, Hui Wei, Dong Yoon Lee, Zhizhang Hu, Shijia Pan
Abstract: This work introduces GraPhy, a graph-based, physics-guided learning framework for high-resolution and accurate air quality modeling in urban areas with limited monitoring data. Fine-grained air quality monitoring information is essential for reducing public exposure to pollutants. However, monitoring networks are often sparse in socioeconomically disadvantaged regions, limiting the accuracy and resolution of air quality modeling. To address this, we propose a physics-guided graph neural network architecture called GraPhy with layers and edge features designed specifically for low-resolution monitoring data. Experiments using data from California's socioeconomically disadvantaged San Joaquin Valley show that GraPhy achieves the overall best performance evaluated by mean squared error (MSE), mean absolute error (MAE), and R-square value (R2), improving the performance by 9%-56% compared to various baseline models. Moreover, GraPhy consistently outperforms baselines across different spatial heterogeneity levels, demonstrating the effectiveness of our model design.
Authors: Wei Min Loh, Jiaqi Shang, Pascal Poupart
Abstract: Dealing with tabular data is challenging due to partial information, noise, and heterogeneous structure. Existing techniques often struggle to simultaneously address key aspects of tabular data such as textual information, a variable number of columns, and unseen data without metadata besides column names. We propose a novel architecture, \textit{basis transformers}, specifically designed to tackle these challenges while respecting inherent invariances in tabular data, including hierarchical structure and the representation of numeric values. We evaluate our design on a multi-task tabular regression benchmark, achieving an improvement of 0.338 in the median $R^2$ score and the lowest standard deviation across 34 tasks from the OpenML-CTR23 benchmark. Furthermore, our model has five times fewer parameters than the best-performing baseline and surpasses pretrained large language model baselines -- even when initialized from randomized weights.
Authors: Mahdi Salmani, Alireza Abdollahpoorrostam, Seyed-Mohsen Moosavi-Dezfooli
Abstract: Traditional decision-based black-box adversarial attacks on image classifiers aim to generate adversarial examples by slightly modifying input images while keeping the number of queries low, where each query involves sending an input to the model and observing its output. Most existing methods assume that all queries have equal cost. However, in practice, queries may incur asymmetric costs; for example, in content moderation systems, certain output classes may trigger additional review, enforcement, or penalties, making them more costly than others. While prior work has considered such asymmetric cost settings, effective algorithms for this scenario remain underdeveloped. In this paper, we propose a general framework for decision-based attacks under asymmetric query costs, which we refer to as asymmetric black-box attacks. We modify two core components of existing attacks: the search strategy and the gradient estimation process. Specifically, we propose Asymmetric Search (AS), a more conservative variant of binary search that reduces reliance on high-cost queries, and Asymmetric Gradient Estimation (AGREST), which shifts the sampling distribution to favor low-cost queries. We design efficient algorithms that minimize total attack cost by balancing different query types, in contrast to earlier methods such as stealthy attacks that focus only on limiting expensive (high-cost) queries. Our method can be integrated into a range of existing black-box attacks with minimal changes. We perform both theoretical analysis and empirical evaluation on standard image classification benchmarks. Across various cost regimes, our method consistently achieves lower total query cost and smaller perturbations than existing approaches, with improvements of up to 40% in some settings.
Authors: Geonhui Yoo, Minhak Song, Chulhee Yun
Abstract: When training deep neural networks with gradient descent, sharpness often increases -- a phenomenon known as progressive sharpening -- before saturating at the edge of stability. Although commonly observed in practice, the underlying mechanisms behind progressive sharpening remain poorly understood. In this work, we study this phenomenon using a minimalist model: a deep linear network with a single neuron per layer. We show that this simple model effectively captures the sharpness dynamics observed in recent empirical studies, offering a simple testbed to better understand neural network training. Moreover, we theoretically analyze how dataset properties, network depth, stochasticity of optimizers, and step size affect the degree of progressive sharpening in the minimalist model. We then empirically demonstrate how these theoretical insights extend to practical scenarios. This study offers a deeper understanding of sharpness dynamics in neural network training, highlighting the interplay between depth, training data, and optimizers.
Authors: Clinton Enwerem, Aniruddh G. Puranic, John S. Baras, Calin Belta
Abstract: Mainstream approximate action-value iteration reinforcement learning (RL) algorithms suffer from overestimation bias, leading to suboptimal policies in high-variance stochastic environments. Quantile-based action-value iteration methods reduce this bias by learning a distribution of the expected cost-to-go using quantile regression. However, ensuring that the learned policy satisfies safety constraints remains a challenge when these constraints are not explicitly integrated into the RL framework. Existing methods often require complex neural architectures or manual tradeoffs due to combined cost functions. To address this, we propose a risk-regularized quantile-based algorithm integrating Conditional Value-at-Risk (CVaR) to enforce safety without complex architectures. We also provide theoretical guarantees on the contraction properties of the risk-sensitive distributional Bellman operator in Wasserstein space, ensuring convergence to a unique cost distribution. Simulations of a mobile robot in a dynamic reach-avoid task show that our approach leads to more goal successes, fewer collisions, and better safety-performance trade-offs compared to risk-neutral methods.
Authors: Pengfei Hu, Xiaoxue Han, Fei Wang, Yue Ning
Abstract: Domain generalization has become a critical challenge in clinical prediction, where patient cohorts often exhibit shifting data distributions that degrade model performance. Typical domain generalization approaches struggle in real-world healthcare settings for two main reasons: (1) patient-specific domain labels are typically unavailable, making domain discovery especially difficult; (2) purely data-driven approaches overlook key clinical insights, leading to a gap in medical knowledge integration. To address these problems, we leverage hierarchical medical ontologies like the ICD-9-CM hierarchy to group diseases into higher-level categories and discover more flexible latent domains. In this paper, we introduce UdonCare, a hierarchy-guided framework that iteratively prunes fine-grained domains, encodes these refined domains, and applies a Siamese-type inference mechanism to separate domain-related signals from patient-level features. Experimental results on clinical datasets (MIMIC-III and MIMIC-IV) show that the proposed model achieves higher performance compared to other domain generalization baselines when substantial domain gaps presents, highlighting the untapped potential of medical knowledge for enhancing domain generalization in practical healthcare applications.
Authors: Zitian Li, Wang Chi Cheung
Abstract: Motivated by an open direction in existing literature, we study the 1-identification problem, a fundamental multi-armed bandit formulation on pure exploration. The goal is to determine whether there exists an arm whose mean reward is at least a known threshold $\mu_0$, or to output None if it believes such an arm does not exist. The agent needs to guarantee its output is correct with probability at least $1-\delta$. Degenne & Koolen 2019 has established the asymptotically tight sample complexity for the 1-identification problem, but they commented that the non-asymptotic analysis remains unclear. We design a new algorithm Sequential-Exploration-Exploitation (SEE), and conduct theoretical analysis from the non-asymptotic perspective. Novel to the literature, we achieve near optimality, in the sense of matching upper and lower bounds on the pulling complexity. The gap between the upper and lower bounds is up to a polynomial logarithmic factor. The numerical result also indicates the effectiveness of our algorithm, compared to existing benchmarks.
Authors: Sajib Acharjee Dip, Uddip Acharjee Shuvo, Dipanwita Mallick, Abrar Rahman Abir, Liqing Zhang
Abstract: Cancer subtype classification is crucial for personalized treatment and prognostic assessment. However, effectively integrating multi-omic data remains challenging due to the heterogeneous nature of genomic, epigenomic, and transcriptomic features. In this work, we propose Modality-Aware Cross-Attention MoXGATE, a novel deep-learning framework that leverages cross-attention and learnable modality weights to enhance feature fusion across multiple omics sources. Our approach effectively captures inter-modality dependencies, ensuring robust and interpretable integration. Through experiments on Gastrointestinal Adenocarcinoma (GIAC) and Breast Cancer (BRCA) datasets from TCGA, we demonstrate that MoXGATE outperforms existing methods, achieving 95\% classification accuracy. Ablation studies validate the effectiveness of cross-attention over simple concatenation and highlight the importance of different omics modalities. Moreover, our model generalizes well to unseen cancer types e.g., breast cancer, underscoring its adaptability. Key contributions include (1) a cross-attention-based multi-omic integration framework, (2) modality-weighted fusion for enhanced interpretability, (3) application of focal loss to mitigate data imbalance, and (4) validation across multiple cancer subtypes. Our results indicate that MoXGATE is a promising approach for multi-omic cancer subtype classification, offering improved performance and biological generalizability.
Authors: Anastasia Koloskova, Youssef Allouah, Animesh Jha, Rachid Guerraoui, Sanmi Koyejo
Abstract: We address the problem of machine unlearning, where the goal is to remove the influence of specific training data from a model upon request, motivated by privacy concerns and regulatory requirements such as the "right to be forgotten." Unfortunately, existing methods rely on restrictive assumptions or lack formal guarantees. To this end, we propose a novel method for certified machine unlearning, leveraging the connection between unlearning and privacy amplification by stochastic post-processing. Our method uses noisy fine-tuning on the retain data, i.e., data that does not need to be removed, to ensure provable unlearning guarantees. This approach requires no assumptions about the underlying loss function, making it broadly applicable across diverse settings. We analyze the theoretical trade-offs in efficiency and accuracy and demonstrate empirically that our method not only achieves formal unlearning guarantees but also performs effectively in practice, outperforming existing baselines. Our code is available at https://github.com/stair-lab/certified-unlearningneural-networks-icml-2025
URLs: https://github.com/stair-lab/certified-unlearningneural-networks-icml-2025
Authors: Austin Snyder, Ryan Gallagher, Boris Kovalerchuk
Abstract: Building on existing work with Hyperblocks, which classify data using minimum and maximum bounds for each attribute, we focus on enhancing interpretability, decreasing training time, and reducing model complexity without sacrificing accuracy. This system allows subject matter experts (SMEs) to directly inspect and understand the model's decision logic without requiring extensive machine learning expertise. To reduce Hyperblock complexity while retaining performance, we introduce a suite of algorithms for Hyperblock simplification. These include removing redundant attributes, removing redundant blocks through overlap analysis, and creating disjunctive units. These methods eliminate unnecessary parameters, dramatically reducing model size without harming classification power. We increase robustness by introducing an interpretable fallback mechanism using k-Nearest Neighbor (k-NN) classifiers for points not covered by any block, ensuring complete data coverage while preserving model transparency. Our results demonstrate that interpretable models can scale to high-dimensional, large-volume datasets while maintaining competitive accuracy. On benchmark datasets such as WBC (9-D), we achieve strong predictive performance with significantly reduced complexity. On MNIST (784-D), our method continues to improve through tuning and simplification, showing promise as a transparent alternative to black-box models in domains where trust, clarity, and control are crucial.
Authors: Mingyi Li, Michael R. Metel, Akiko Takeda
Abstract: The K-means algorithm is one of the most widely studied clustering algorithms in machine learning. While extensive research has focused on its ability to achieve a globally optimal solution, there still lacks a rigorous analysis of its local optimality guarantees. In this paper, we first present conditions under which the K-means algorithm converges to a locally optimal solution. Based on this, we propose simple modifications to the K-means algorithm which ensure local optimality in both the continuous and discrete sense, with the same computational complexity as the original K-means algorithm. As the dissimilarity measure, we consider a general Bregman divergence, which is an extension of the squared Euclidean distance often used in the K-means algorithm. Numerical experiments confirm that the K-means algorithm does not always find a locally optimal solution in practice, while our proposed methods provide improved locally optimal solutions with reduced clustering loss. Our code is available at https://github.com/lmingyi/LO-K-means.
Authors: Arun Sharma, Mingzhou Yang, Majid Farhadloo, Subhankar Ghosh, Bharat Jayaprakash, Shashi Shekhar
Abstract: Given trajectory data, a domain-specific study area, and a user-defined threshold, we aim to find anomalous trajectories indicative of possible GPS spoofing (e.g., fake trajectory). The problem is societally important to curb illegal activities in international waters, such as unauthorized fishing and illicit oil transfers. The problem is challenging due to advances in AI generated in deep fakes generation (e.g., additive noise, fake trajectories) and lack of adequate amount of labeled samples for ground-truth verification. Recent literature shows promising results for anomalous trajectory detection using generative models despite data sparsity. However, they do not consider fine-scale spatiotemporal dependencies and prior physical knowledge, resulting in higher false-positive rates. To address these limitations, we propose a physics-informed diffusion model that integrates kinematic constraints to identify trajectories that do not adhere to physical laws. Experimental results on real-world datasets in the maritime and urban domains show that the proposed framework results in higher prediction accuracy and lower estimation error rate for anomaly detection and trajectory generation methods, respectively. Our implementation is available at https://github.com/arunshar/Physics-Informed-Diffusion-Probabilistic-Model.
URLs: https://github.com/arunshar/Physics-Informed-Diffusion-Probabilistic-Model.
Authors: Utkarsh Utkarsh, Danielle C. Maddix, Ruijun Ma, Michael W. Mahoney, Yuyang Wang
Abstract: We present a general purpose probabilistic forecasting framework, ProbHardE2E, to learn systems that can incorporate operational/physical constraints as hard requirements. ProbHardE2E enforces hard constraints by exploiting variance information in a novel way; and thus it is also capable of performing uncertainty quantification (UQ) on the model. Our methodology uses a novel differentiable probabilistic projection layer (DPPL) that can be combined with a wide range of neural network architectures. This DPPL allows the model to learn the system in an end-to-end manner, compared to other approaches where the constraints are satisfied either through a post-processing step or at inference. In addition, ProbHardE2E can optimize a strictly proper scoring rule, without making any distributional assumptions on the target, which enables it to obtain robust distributional estimates (in contrast to existing approaches that generally optimize likelihood-based objectives, which are heavily biased by their distributional assumptions and model choices); and it can incorporate a range of non-linear constraints (increasing the power of modeling and flexibility). We apply ProbHardE2E to problems in learning partial differential equations with uncertainty estimates and to probabilistic time-series forecasting, showcasing it as a broadly applicable general setup that connects these seemingly disparate domains.
Authors: Yutaro Nakagama, Daisuke Ishii, Kazuki Yoshizoe
Abstract: Driver drowsiness detection (DDD) prevents road accidents caused by driver fatigue. Vehicle dynamics-based DDD has been proposed as a method that is both economical and high performance. However, there are concerns about the reliability of performance metrics and the reproducibility of many of the existing methods. For instance, some previous studies seem to have a data leakage issue among training and test datasets, and many do not openly provide the datasets they used. To this end, this paper aims to compare the performance of representative vehicle dynamics-based DDD methods under a transparent and fair framework that uses a public dataset. We first develop a framework for extracting features from an open dataset by Aygun et al. and performing DDD with lightweight ML models; the framework is carefully designed to support a variety of onfigurations. Second, we implement three existing representative methods and a concise random forest (RF)-based method in the framework. Finally, we report the results of experiments to verify the reproducibility and clarify the performance of DDD based on common metrics. Among the evaluated methods, the RF-based method achieved the highest accuracy of 88 %. Our findings imply the issues inherent in DDD methods developed in a non-standard manner, and demonstrate a high performance method implemented appropriately.
Authors: Leheng Sheng, Changshuo Shen, Weixiang Zhao, Junfeng Fang, Xiaohao Liu, Zhenkai Liang, Xiang Wang, An Zhang, Tat-Seng Chua
Abstract: As LLMs are increasingly deployed in real-world applications, ensuring their ability to refuse malicious prompts, especially jailbreak attacks, is essential for safe and reliable use. Recently, activation steering has emerged as an effective approach for enhancing LLM safety by adding a refusal direction vector to internal activations of LLMs during inference, which will further induce the refusal behaviors of LLMs. However, indiscriminately applying activation steering fundamentally suffers from the trade-off between safety and utility, since the same steering vector can also lead to over-refusal and degraded performance on benign prompts. Although prior efforts, such as vector calibration and conditional steering, have attempted to mitigate this trade-off, their lack of theoretical grounding limits their robustness and effectiveness. To better address the trade-off between safety and utility, we present a theoretically grounded and empirically effective activation steering method called AlphaSteer. Specifically, it considers activation steering as a learnable process with two principled learning objectives: utility preservation and safety enhancement. For utility preservation, it learns to construct a nearly zero vector for steering benign data, with the null-space constraints. For safety enhancement, it learns to construct a refusal direction vector for steering malicious data, with the help of linear regression. Experiments across multiple jailbreak attacks and utility benchmarks demonstrate the effectiveness of AlphaSteer, which significantly improves the safety of LLMs without compromising general capabilities. Our codes are available at https://github.com/AlphaLab-USTC/AlphaSteer.
Authors: Yung-Chien Wang, Kuang-Da Wang, Wei-Yao Wang, Wen-Chih Peng
Abstract: Tabular data serve as a fundamental and ubiquitous representation of structured information in numerous real-world applications, e.g., finance and urban planning. In the realm of tabular imbalanced applications, data imbalance has been investigated in classification tasks with insufficient instances in certain labels, causing the model's ineffective generalizability. However, the imbalance issue of tabular regression tasks is underexplored, and yet is critical due to unclear boundaries for continuous labels and simplifying assumptions in existing imbalance regression work, which often rely on known and balanced test distributions. Such assumptions may not hold in practice and can lead to performance degradation. To address these issues, we propose MATI: Mixture Experts with Test-Time Self-Supervised Aggregation for Tabular Imbalance Regression, featuring two key innovations: (i) the Region-Aware Mixture Expert, which adopts a Gaussian Mixture Model to capture the underlying related regions. The statistical information of each Gaussian component is then used to synthesize and train region-specific experts to capture the unique characteristics of their respective regions. (ii) Test-Time Self-Supervised Expert Aggregation, which dynamically adjusts region expert weights based on test data features to reinforce expert adaptation across varying test distributions. We evaluated MATI on four real-world tabular imbalance regression datasets, including house pricing, bike sharing, and age prediction. To reflect realistic deployment scenarios, we adopted three types of test distributions: a balanced distribution with uniform target frequencies, a normal distribution that follows the training data, and an inverse distribution that emphasizes rare target regions. On average across these three test distributions, MATI achieved a 7.1% improvement in MAE compared to existing methods.
Authors: Yang Xu, Swetha Ganesh, Vaneet Aggarwal
Abstract: We present the first $Q$-learning and actor-critic algorithms for robust average reward Markov Decision Processes (MDPs) with non-asymptotic convergence under contamination, TV distance and Wasserstein distance uncertainty sets. We show that the robust $Q$ Bellman operator is a strict contractive mapping with respect to a carefully constructed semi-norm with constant functions being quotiented out. This property supports a stochastic approximation update, that learns the optimal robust $Q$ function in $\tilde{\cO}(\epsilon^{-2})$ samples. We also show that the same idea can be used for robust $Q$ function estimation, which can be further used for critic estimation. Coupling it with theories in robust policy mirror descent update, we present a natural actor-critic algorithm that attains an $\epsilon$-optimal robust policy in $\tilde{\cO}(\epsilon^{-3})$ samples. These results advance the theory of distributionally robust reinforcement learning in the average reward setting.
Authors: Jake Robertson, Noah Hollmann, Samuel M\"uller, Noor Awad, Frank Hutter
Abstract: Machine learning (ML) systems are utilized in critical sectors, such as healthcare, law enforcement, and finance. However, these systems are often trained on historical data that contains demographic biases, leading to ML decisions that perpetuate or exacerbate existing social inequalities. Causal fairness provides a transparent, human-in-the-loop framework to mitigate algorithmic discrimination, aligning closely with legal doctrines of direct and indirect discrimination. However, current causal fairness frameworks hold a key limitation in that they assume prior knowledge of the correct causal model, restricting their applicability in complex fairness scenarios where causal models are unknown or difficult to identify. To bridge this gap, we propose FairPFN, a tabular foundation model pre-trained on synthetic causal fairness data to identify and mitigate the causal effects of protected attributes in its predictions. FairPFN's key contribution is that it requires no knowledge of the causal model and still demonstrates strong performance in identifying and removing protected causal effects across a diverse set of hand-crafted and real-world scenarios relative to robust baseline methods. FairPFN paves the way for promising future research, making causal fairness more accessible to a wider variety of complex fairness problems.
Authors: Uri Koren, Navdeep Kumar, Uri Gadot, Giorgia Ramponi, Kfir Yehuda Levy, Shie Mannor
Abstract: Classical policy gradient (PG) methods in reinforcement learning frequently converge to suboptimal local optima, a challenge exacerbated in large or complex environments. This work investigates Policy Gradient with Tree Search (PGTS), an approach that integrates an $m$-step lookahead mechanism to enhance policy optimization. We provide theoretical analysis demonstrating that increasing the tree search depth $m$-monotonically reduces the set of undesirable stationary points and, consequently, improves the worst-case performance of any resulting stationary policy. Critically, our analysis accommodates practical scenarios where policy updates are restricted to states visited by the current policy, rather than requiring updates across the entire state space. Empirical evaluations on diverse MDP structures, including Ladder, Tightrope, and Gridworld environments, illustrate PGTS's ability to exhibit "farsightedness," navigate challenging reward landscapes, escape local traps where standard PG fails, and achieve superior solutions.
Authors: Jiaheng Dong, Hong Jia, Soumyajit Chatterjee, Abhirup Ghosh, James Bailey, Ting Dang
Abstract: Speech Foundation Models encounter significant performance degradation when deployed in real-world scenarios involving acoustic domain shifts, such as background noise and speaker accents. Test-time adaptation (TTA) has recently emerged as a viable strategy to address such domain shifts at inference time without requiring access to source data or labels. However, existing TTA approaches, particularly those relying on backpropagation, are memory-intensive, limiting their applicability in speech tasks and resource-constrained settings. Although backpropagation-free methods offer improved efficiency, existing ones exhibit poor accuracy. This is because they are predominantly developed for vision tasks, which fundamentally differ from speech task formulations, noise characteristics, and model architecture, posing unique transferability challenges. In this paper, we introduce E-BATS, the first Efficient BAckpropagation-free TTA framework designed explicitly for speech foundation models. E-BATS achieves a balance between adaptation effectiveness and memory efficiency through three key components: (i) lightweight prompt adaptation for a forward-pass-based feature alignment, (ii) a multi-scale loss to capture both global (utterance-level) and local distribution shifts (token-level) and (iii) a test-time exponential moving average mechanism for stable adaptation across utterances. Experiments conducted on four noisy speech datasets spanning sixteen acoustic conditions demonstrate consistent improvements, with 4.1%-13.5% accuracy gains over backpropagation-free baselines and 2.0-6.4 times GPU memory savings compared to backpropagation-based methods. By enabling scalable and robust adaptation under acoustic variability, this work paves the way for developing more efficient adaptation approaches for practical speech processing systems in real-world environments.
Authors: Uri Koren, Yonatan Ashlag, Mirco Mutti, Esther Derman, Pierre-Luc Bacon, Shie Mannor
Abstract: State entropy regularization has empirically shown better exploration and sample complexity in reinforcement learning (RL). However, its theoretical guarantees have not been studied. In this paper, we show that state entropy regularization improves robustness to structured and spatially correlated perturbations. These types of variation are common in transfer learning but often overlooked by standard robust RL methods, which typically focus on small, uncorrelated changes. We provide a comprehensive characterization of these robustness properties, including formal guarantees under reward and transition uncertainty, as well as settings where the method performs poorly. Much of our analysis contrasts state entropy with the widely used policy entropy regularization, highlighting their different benefits. Finally, from a practical standpoint, we illustrate that compared with policy entropy, the robustness advantages of state entropy are more sensitive to the number of rollouts used for policy evaluation.
Authors: Ilja Kuzborskij, Yasin Abbasi Yadkori
Abstract: We consider a high-probability non-asymptotic confidence estimation in the $\ell^2$-regularized non-linear least-squares setting with fixed design. In particular, we study confidence estimation for local minimizers of the regularized training loss. We show a pointwise confidence bound, meaning that it holds for the prediction on any given fixed test input $x$. Importantly, the proposed confidence bound scales with similarity of the test input to the training data in the implicit feature space of the predictor (for instance, becoming very large when the test input lies far outside of the training data). This desirable last feature is captured by the weighted norm involving the inverse-Hessian matrix of the objective function, which is a generalized version of its counterpart in the linear setting, $x^{\top} \text{Cov}^{-1} x$. Our generalized result can be regarded as a non-asymptotic counterpart of the classical confidence interval based on asymptotic normality of the MLE estimator. We propose an efficient method for computing the weighted norm, which only mildly exceeds the cost of a gradient computation of the loss function. Finally, we complement our analysis with empirical evidence showing that the proposed confidence bound provides better coverage/width trade-off compared to a confidence estimation by bootstrapping, which is a gold-standard method in many applications involving non-linear predictors such as neural networks.
Authors: Joydeb Kumar Sana, Mohammad M. Masud, M Sohel Rahman, M Saifur Rahman
Abstract: Patient similarity computation (PSC) is a fundamental problem in healthcare informatics. The aim of the patient similarity computation is to measure the similarity among patients according to their historical clinical records, which helps to improve clinical decision support. This paper presents a novel distributed patient similarity computation (DPSC) technique based on data transformation (DT) methods, utilizing an effective combination of time series and static data. Time series data are sensor-collected patients' information, including metrics like heart rate, blood pressure, Oxygen saturation, respiration, etc. The static data are mainly patient background and demographic data, including age, weight, height, gender, etc. Static data has been used for clustering the patients. Before feeding the static data to the machine learning model adaptive Weight-of-Evidence (aWOE) and Z-score data transformation (DT) methods have been performed, which improve the prediction performances. In aWOE-based patient similarity models, sensitive patient information has been processed using aWOE which preserves the data privacy of the trained models. We used the Dynamic Time Warping (DTW) approach, which is robust and very popular, for time series similarity. However, DTW is not suitable for big data due to the significant computational run-time. To overcome this problem, distributed DTW computation is used in this study. For Coronary Artery Disease, our DT based approach boosts prediction performance by as much as 11.4%, 10.20%, and 12.6% in terms of AUC, accuracy, and F-measure, respectively. In the case of Congestive Heart Failure (CHF), our proposed method achieves performance enhancement up to 15.9%, 10.5%, and 21.9% for the same measures, respectively. The proposed method reduces the computation time by as high as 40%.
Authors: Wenying He, Jieling Huang, Junhua Gu, Ji Zhang, Yude Bai
Abstract: Missing data in spatiotemporal systems presents a significant challenge for modern applications, ranging from environmental monitoring to urban traffic management. The integrity of spatiotemporal data often deteriorates due to hardware malfunctions and software failures in real-world deployments. Current approaches based on machine learning and deep learning struggle to model the intricate interdependencies between spatial and temporal dimensions effectively and, more importantly, suffer from cumulative errors during the data imputation process, which propagate and amplify through iterations. To address these limitations, we propose CoFILL, a novel Conditional Diffusion Model for spatiotemporal data imputation. CoFILL builds on the inherent advantages of diffusion models to generate high-quality imputations without relying on potentially error-prone prior estimates. It incorporates an innovative dual-stream architecture that processes temporal and frequency domain features in parallel. By fusing these complementary features, CoFILL captures both rapid fluctuations and underlying patterns in the data, which enables more robust imputation. The extensive experiments reveal that CoFILL's noise prediction network successfully transforms random noise into meaningful values that align with the true data distribution. The results also show that CoFILL outperforms state-of-the-art methods in imputation accuracy. The source code is publicly available at https://github.com/joyHJL/CoFILL.
Authors: Rong-Xi Tan, Ming Chen, Ke Xue, Yao Wang, Yaoyuan Wang, Sheng Fu, Chao Qian
Abstract: The pursuit of universal black-box optimization (BBO) algorithms is a longstanding goal. However, unlike domains such as language or vision, where scaling structured data has driven generalization, progress in offline BBO remains hindered by the lack of unified representations for heterogeneous numerical spaces. Thus, existing offline BBO approaches are constrained to single-task and fixed-dimensional settings, failing to achieve cross-domain universal optimization. Recent advances in language models (LMs) offer a promising path forward: their embeddings capture latent relationships in a unifying way, enabling universal optimization across different data types possible. In this paper, we discuss multiple potential approaches, including an end-to-end learning framework in the form of next-token prediction, as well as prioritizing the learning of latent spaces with strong representational capabilities. To validate the effectiveness of these methods, we collect offline BBO tasks and data from open-source academic works for training. Experiments demonstrate the universality and effectiveness of our proposed methods. Our findings suggest that unifying language model priors and learning string embedding space can overcome traditional barriers in universal BBO, paving the way for general-purpose BBO algorithms. The code is provided at https://github.com/lamda-bbo/universal-offline-bbo.
Authors: Ren-Jian Wang, Ke Xue, Zeyu Qin, Ziniu Li, Sheng Tang, Hao-Tian Li, Shengcai Liu, Chao Qian
Abstract: Ensuring safety of large language models (LLMs) is important. Red teaming--a systematic approach to identifying adversarial prompts that elicit harmful responses from target LLMs--has emerged as a crucial safety evaluation method. Within this framework, the diversity of adversarial prompts is essential for comprehensive safety assessments. We find that previous approaches to red-teaming may suffer from two key limitations. First, they often pursue diversity through simplistic metrics like word frequency or sentence embedding similarity, which may not capture meaningful variation in attack strategies. Second, the common practice of training a single attacker model restricts coverage across potential attack styles and risk categories. This paper introduces Quality-Diversity Red-Teaming (QDRT), a new framework designed to address these limitations. QDRT achieves goal-driven diversity through behavior-conditioned training and implements a behavioral replay buffer in an open-ended manner. Additionally, it trains multiple specialized attackers capable of generating high-quality attacks across diverse styles and risk categories. Our empirical evaluation demonstrates that QDRT generates attacks that are both more diverse and more effective against a wide range of target LLMs, including GPT-2, Llama-3, Gemma-2, and Qwen2.5. This work advances the field of LLM safety by providing a systematic and effective approach to automated red-teaming, ultimately supporting the responsible deployment of LLMs.
Authors: Eshwar S. R., Gugan Thoppe, Aditya Gopalan, Gal Dalal
Abstract: Despite decades of research, it remains challenging to correctly use Reinforcement Learning (RL) algorithms with function approximation. A prime example is policy iteration, whose fundamental guarantee of monotonic improvement collapses even under linear function approximation. To address this issue, we introduce Reliable Policy Iteration (RPI). It replaces the common projection or Bellman-error minimization during policy evaluation with a Bellman-based constrained optimization. We prove that not only does RPI confer textbook monotonicity on its value estimates but these estimates also lower bound the true return. Also, their limit partially satisfies the unprojected Bellman equation, emphasizing RPI's natural fit within RL. RPI is the first algorithm with such monotonicity and convergence guarantees under function approximation. For practical use, we provide a model-free variant of RPI that amounts to a novel critic. It can be readily integrated into primary model-free PI implementations such as DQN and DDPG. In classical control tasks, such RPI-enhanced variants consistently maintain their lower-bound guarantee while matching or surpassing the performance of all baseline methods.
Authors: Qi Liu, Jingqing Ruan, Hao Li, Haodong Zhao, Desheng Wang, Jiansong Chen, Wan Guanglu, Xunliang Cai, Zhi Zheng, Tong Xu
Abstract: Existing multi-objective preference alignment methods for large language models (LLMs) face limitations: (1) the inability to effectively balance various preference dimensions, and (2) reliance on auxiliary reward/reference models introduces computational complexity. To address these challenges, we propose Adaptive Multi-objective Preference Optimization (AMoPO), a novel framework that achieves dynamic balance across preference dimensions. By introducing the multi-objective optimization paradigm to use the dimension-aware generation metrics as implicit rewards, AMoPO aligns LLMs with diverse preferences without additional reward models or reference models. We introduce an adaptive weight assignment mechanism that models the generation space as a Gaussian distribution, allowing dynamic prioritization of preference dimensions. Empirical results demonstrate that AMoPO outperforms state-of-the-art baselines by 28.5%, and the experiments on 7B, 14B, and 32B models reveal the scaling ability of AMoPO. Moreover, additional analysis of multiple dimensions verifies its adaptability and effectiveness. These findings validate AMoPO's capability to achieve dimension-aware preference alignment, highlighting its superiority. Our codes and datasets are available at https://github.com/Javkonline/AMoPO.
Authors: Huanyi Xie, Lijie Hu, Lu Yu, Tianhao Huang, Longfei Li, Meng Li, Jun Zhou, Huan Wang, Di Wang
Abstract: In the realm of Text-attributed Graphs (TAGs), traditional graph neural networks (GNNs) often fall short due to the complex textual information associated with each node. Recent methods have improved node representations by leveraging large language models (LLMs) to enhance node text features, but these approaches typically require extensive annotations or fine-tuning across all nodes, which is both time-consuming and costly. To overcome these challenges, we introduce GAGA, an efficient framework for TAG representation learning. GAGA reduces annotation time and cost by focusing on annotating only representative nodes and edges. It constructs an annotation graph that captures the topological relationships among these annotations. Furthermore, GAGA employs a two-level alignment module to effectively integrate the annotation graph with the TAG, aligning their underlying structures. Experiments show that GAGA achieves classification accuracies on par with or surpassing state-of-the-art methods while requiring only 1% of the data to be annotated, demonstrating its high efficiency.
Authors: Kaiqi Wu, Weiyang Kong, Sen Zhang, Yubao Liu, Zitong Chen
Abstract: Traffic prediction is a critical task in spatial-temporal forecasting with broad applications in travel planning and urban management. Adaptive graph convolution networks have emerged as mainstream solutions due to their ability to learn node embeddings in a data-driven manner and capture complex latent dependencies. However, existing adaptive graph learning methods for traffic forecasting often either ignore the regularization of node embeddings, which account for a significant proportion of model parameters, or face scalability issues from expensive graph convolution operations. To address these challenges, we propose a Regularized Adaptive Graph Learning (RAGL) model. First, we introduce a regularized adaptive graph learning framework that synergizes Stochastic Shared Embedding (SSE) and adaptive graph convolution via a residual difference mechanism, achieving both embedding regularization and noise suppression. Second, to ensure scalability on large road networks, we develop the Efficient Cosine Operator (ECO), which performs graph convolution based on the cosine similarity of regularized embeddings with linear time complexity. Extensive experiments on four large-scale real-world traffic datasets show that RAGL consistently outperforms state-of-the-art methods in terms of prediction accuracy and exhibits competitive computational efficiency.
Authors: J. C. Husillos, A. Gallego, A. Roma, A. Troncoso
Abstract: In this paper, we present a novel learning approach based on Neurovectors, an innovative paradigm that structures information through interconnected nodes and vector relationships for tabular data processing. Unlike traditional artificial neural networks that rely on weight adjustment through backpropagation, Neurovectors encode information by structuring data in vector spaces where energy propagation, rather than traditional weight updates, drives the learning process, enabling a more adaptable and explainable learning process. Our method generates dynamic representations of knowledge through neurovectors, thereby improving both the interpretability and efficiency of the predictive model. Experimental results using datasets from well-established repositories such as the UCI machine learning repository and Kaggle are reported both for classification and regression. To evaluate its performance, we compare our approach with standard machine learning and deep learning models, showing that Neurovectors achieve competitive accuracy.
Authors: Ramisa Farha, Joshua O. Olukoya
Abstract: This study employs a robust analytical framework to uncover patterns in survival outcomes among breast cancer patients from diverse racial and geographical backgrounds. This research uses the SEER 2021 dataset to analyze breast cancer survival outcomes to identify and comprehend dissimilarities. Our approach integrates exploratory data analysis (EDA), through this we identify key variables that influence survival rates and employ survival analysis techniques, including the Kaplan-Meier estimator and log-rank test and the advanced modeling Cox Proportional Hazards model to determine how survival rates vary across racial groups and countries. Model validation and interpretation are undertaken to ensure the reliability of our findings, which are documented comprehensively to inform policymakers and healthcare professionals. The outcome of this paper is a detailed version of statistical analysis that not just highlights disparities in breast cancer treatment and care but also serves as a foundational tool for developing targeted interventions to address the inequalities effectively. Through this research, our aim is to contribute to the global efforts to improve breast cancer outcomes and reduce treatment disparities.
Authors: Tianci Bu, Chuanrui Wang, Hao Ma, Haoren Zheng, Xin Lu, Tailin Wu
Abstract: Generating graphs with hierarchical structures remains a fundamental challenge due to the limitations of Euclidean geometry in capturing exponential complexity. Here we introduce \textbf{GGBall}, a novel hyperbolic framework for graph generation that integrates geometric inductive biases with modern generative paradigms. GGBall combines a Hyperbolic Vector-Quantized Autoencoder (HVQVAE) with a Riemannian flow matching prior defined via closed-form geodesics. This design enables flow-based priors to model complex latent distributions, while vector quantization helps preserve the curvature-aware structure of the hyperbolic space. We further develop a suite of hyperbolic GNN and Transformer layers that operate entirely within the manifold, ensuring stability and scalability. Empirically, our model reduces degree MMD by over 75\% on Community-Small and over 40\% on Ego-Small compared to state-of-the-art baselines, demonstrating an improved ability to preserve topological hierarchies. These results highlight the potential of hyperbolic geometry as a powerful foundation for the generative modeling of complex, structured, and hierarchical data domains. Our code is available at \href{https://github.com/AI4Science-WestlakeU/GGBall}{here}.
Authors: Tong Xiao, Xin Xu, Zhenya Huang, Hongyu Gao, Quan Liu, Qi Liu, Enhong Chen
Abstract: Enhancing the multimodal reasoning capabilities of Multimodal Large Language Models (MLLMs) is a challenging task that has attracted increasing attention in the community. Recently, several studies have applied Reinforcement Learning with Verifiable Rewards (RLVR) to the multimodal domain in order to enhance the reasoning abilities of MLLMs. However, these works largely overlook the enhancement of multimodal perception capabilities in MLLMs, which serve as a core prerequisite and foundational component of complex multimodal reasoning. Through McNemar's test, we find that existing RLVR method fails to effectively enhance the multimodal perception capabilities of MLLMs, thereby limiting their further improvement in multimodal reasoning. To address this limitation, we propose Perception-R1, which introduces a novel visual perception reward that explicitly encourages MLLMs to perceive the visual content accurately, thereby can effectively incentivizing both their multimodal perception and reasoning capabilities. Specifically, we first collect textual visual annotations from the CoT trajectories of multimodal problems, which will serve as visual references for reward assignment. During RLVR training, we employ a judging LLM to assess the consistency between the visual annotations and the responses generated by MLLM, and assign the visual perception reward based on these consistency judgments. Extensive experiments on several multimodal reasoning benchmarks demonstrate the effectiveness of our Perception-R1, which achieves state-of-the-art performance on most benchmarks using only 1,442 training data.
Authors: Mateusz Gajewski, Miko{\l}aj Morzy, Adam Karczmarz, Piotr Sankowski
Abstract: Existing feature attribution methods like SHAP often suffer from global dependence, failing to capture true local model behavior. This paper introduces VARSHAP, a novel model-agnostic local feature attribution method which uses the reduction of prediction variance as the key importance metric of features. Building upon Shapley value framework, VARSHAP satisfies the key Shapley axioms, but, unlike SHAP, is resilient to global data distribution shifts. Experiments on synthetic and real-world datasets demonstrate that VARSHAP outperforms popular methods such as KernelSHAP or LIME, both quantitatively and qualitatively.
Authors: Roy Eisenstadt, Itamar Zimerman, Lior Wolf
Abstract: Recently, techniques such as explicit structured reasoning have demonstrated strong test-time scaling behavior by enforcing a separation between the model's internal "thinking" process and the final response. A key factor influencing answer quality in this setting is the length of the thinking stage. When the reasoning is too short, the model may fail to capture the complexity of the task. Conversely, when it is too long, the model may overthink, leading to unnecessary computation and degraded performance. This paper explores and exploits the underlying mechanisms by which LLMs understand and regulate the length of their reasoning during explicit thought processes. First, we show that LLMs encode their progress through the reasoning process and introduce an interactive progress bar visualization, which is then used to reveal insights on the model's planning dynamics. Second, we manipulate the internal progress encoding during inference to reduce unnecessary steps and generate a more concise and decisive chain of thoughts. Our empirical results demonstrate that this "overclocking" method mitigates overthinking, improves answer accuracy, and reduces inference latency. Our code is publicly available.
Authors: Ngoc-Quan Pham, Tuan Truong, Quyen Tran, Tan Nguyen, Dinh Phung, Trung Le
Abstract: We introduce Interactive Bayesian Distributional Robustness (IBDR), a novel Bayesian inference framework that allows modeling the interactions between particles, thereby enhancing ensemble quality through increased particle diversity. IBDR is grounded in a generalized theoretical framework that connects the distributional population loss with the approximate posterior, motivating a practical dual optimization procedure that enforces distributional robustness while fostering particle diversity. We evaluate IBDR's performance against various baseline methods using the VTAB-1K benchmark and the common reasoning language task. The results consistently show that IBDR outperforms these baselines, underscoring its effectiveness in real-world applications.
Authors: Kevin Frans, Sergey Levine, Pieter Abbeel
Abstract: In this work, we take an experimentally grounded look at neural network optimization. Building on the Shampoo family of algorithms, we identify and alleviate three key issues, resulting in the proposed SPlus method. First, we find that naive Shampoo is prone to divergence when matrix-inverses are cached for long periods. We introduce an alternate bounded update combining a historical eigenbasis with instantaneous normalization, resulting in across-the-board stability and significantly lower computational requirements. Second, we adapt a shape-aware scaling to enable learning rate transfer across network width. Third, we find that high learning rates result in large parameter noise, and propose a simple iterate-averaging scheme which unblocks faster learning. To properly confirm these findings, we introduce a pointed Transformer training benchmark, considering three objectives (language modelling, image classification, and diffusion modelling) across different stages of training. On average, SPlus is able to reach the validation performance of Adam within 44% of the gradient steps and 62% of the wallclock time.
Authors: Alex Clinton, Thomas Zeng, Yiding Chen, Xiaojin Zhu, Kirthevasan Kandasamy
Abstract: Modern data marketplaces and data sharing consortia increasingly rely on incentive mechanisms to encourage agents to contribute data. However, schemes that reward agents based on the quantity of submitted data are vulnerable to manipulation, as agents may submit fabricated or low-quality data to inflate their rewards. Prior work has proposed comparing each agent's data against others' to promote honesty: when others contribute genuine data, the best way to minimize discrepancy is to do the same. Yet prior implementations of this idea rely on very strong assumptions about the data distribution (e.g. Gaussian), limiting their applicability. In this work, we develop reward mechanisms based on a novel, two-sample test inspired by the Cram\'er-von Mises statistic. Our methods strictly incentivize agents to submit more genuine data, while disincentivizing data fabrication and other types of untruthful reporting. We establish that truthful reporting constitutes a (possibly approximate) Nash equilibrium in both Bayesian and prior-agnostic settings. We theoretically instantiate our method in three canonical data sharing problems and show that it relaxes key assumptions made by prior work. Empirically, we demonstrate that our mechanism incentivizes truthful data sharing via simulations and on real-world language and image data.
Authors: Haochen Song, Dominik Hofer, Rania Islambouli, Laura Hawkins, Ananya Bhattacharjee, Meredith Franklin, Joseph Jay Williams
Abstract: Machine learning approaches, such as contextual multi-armed bandit (cMAB) algorithms, offer a promising strategy to reduce sedentary behavior by delivering personalized interventions to encourage physical activity. However, cMAB algorithms typically require large participant samples to learn effectively and may overlook key psychological factors that are not explicitly encoded in the model. In this study, we propose a hybrid approach that combines cMAB for selecting intervention types with large language models (LLMs) to personalize message content. We evaluate four intervention types: behavioral self-monitoring, gain-framed, loss-framed, and social comparison, each delivered as a motivational message aimed at increasing motivation for physical activity and daily step count. Message content is further personalized using dynamic contextual factors including daily fluctuations in self-efficacy, social influence, and regulatory focus. Over a seven-day trial, participants receive daily messages assigned by one of four models: cMAB alone, LLM alone, combined cMAB with LLM personalization (cMABxLLM), or equal randomization (RCT). Outcomes include daily step count and message acceptance, assessed via ecological momentary assessments (EMAs). We apply a causal inference framework to evaluate the effects of each model. Our findings offer new insights into the complementary roles of LLM-based personalization and cMAB adaptation in promoting physical activity through personalized behavioral messaging.
Authors: Suho Shin, Chenghao Yang, Haifeng Xu, Mohammad T. Hajiaghayi
Abstract: We introduce the tokenized linear bandit (TLB) and multi-armed bandit (TMAB), variants of linear and stochastic multi-armed bandit problems inspired by LLM decoding and alignment. In these problems, at each round $t \in [T]$, a user submits a query (context), and the decision maker (DM) sequentially selects a token irrevocably from a token set. Once the sequence is complete, the DM observes a random utility from the user, whose expectation is presented by a sequence function mapping the chosen token sequence to a nonnegative real value that depends on the query. In both problems, we first show that learning is impossible without any structure on the sequence function. We introduce a natural assumption, diminishing distance with more commons (DDMC), and propose algorithms with regret $\tilde{O}(L\sqrt{T})$ and $\tilde{O}(L\sqrt{T^{2/3}})$ for TLB and TMAB, respectively. As a side product, we obtain an (almost) optimality of the greedy decoding for LLM decoding algorithm under DDMC, which justifies the unresaonable effectiveness of greedy decoding in several tasks. This also has an immediate application to decoding-time LLM alignment, when the misaligned utility can be represented as the frozen LLM's utility and a linearly realizable latent function. We finally validate our algorithm's performance empirically as well as verify our assumptions using synthetic and real-world datasets.
Authors: Weijie Guan, Haohui Wang, Jian Kang, Lihui Liu, Dawei Zhou
Abstract: Graph learning has been crucial to many real-world tasks, but they are often studied with a closed-world assumption, with all possible labels of data known a priori. To enable effective graph learning in an open and noisy environment, it is critical to inform the model users when the model makes a wrong prediction to in-distribution data of a known class, i.e., misclassification detection or when the model encounters out-of-distribution from novel classes, i.e., out-of-distribution detection. This paper introduces Evidential Reasoning Network (EVINET), a framework that addresses these two challenges by integrating Beta embedding within a subjective logic framework. EVINET includes two key modules: Dissonance Reasoning for misclassification detection and Vacuity Reasoning for out-of-distribution detection. Extensive experiments demonstrate that EVINET outperforms state-of-the-art methods across multiple metrics in the tasks of in-distribution classification, misclassification detection, and out-of-distribution detection. EVINET demonstrates the necessity of uncertainty estimation and logical reasoning for misclassification detection and out-of-distribution detection and paves the way for open-world graph learning. Our code and data are available at https://github.com/SSSKJ/EviNET.
Authors: Yijia Dai, Zhaolin Gao, Yahya Satter, Sarah Dean, Jennifer J. Sun
Abstract: Hidden Markov Models (HMMs) are foundational tools for modeling sequential data with latent Markovian structure, yet fitting them to real-world data remains computationally challenging. In this work, we show that pre-trained large language models (LLMs) can effectively model data generated by HMMs via in-context learning (ICL)$\unicode{x2013}$their ability to infer patterns from examples within a prompt. On a diverse set of synthetic HMMs, LLMs achieve predictive accuracy approaching the theoretical optimum. We uncover novel scaling trends influenced by HMM properties, and offer theoretical conjectures for these empirical observations. We also provide practical guidelines for scientists on using ICL as a diagnostic tool for complex data. On real-world animal decision-making tasks, ICL achieves competitive performance with models designed by human experts. To our knowledge, this is the first demonstration that ICL can learn and predict HMM-generated sequences$\unicode{x2013}$an advance that deepens our understanding of in-context learning in LLMs and establishes its potential as a powerful tool for uncovering hidden structure in complex scientific data.
Authors: Yizhuo Chen (Richard), Chun-Fu (Richard), Chen, Hsiang Hsu, Shaohan Hu, Tarek Abdelzaher
Abstract: The growing Machine Learning (ML) services require extensive collections of user data, which may inadvertently include people's private information irrelevant to the services. Various studies have been proposed to protect private attributes by removing them from the data while maintaining the utilities of the data for downstream tasks. Nevertheless, as we theoretically and empirically show in the paper, these methods reveal severe vulnerability because of a common weakness rooted in their adversarial training based strategies. To overcome this limitation, we propose a novel approach, PASS, designed to stochastically substitute the original sample with another one according to certain probabilities, which is trained with a novel loss function soundly derived from information-theoretic objective defined for utility-preserving private attributes protection. The comprehensive evaluation of PASS on various datasets of different modalities, including facial images, human activity sensory signals, and voice recording datasets, substantiates PASS's effectiveness and generalizability.
Authors: Thomas Joshi, Herman Saini, Neil Dhillon, Antoni Viros i Martin, Kaoutar El Maghraoui
Abstract: Large Language Models (LLMs) encounter severe memory inefficiencies during long-context inference due to conventional handling of key-value (KV) caches. In this work, we introduce a novel integration of PagedAttention with PyTorch's FlexAttention, addressing internal fragmentation and inefficiencies associated with monolithic KV cache allocations. Implemented within IBM's Foundation Model Stack (FMS), our fused attention kernel efficiently gathers scattered KV data. Our benchmarks on an NVIDIA L4 GPU (24GB) demonstrate significantly reduced inference latency, growing only linearly (~2x) with sequence length from 128 to 2048 tokens when utilizing a global KV cache, compared to exponential latency increases without caching. While peak memory usage remains largely unchanged for single-step evaluations (dominated by model weights and activations), paged attention causes minimal incremental memory usage, observable only at sequence lengths exceeding 2048 tokens due to its power-of-two cache allocations. We open-source the full implementation and discuss its implications for future long-context model deployment.
Authors: Yusuf Elnady
Abstract: Many security and network applications require having large datasets to train the machine learning models. Limited data access is a well-known problem in the security domain. Recent studies have shown the potential of Transformer models to enlarge the size of data by synthesizing new samples, but the synthesized samples don't improve the models over the real data. To address this issue, we design an efficient transformer-based model as a generative framework to generate time-series data, that can be used to boost the performance of existing and new ML workflows. Our new transformer model achieves the SOTA results. We style our model to be generalizable and work across different datasets, and produce high-quality samples.
Authors: David Millard, Arielle Carr, St\'ephane Gaudreault, Ali Baheri
Abstract: We present DEF (\textbf{\ul{D}}iffusion-augmented \textbf{\ul{E}}nsemble \textbf{\ul{F}}orecasting), a novel approach for generating initial condition perturbations. Modern approaches to initial condition perturbations are primarily designed for numerical weather prediction (NWP) solvers, limiting their applicability in the rapidly growing field of machine learning for weather prediction. Consequently, stochastic models in this domain are often developed on a case-by-case basis. We demonstrate that a simple conditional diffusion model can (1) generate meaningful structured perturbations, (2) be applied iteratively, and (3) utilize a guidance term to intuitivey control the level of perturbation. This method enables the transformation of any deterministic neural forecasting system into a stochastic one. With our stochastic extended systems, we show that the model accumulates less error over long-term forecasts while producing meaningful forecast distributions. We validate our approach on the 5.625$^\circ$ ERA5 reanalysis dataset, which comprises atmospheric and surface variables over a discretized global grid, spanning from the 1960s to the present. On this dataset, our method demonstrates improved predictive performance along with reasonable spread estimates.
Authors: Jintao Yan, Tan Chen, Yuxuan Sun, Zhaojun Nan, Sheng Zhou, Zhisheng Niu
Abstract: Asynchronous Federated Learning (AFL) enables distributed model training across multiple mobile devices, allowing each device to independently update its local model without waiting for others. However, device mobility introduces intermittent connectivity, which necessitates gradient sparsification and leads to model staleness, jointly affecting AFL convergence. This paper develops a theoretical model to characterize the interplay among sparsification, model staleness and mobility-induced contact patterns, and their joint impact on AFL convergence. Based on the analysis, we propose a mobility-aware dynamic sparsification (MADS) algorithm that optimizes the sparsification degree based on contact time and model staleness. Closed-form solutions are derived, showing that under low-speed conditions, MADS increases the sparsification degree to enhance convergence, while under high-speed conditions, it reduces the sparsification degree to guarantee reliable uploads within limited contact time. Experimental results validate the theoretical findings. Compared with the state-of-the-art benchmarks, the MADS algorithm increases the image classification accuracy on the CIFAR-10 dataset by 8.76% and reduces the average displacement error in the Argoverse trajectory prediction dataset by 9.46%.
Authors: Yash Datta, Sharath Rajasekar
Abstract: We present JavelinGuard, a suite of low-cost, high-performance model architectures designed for detecting malicious intent in Large Language Model (LLM) interactions, optimized specifically for production deployment. Recent advances in transformer architectures, including compact BERT(Devlin et al. 2019) variants (e.g., ModernBERT (Warner et al. 2024)), allow us to build highly accurate classifiers with as few as approximately 400M parameters that achieve rapid inference speeds even on standard CPU hardware. We systematically explore five progressively sophisticated transformer-based architectures: Sharanga (baseline transformer classifier), Mahendra (enhanced attention-weighted pooling with deeper heads), Vaishnava and Ashwina (hybrid neural ensemble architectures), and Raudra (an advanced multi-task framework with specialized loss functions). Our models are rigorously benchmarked across nine diverse adversarial datasets, including popular sets like the NotInject series, BIPIA, Garak, ImprovedLLM, ToxicChat, WildGuard, and our newly introduced JavelinBench, specifically crafted to test generalization on challenging borderline and hard-negative cases. Additionally, we compare our architectures against leading open-source guardrail models as well as large decoder-only LLMs such as gpt-4o, demonstrating superior cost-performance trade-offs in terms of accuracy, and latency. Our findings reveal that while Raudra's multi-task design offers the most robust performance overall, each architecture presents unique trade-offs in speed, interpretability, and resource requirements, guiding practitioners in selecting the optimal balance of complexity and efficiency for real-world LLM security applications.
Authors: Haoyu Wang, Peihao Wang, Mufei Li, Shikun Liu, Siqi Miao, Zhangyang Wang, Pan Li
Abstract: Modern large language models (LLMs) are inherently auto-regressive, requiring input to be serialized into flat sequences regardless of their structural dependencies. This serialization hinders the model's ability to leverage structural inductive biases, especially in tasks such as retrieval-augmented generation (RAG) and reasoning on data with native graph structures, where inter-segment dependencies are crucial. We introduce Graph-KV with the potential to overcome this limitation. Graph-KV leverages the KV-cache of text segments as condensed representations and governs their interaction through structural inductive biases. In this framework, 'target' segments selectively attend only to the KV-caches of their designated 'source' segments, rather than all preceding segments in a serialized sequence. This approach induces a graph-structured block mask, sparsifying attention and enabling a message-passing-like step within the LLM. Furthermore, strategically allocated positional encodings for source and target segments reduce positional bias and context window consumption. We evaluate Graph-KV across three scenarios: (1) seven RAG benchmarks spanning direct inference, multi-hop reasoning, and long-document understanding; (2) Arxiv-QA, a novel academic paper QA task with full-text scientific papers structured as citation ego-graphs; and (3) paper topic classification within a citation network. By effectively reducing positional bias and harnessing structural inductive biases, Graph-KV substantially outperforms baselines, including standard costly sequential encoding, across various settings. Code and the Graph-KV data are publicly available.
Authors: Yuya Okada, Takayuki Nishio
Abstract: We propose SALT (Split-Adaptive Lightweight Tuning), a lightweight model adaptation framework for Split Computing under closed constraints, where the head and tail networks are proprietary and inaccessible to users. In such closed environments, conventional adaptation methods are infeasible since they require access to model parameters or architectures. SALT addresses this challenge by introducing a compact, trainable adapter on the client side to refine latent features from the head network, enabling user-specific adaptation without modifying the original models or increasing communication overhead. We evaluate SALT on user-specific classification tasks with CIFAR-10 and CIFAR-100, demonstrating improved accuracy with lower training latency compared to fine-tuning methods. Furthermore, SALT facilitates model adaptation for robust inference over lossy networks, a common challenge in edge-cloud environments. With minimal deployment overhead, SALT offers a practical solution for personalized inference in edge AI systems under strict system constraints.
Authors: Haiyue Ma, Zhixu Du, Yiran Chen
Abstract: In multi-GPU Mixture-of-Experts (MoE) network, experts are distributed across different GPUs, which creates load imbalance as each expert processes different number of tokens. Recent works improve MoE inference load balance by dynamically duplicating popular experts to more GPUs to process excessive tokens, which requires predicting the distribution before routing. In this paper, we discuss the tradeoff of prediction strategies, accuracies, overhead, and end-to-end system performance. We propose MoE-GPS, a framework that guides the selection of the optimal predictor design under various system configurations, by quantifying the performance impact to system-level model runtime. Specifically, we advocate for Distribution-Only Prediction, a prediction strategy that only predicts overall token distribution which significantly reduces overhead compared to the traditional Token-to-Expert Prediction. On Mixtral 8x7B MMLU dataset, MoE-GPS suggests Distribution-Only Prediction which improves end-to-end inference performance by more than 23% compared with Token-to-Expert Prediction.
Authors: Yuen Chen, Haozhe Si, Guojun Zhang, Han Zhao
Abstract: Domain generalization (DG) seeks to develop models that generalize well to unseen target domains, addressing the prevalent issue of distribution shifts in real-world applications. One line of research in DG focuses on aligning domain-level gradients and Hessians to enhance generalization. However, existing methods are computationally inefficient and the underlying principles of these approaches are not well understood. In this paper, we develop the theory of moment alignment for DG. Grounded in \textit{transfer measure}, a principled framework for quantifying generalizability between two domains, we first extend the definition of transfer measure to domain generalization that includes multiple source domains and establish a target error bound. Then, we prove that aligning derivatives across domains improves transfer measure both when the feature extractor induces an invariant optimal predictor across domains and when it does not. Notably, moment alignment provides a unifying understanding of Invariant Risk Minimization, gradient matching, and Hessian matching, three previously disconnected approaches to DG. We further connect feature moments and derivatives of the classifier head, and establish the duality between feature learning and classifier fitting. Building upon our theory, we introduce \textbf{C}losed-Form \textbf{M}oment \textbf{A}lignment (CMA), a novel DG algorithm that aligns domain-level gradients and Hessians in closed-form. Our method overcomes the computational inefficiencies of existing gradient and Hessian-based techniques by eliminating the need for repeated backpropagation or sampling-based Hessian estimation. We validate the efficacy of our approach through two sets of experiments: linear probing and full fine-tuning. CMA demonstrates superior performance in both settings compared to Empirical Risk Minimization and state-of-the-art algorithms.
Authors: Zhongping Ji
Abstract: This research endeavors to offer insights into unlocking the further potential of transformer-based architectures. One of the primary motivations is to offer a geometric interpretation for the attention mechanism in transformers. In our framework, the attention mainly involves metric tensors, tangent spaces, inner product, and how they relate to each other. These quantities and structures at discrete positions are intricately interconnected via the parallel transport of tangent vectors. To make the learning process more efficient, we reduce the number of parameters through ingenious predefined configurations. Moreover, we introduce an explicit mechanism to highlight a neighborhood by attenuating the remote values, given that transformers inherently neglect local inductive bias. Experimental results demonstrate that our modules deliver significant performance improvements relative to the baseline. More evaluation experiments on visual and large language models will be launched successively.
Authors: Yifan Luo, Zhennan Zhou, Bin Dong
Abstract: Understanding the internal representations of large language models (LLMs) is a central challenge in interpretability research. Existing feature interpretability methods often rely on strong assumptions about the structure of representations that may not hold in practice. In this work, we introduce InverseScope, an assumption-light and scalable framework for interpreting neural activations via input inversion. Given a target activation, we define a distribution over inputs that generate similar activations and analyze this distribution to infer the encoded features. To address the inefficiency of sampling in high-dimensional spaces, we propose a novel conditional generation architecture that significantly improves sample efficiency compared to previous methods. We further introduce a quantitative evaluation protocol that tests interpretability hypotheses using feature consistency rate computed over the sampled inputs. InverseScope scales inversion-based interpretability methods to larger models and practical tasks, enabling systematic and quantitative analysis of internal representations in real-world LLMs.
Authors: Yihong Jin, Ze Yang, Juntian Liu, Xinhe Xu
Abstract: With the rapid development of multi-cloud environments, it is increasingly important to ensure the security and reliability of intelligent monitoring systems. In this paper, we propose an anomaly detection and early warning mechanism for intelligent monitoring system in multi-cloud environment based on Large-Scale Language Model (LLM). On the basis of the existing monitoring framework, the proposed model innovatively introduces a multi-level feature extraction method, which combines the natural language processing ability of LLM with traditional machine learning methods to enhance the accuracy of anomaly detection and improve the real-time response efficiency. By introducing the contextual understanding capabilities of LLMs, the model dynamically adapts to different cloud service providers and environments, so as to more effectively detect abnormal patterns and predict potential failures. Experimental results show that the proposed model is significantly better than the traditional anomaly detection system in terms of detection accuracy and latency, and significantly improves the resilience and active management ability of cloud infrastructure.
Authors: Xiaojun zhou, Chunna Zhao, Yaqun Huang, Chengli Zhou, Junjie Ye, Kemeng Xiang
Abstract: Fractional-order differentiation has many characteristics different from integer-order differentiation. These characteristics can be applied to the optimization algorithms of artificial neural networks to obtain better results. However, due to insufficient theoretical research, at present, there is no fractional-order matrix differentiation method that is perfectly compatible with automatic differentiation (Autograd) technology. Therefore, we propose a fractional-order matrix differentiation calculation method. This method is introduced by the definition of the integer-order Jacobian matrix. We denote it as fractional-order Jacobian matrix differentiation (${{\bf{J}}^\alpha }$). Through ${{\bf{J}}^\alpha }$, we can carry out the matrix-based fractional-order chain rule. Based on the Linear module and the fractional-order differentiation, we design the fractional-order Autograd technology to enable the use of fractional-order differentiation in hidden layers, thereby enhancing the practicality of fractional-order differentiation in deep learning. In the experiment, according to the PyTorch framework, we design fractional-order Linear (FLinear) and replace nn.Linear in the multilayer perceptron with FLinear. Through the qualitative analysis of the training set and validation set $Loss$, the quantitative analysis of the test set indicators, and the analysis of time consumption and GPU memory usage during model training, we verify the superior performance of ${{\bf{J}}^\alpha }$ and prove that it is an excellent fractional-order gradient descent method in the field of deep learning.
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: Jin Huang, Yuchao Jin, Le An, Josh Park
Abstract: This paper introduces an efficient Vision-Language Model (VLM) pipeline specifically optimized for deployment on embedded devices, such as those used in robotics and autonomous driving. The pipeline significantly reduces the computational overhead by jointly leveraging patch selection to filter irrelevant camera views, a token selection module to reduce input sequence length for the LLM, and speculative decoding to accelerate token generation. Evaluation on the NVIDIA DRIVE Thor platform for automonous driving application, our pipeline achieves $2.5\times$ end-to-end latency reduction without compromising task accuracy. The speed-up further increases to $3.2\times$ when applying FP8 post-training quantization. These results demonstrate our pipeline as a viable solution for enabling real-time VLM deployment in resource-constrained environments.
Authors: Nan Sun, Xixun Lin, Zhiheng Zhou, Yanmin Shang, Zhenlin Cheng, Yanan Cao
Abstract: Recently, Out-of-distribution (OOD) detection in dynamic graphs, which aims to identify whether incoming data deviates from the distribution of the in-distribution (ID) training set, has garnered considerable attention in security-sensitive fields. Current OOD detection paradigms primarily focus on static graphs and confront two critical challenges: i) high bias and high variance caused by single-point estimation, which makes the predictions sensitive to randomness in the data; ii) score homogenization resulting from the lack of OOD training data, where the model only learns ID-specific patterns, resulting in overall low OOD scores and a narrow score gap between ID and OOD data. To tackle these issues, we first investigate OOD detection in dynamic graphs through the lens of Evidential Deep Learning (EDL). Specifically, we propose EviSEC, an innovative and effective OOD detector via Evidential Spectrum-awarE Contrastive Learning. We design an evidential neural network to redefine the output as the posterior Dirichlet distribution, explaining the randomness of inputs through the uncertainty of distribution, which is overlooked by single-point estimation. Moreover, spectrum-aware augmentation module generates OOD approximations to identify patterns with high OOD scores, thereby widening the score gap between ID and OOD data and mitigating score homogenization. Extensive experiments on real-world datasets demonstrate that EviSAC effectively detects OOD samples in dynamic graphs.
Authors: Ruhan Wang, Zhiyong Wang, Chengkai Huang, Rui Wang, Tong Yu, Lina Yao, John C. S. Lui, Dongruo Zhou
Abstract: For question-answering (QA) tasks, in-context learning (ICL) enables language models to generate responses without modifying their parameters by leveraging examples provided in the input. However, the effectiveness of ICL heavily depends on the availability of high-quality examples, which are often scarce due to data privacy constraints, annotation costs, and distribution disparities. A natural solution is to utilize examples stored on client devices, but existing approaches either require transmitting model parameters - incurring significant communication overhead - or fail to fully exploit local datasets, limiting their effectiveness. To address these challenges, we propose Federated In-Context Learning (Fed-ICL), a general framework that enhances ICL through an iterative, collaborative process. Fed-ICL progressively refines responses by leveraging multi-round interactions between clients and a central server, improving answer quality without the need to transmit model parameters. We establish theoretical guarantees for the convergence of Fed-ICL and conduct extensive experiments on standard QA benchmarks, demonstrating that our proposed approach achieves strong performance while maintaining low communication costs.
Authors: T. Duy Nguyen-Hien, Desi R. Ivanova, Yee Whye Teh, Wee Sun Lee
Abstract: Although large language models (LLMs) are highly interactive and extendable, current approaches to ensure reliability in deployments remain mostly limited to rejecting outputs with high uncertainty in order to avoid misinformation. This conservative strategy reflects the current lack of tools to systematically distinguish and respond to different sources of uncertainty. In this paper, we advocate for the adoption of Bayesian Modeling of Experiments -- a framework that provides a coherent foundation to reason about uncertainty and clarify the reducibility of uncertainty -- for managing and proactively addressing uncertainty that arises in LLM deployments. This framework enables LLMs and their users to take contextually appropriate steps, such as requesting clarification, retrieving external information, or refining inputs. By supporting active resolution rather than passive avoidance, it opens the door to more reliable, transparent, and broadly applicable LLM systems, particularly in high-stakes, real-world settings.
Authors: Yuxin Xiao, Sana Tonekaboni, Walter Gerych, Vinith Suriyakumar, Marzyeh Ghassemi
Abstract: Large language models (LLMs) can be prompted with specific styles (e.g., formatting responses as lists), including in jailbreak queries. Although these style patterns are semantically unrelated to the malicious intents behind jailbreak queries, their safety impact remains unclear. In this work, we seek to understand whether style patterns compromise LLM safety, how superficial style alignment increases model vulnerability, and how best to mitigate these risks during alignment. We evaluate 32 LLMs across seven jailbreak benchmarks, and find that malicious queries with style patterns inflate the attack success rate (ASR) for nearly all models. Notably, ASR inflation correlates with both the length of style patterns and the relative attention an LLM exhibits on them. We then investigate superficial style alignment, and find that fine-tuning with specific styles makes LLMs more vulnerable to jailbreaks of those same styles. Finally, we propose SafeStyle, a defense strategy that incorporates a small amount of safety training data augmented to match the distribution of style patterns in the fine-tuning data. Across three LLMs and five fine-tuning style settings, SafeStyle consistently outperforms baselines in maintaining LLM safety.
Authors: Ziwen Wang, Jiajun Fan, Ruihan Guo, Thao Nguyen, Heng Ji, Ge Liu
Abstract: Protein generative models have shown remarkable promise in protein design but still face limitations in success rate, due to the scarcity of high-quality protein datasets for supervised pretraining. We present ProteinZero, a novel framework that enables scalable, automated, and continuous self-improvement of the inverse folding model through online reinforcement learning. To achieve computationally tractable online feedback, we introduce efficient proxy reward models based on ESM-fold and a novel rapid ddG predictor that significantly accelerates evaluation speed. ProteinZero employs a general RL framework balancing multi-reward maximization, KL-divergence from a reference model, and a novel protein-embedding level diversity regularization that prevents mode collapse while promoting higher sequence diversity. Through extensive experiments, we demonstrate that ProteinZero substantially outperforms existing methods across every key metric in protein design, achieving significant improvements in structural accuracy, designability, thermodynamic stability, and sequence diversity. Most impressively, ProteinZero reduces design failure rates by approximately 36% - 48% compared to widely-used methods like ProteinMPNN, ESM-IF and InstructPLM, consistently achieving success rates exceeding 90% across diverse and complex protein folds. Notably, the entire RL run on CATH-4.3 can be done with a single 8 X GPU node in under 3 days, including reward computation. Our work establishes a new paradigm for protein design where models evolve continuously from their own generated outputs, opening new possibilities for exploring the vast protein design space.
Authors: Jie Peng, Hongwei Yang, Jing Zhao, Hengji Dong, Hui He, Weizhe Zhang, Haoyu He
Abstract: Deep neural networks are vulnerable to backdoor attacks, where malicious behaviors are implanted during training. While existing defenses can effectively purify compromised models, they typically require labeled data or specific training procedures, making them difficult to apply beyond supervised learning settings. Notably, recent studies have shown successful backdoor attacks across various learning paradigms, highlighting a critical security concern. To address this gap, we propose Two-stage Symmetry Connectivity (TSC), a novel backdoor purification defense that operates independently of data format and requires only a small fraction of clean samples. Through theoretical analysis, we prove that by leveraging permutation invariance in neural networks and quadratic mode connectivity, TSC amplifies the loss on poisoned samples while maintaining bounded clean accuracy. Experiments demonstrate that TSC achieves robust performance comparable to state-of-the-art methods in supervised learning scenarios. Furthermore, TSC generalizes to self-supervised learning frameworks, such as SimCLR and CLIP, maintaining its strong defense capabilities. Our code is available at https://github.com/JiePeng104/TSC.
Authors: Mickel Liu, Liwei Jiang, Yancheng Liang, Simon Shaolei Du, Yejin Choi, Tim Althoff, Natasha Jaques
Abstract: Conventional language model (LM) safety alignment relies on a reactive, disjoint procedure: attackers exploit a static model, followed by defensive fine-tuning to patch exposed vulnerabilities. This sequential approach creates a mismatch -- attackers overfit to obsolete defenses, while defenders perpetually lag behind emerging threats. To address this, we propose Self-RedTeam, an online self-play reinforcement learning algorithm where an attacker and defender agent co-evolve through continuous interaction. We cast safety alignment as a two-player zero-sum game, where a single model alternates between attacker and defender roles -- generating adversarial prompts and safeguarding against them -- while a reward LM adjudicates outcomes. This enables dynamic co-adaptation. Grounded in the game-theoretic framework of zero-sum games, we establish a theoretical safety guarantee which motivates the design of our method: if self-play converges to a Nash Equilibrium, the defender will reliably produce safe responses to any adversarial input. Empirically, Self-RedTeam uncovers more diverse attacks (+21.8% SBERT) compared to attackers trained against static defenders and achieves higher robustness on safety benchmarks (e.g., +65.5% on WildJailBreak) than defenders trained against static attackers. We further propose hidden Chain-of-Thought, allowing agents to plan privately, which boosts adversarial diversity and reduces over-refusals. Our results motivate a shift from reactive patching to proactive co-evolution in LM safety training, enabling scalable, autonomous, and robust self-improvement of LMs via multi-agent reinforcement learning (MARL).
Authors: Thomas Zhu, Joshua Clune, Jeremy Avigad, Albert Qiaochu Jiang, Sean Welleck
Abstract: Neural methods are transforming automated reasoning for proof assistants, yet integrating these advances into practical verification workflows remains challenging. Hammers are tools that interface with external automatic theorem provers to automate tedious reasoning steps. They have dramatically improved productivity in proof assistants, but the Lean proof assistant still does not have a hammer despite its growing popularity. We present LeanHammer, the first end-to-end domain-general hammer for Lean, built on a novel neural premise selection system for a hammer in dependent type theory. Unlike existing Lean premise selectors, our approach dynamically adapts to user-specific contexts and combines with symbolic proof search and reconstruction to create a practical hammer. With comprehensive evaluations, we show that our premise selector enables LeanHammer to solve 21\% more goals relative to existing premise selectors, and generalize well to diverse domains. Our work bridges the gap between neural retrieval and symbolic reasoning, making formal verification more accessible to researchers and practitioners.
Authors: Xiangkun Hu, Lemin Kong, Tong He, David Wipf
Abstract: The generated responses of large language models (LLMs) are often fine-tuned to human preferences through a process called reinforcement learning from human feedback (RLHF). As RLHF relies on a challenging training sequence, whereby a separate reward model is independently learned and then later applied to LLM policy updates, ongoing research effort has targeted more straightforward alternatives. In this regard, direct preference optimization (DPO) and its many offshoots circumvent the need for a separate reward training step. Instead, through the judicious use of a reparameterization trick that induces an \textit{implicit} reward, DPO and related methods consolidate learning to the minimization of a single loss function. And yet despite demonstrable success in some real-world settings, we prove that DPO-based objectives are nonetheless subject to sub-optimal regularization and counter-intuitive interpolation behaviors, underappreciated artifacts of the reparameterizations upon which they are based. To this end, we introduce an \textit{explicit} preference optimization framework termed EXPO that requires no analogous reparameterization to achieve an implicit reward. Quite differently, we merely posit intuitively-appealing regularization factors from scratch that transparently avoid the potential pitfalls of key DPO variants, provably satisfying regularization desiderata that prior methods do not. Empirical results serve to corroborate our analyses and showcase the efficacy of EXPO.
Authors: Shakir Yousefi, Andreas Plesner, Till Aczel, Roger Wattenhofer
Abstract: Modern neural networks demonstrate state-of-the-art performance on numerous existing benchmarks; however, their high computational requirements and energy consumption prompt researchers to seek more efficient solutions for real-world deployment. Logic gate networks (LGNs) learns a large network of logic gates for efficient image classification. However, learning a network that can solve a simple problem like CIFAR-10 can take days to weeks to train. Even then, almost half of the network remains unused, causing a discretization gap. This discretization gap hinders real-world deployment of LGNs, as the performance drop between training and inference negatively impacts accuracy. We inject Gumbel noise with a straight-through estimator during training to significantly speed up training, improve neuron utilization, and decrease the discretization gap. We theoretically show that this results from implicit Hessian regularization, which improves the convergence properties of LGNs. We train networks $4.5 \times$ faster in wall-clock time, reduce the discretization gap by $98\%$, and reduce the number of unused gates by $100\%$.
Authors: Libo Wang
Abstract: In view of the problem that each subchain in the chain-of-model (CoM) relies only on the information of the previous subchain and may lose long-range dependencies due to the causal mask blocking the global context flow between multi-level subchains, this work proposes a graph of causal evolution (GoCE). Its core principle is to map the implicit token representation into a differentiable and sparse causal adjacency matrix, then permeate causal constraints through each layer of calculation using causal-masked attention and causal-MoE. By combining intervention consistency loss test and self-evolution gate, the dynamic balance between causal structure learning and adaptive updating of transformer architecture is realized. The researcher built experimental environments in sandboxes built with Claude Sonnet 4, o4-mini-high, and DeepSeek R1 respectively with the transformer variant architecture introduced in GoCE. It is evaluated on publicly available datasets including CLUTRR, CLADDER, EX-FEVER, and CausalQA and compared with the baseline LLMs. The finding proves that GoCE strengthens the transformer's ability to capture long-range causal dependencies, while the ability to self-evolve is improved. It not only surpasses the design of CoM in terms of design principles, but also provides experience for future research on causal learning and continuous adaptive improvement.
Authors: Perry Dong, Alec M. Lessing, Annie S. Chen, Chelsea Finn
Abstract: We study the problem of sample efficient reinforcement learning, where prior data such as demonstrations are provided for initialization in lieu of a dense reward signal. A natural approach is to incorporate an imitation learning objective, either as regularization during training or to acquire a reference policy. However, imitation learning objectives can ultimately degrade long-term performance, as it does not directly align with reward maximization. In this work, we propose to use prior data solely for guiding exploration via noise added to the policy, sidestepping the need for explicit behavior cloning constraints. The key insight in our framework, Data-Guided Noise (DGN), is that demonstrations are most useful for identifying which actions should be explored, rather than forcing the policy to take certain actions. Our approach achieves up to 2-3x improvement over prior reinforcement learning from offline data methods across seven simulated continuous control tasks.
Authors: Shuqiang Zhang, Yuchao Zhang, Jinkun Chen, Haochen Sui
Abstract: Recommendation systems (RS) aim to provide personalized content, but they face a challenge in unbiased learning due to selection bias, where users only interact with items they prefer. This bias leads to a distorted representation of user preferences, which hinders the accuracy and fairness of recommendations. To address the issue, various methods such as error imputation based, inverse propensity scoring, and doubly robust techniques have been developed. Despite the progress, from the structural causal model perspective, previous debiasing methods in RS assume the independence of the exogenous variables. In this paper, we release this assumption and propose a learning algorithm based on likelihood maximization to learn a prediction model. We first discuss the correlation and difference between unmeasured confounding and our scenario, then we propose a unified method that effectively handles latent exogenous variables. Specifically, our method models the data generation process with latent exogenous variables under mild normality assumptions. We then develop a Monte Carlo algorithm to numerically estimate the likelihood function. Extensive experiments on synthetic datasets and three real-world datasets demonstrate the effectiveness of our proposed method. The code is at https://github.com/WallaceSUI/kdd25-background-variable.
URLs: https://github.com/WallaceSUI/kdd25-background-variable.
Authors: Cl\'ement Bonet, Christophe Vauthier, Anna Korba
Abstract: Many applications in machine learning involve data represented as probability distributions. The emergence of such data requires radically novel techniques to design tractable gradient flows on probability distributions over this type of (infinite-dimensional) objects. For instance, being able to flow labeled datasets is a core task for applications ranging from domain adaptation to transfer learning or dataset distillation. In this setting, we propose to represent each class by the associated conditional distribution of features, and to model the dataset as a mixture distribution supported on these classes (which are themselves probability distributions), meaning that labeled datasets can be seen as probability distributions over probability distributions. We endow this space with a metric structure from optimal transport, namely the Wasserstein over Wasserstein (WoW) distance, derive a differential structure on this space, and define WoW gradient flows. The latter enables to design dynamics over this space that decrease a given objective functional. We apply our framework to transfer learning and dataset distillation tasks, leveraging our gradient flow construction as well as novel tractable functionals that take the form of Maximum Mean Discrepancies with Sliced-Wasserstein based kernels between probability distributions.
Authors: Zhangchi Zhao, Jun Shu, Deyu Meng, Zongben Xu
Abstract: Inspired by the Kolmogorov-Arnold representation theorem, KANs offer a novel framework for function approximation by replacing traditional neural network weights with learnable univariate functions. This design demonstrates significant potential as an efficient and interpretable alternative to traditional MLPs. However, KANs are characterized by a substantially larger number of trainable parameters, leading to challenges in memory efficiency and higher training costs compared to MLPs. To address this limitation, we propose to generate weights for KANs via a smaller meta-learner, called MetaKANs. By training KANs and MetaKANs in an end-to-end differentiable manner, MetaKANs achieve comparable or even superior performance while significantly reducing the number of trainable parameters and maintaining promising interpretability. Extensive experiments on diverse benchmark tasks, including symbolic regression, partial differential equation solving, and image classification, demonstrate the effectiveness of MetaKANs in improving parameter efficiency and memory usage. The proposed method provides an alternative technique for training KANs, that allows for greater scalability and extensibility, and narrows the training cost gap with MLPs stated in the original paper of KANs. Our code is available at https://github.com/Murphyzc/MetaKAN.
Authors: Mengsong Wu, YaFei Wang, Yidong Ming, Yuqi An, Yuwei Wan, Wenliang Chen, Binbin Lin, Yuqiang Li, Tong Xie, Dongzhan Zhou
Abstract: Large language models (LLMs) have recently demonstrated promising capabilities in chemistry tasks while still facing challenges due to outdated pretraining knowledge and the difficulty of incorporating specialized chemical expertise. To address these issues, we propose an LLM-based agent that synergistically integrates 137 external chemical tools created ranging from basic information retrieval to complex reaction predictions, and a dataset curation pipeline to generate the dataset ChemToolBench that facilitates both effective tool selection and precise parameter filling during fine-tuning and evaluation. We introduce a Hierarchical Evolutionary Monte Carlo Tree Search (HE-MCTS) framework, enabling independent optimization of tool planning and execution. By leveraging self-generated data, our approach supports step-level fine-tuning (FT) of the policy model and training task-adaptive PRM and ORM that surpass GPT-4o. Experimental evaluations demonstrate that our approach significantly improves performance in Chemistry QA and discovery tasks, offering a robust solution to integrate specialized tools with LLMs for advanced chemical applications. All datasets and code are available at https://github.com/AI4Chem/ChemistryAgent .
Authors: Florian Andreas Marwitz, Ralf M\"oller, Magnus Bender, Marcel Gehrke
Abstract: Inference in dynamic probabilistic models is a complex task involving expensive operations. In particular, for Hidden Markov Models, the whole state space has to be enumerated for advancing in time. Even states with negligible probabilities are considered, resulting in computational inefficiency and increased noise due to the propagation of unlikely probability mass. We propose to denoise the future and speed up inference by using only the top-p states, i.e., the most probable states with accumulated probability p. We show that the error introduced by using only the top-p states is bound by p and the so-called minimal mixing rate of the underlying model. Moreover, in our empirical evaluation, we show that we can expect speedups of at least an order of magnitude, while the error in terms of total variation distance is below 0.09.
Authors: Tan Chen, Jintao Yan, Yuxuan Sun, Sheng Zhou, Zhisheng Niu
Abstract: Federated learning (FL) is a promising paradigm for multiple devices to cooperatively train a model. When applied in wireless networks, two issues consistently affect the performance of FL, i.e., data heterogeneity of devices and limited bandwidth. Many papers have investigated device scheduling strategies considering the two issues. However, most of them recognize data heterogeneity as a property of individual devices. In this paper, we prove that the convergence speed of FL is affected by the sum of device-level and sample-level collective gradient divergence (CGD). The device-level CGD refers to the gradient divergence of the scheduled device group, instead of the sum of the individual device divergence. The sample-level CGD is statistically upper bounded by sampling variance, which is inversely proportional to the total number of samples scheduled for local update. To derive a tractable form of the device-level CGD, we further consider a classification problem and transform it into the weighted earth moving distance (WEMD) between the group distribution and the global distribution. Then we propose FedCGD algorithm to minimize the sum of multi-level CGDs by balancing WEMD and sampling variance, within polynomial time. Simulation shows that the proposed strategy increases classification accuracy on the CIFAR-10 dataset by up to 4.2\% while scheduling 41.8\% fewer devices, and flexibly switches between reducing WEMD and reducing sampling variance.
Authors: Hao Li, Bowen Deng, Chang Xu, Zhiyuan Feng, Viktor Schlegel, Yu-Hao Huang, Yizheng Sun, Jingyuan Sun, Kailai Yang, Yiyao Yu, Jiang Bian
Abstract: A unified foundation model for medical time series -- pretrained on open access and ethics board-approved medical corpora -- offers the potential to reduce annotation burdens, minimize model customization, and enable robust transfer across clinical institutions, modalities, and tasks, particularly in data-scarce or privacy-constrained environments. However, existing generalist time series foundation models struggle to handle medical time series data due to their inherent challenges, including irregular intervals, heterogeneous sampling rates, and frequent missing values. To address these challenges, we introduce MIRA, a unified foundation model specifically designed for medical time series forecasting. MIRA incorporates a Continuous-Time Rotary Positional Encoding that enables fine-grained modeling of variable time intervals, a frequency-specific mixture-of-experts layer that routes computation across latent frequency regimes to further promote temporal specialization, and a Continuous Dynamics Extrapolation Block based on Neural ODE that models the continuous trajectory of latent states, enabling accurate forecasting at arbitrary target timestamps. Pretrained on a large-scale and diverse medical corpus comprising over 454 billion time points collect from publicly available datasets, MIRA achieves reductions in forecasting errors by an average of 10% and 7% in out-of-distribution and in-distribution scenarios, respectively, when compared to other zero-shot and fine-tuned baselines. We also introduce a comprehensive benchmark spanning multiple downstream clinical tasks, establishing a foundation for future research in medical time series modeling.
Authors: Seokbin Yoon, Keumjin Lee
Abstract: Aircraft trajectory modeling plays a crucial role in Air Traffic Management (ATM) and is important for various downstream tasks, including conflict detection and landing time prediction. Dataset augmentation through the addition of synthetically generated trajectory data is necessary to develop a more robust aircraft trajectory model and ensure that the trajectory dataset is sufficient and balanced. In this work, we propose a novel framework called ATRADA for aircraft trajectory dataset augmentation. In the proposed framework, a Transformer encoder learns the underlying patterns in the original trajectory dataset and converts each data point into a context vector in the learned latent space. The converted dataset in the latent space is projected into reduced dimensions using principal component analysis (PCA), and a Gaussian mixture model (GMM) is applied to fit the probability distribution of the data points in the reduced-dimensional space. Finally, new samples are drawn from the fitted GMM, the dimension of the samples is reverted to the original dimension, and they are decoded with a Multi-Layer Perceptron (MLP). Several experiments demonstrate that the framework effectively generates new, high-quality synthetic aircraft trajectory data, which were compared to the results of several baselines.
Authors: Tongzhou Yu, Zhuhao Zhang, Guanghui Zhu, Shen Jiang, Meikang Qiu, Yihua Huang
Abstract: Parameter Efficient Fine-Tuning (PEFT) methods have emerged as effective and promising approaches for fine-tuning pre-trained language models. Compared with Full parameter Fine-Tuning (FFT), PEFT achieved comparable task performance with a substantial reduction of trainable parameters, which largely saved the training and storage costs. However, using the PEFT method requires considering a vast design space, such as the type of PEFT modules and their insertion layers. Inadequate configurations can lead to sub-optimal results. Conventional solutions such as architectural search techniques, while effective, tend to introduce substantial additional overhead. In this paper, we propose a novel approach, PrunePEFT, which formulates the PEFT strategy search as a pruning problem and introduces a hybrid pruning strategy that capitalizes on the sensitivity of pruning methods to different PEFT modules. This method extends traditional pruning techniques by iteratively removing redundant or conflicting PEFT modules, thereby optimizing the fine-tuned configuration. By efficiently identifying the most relevant modules, our approach significantly reduces the computational burden typically associated with architectural search processes, making it a more scalable and efficient solution for fine-tuning large pre-trained models.
Authors: Hao Qiu, Emmanuel Esposito, Mengxiao Zhang
Abstract: In this work, we study the online convex optimization problem with curved losses and delayed feedback. When losses are strongly convex, existing approaches obtain regret bounds of order $d_{\max} \ln T$, where $d_{\max}$ is the maximum delay and $T$ is the time horizon. However, in many cases, this guarantee can be much worse than $\sqrt{d_{\mathrm{tot}}}$ as obtained by a delayed version of online gradient descent, where $d_{\mathrm{tot}}$ is the total delay. We bridge this gap by proposing a variant of follow-the-regularized-leader that obtains regret of order $\min\{\sigma_{\max}\ln T, \sqrt{d_{\mathrm{tot}}}\}$, where $\sigma_{\max}$ is the maximum number of missing observations. We then consider exp-concave losses and extend the Online Newton Step algorithm to handle delays with an adaptive learning rate tuning, achieving regret $\min\{d_{\max} n\ln T, \sqrt{d_{\mathrm{tot}}}\}$ where $n$ is the dimension. To our knowledge, this is the first algorithm to achieve such a regret bound for exp-concave losses. We further consider the problem of unconstrained online linear regression and achieve a similar guarantee by designing a variant of the Vovk-Azoury-Warmuth forecaster with a clipping trick. Finally, we implement our algorithms and conduct experiments under various types of delay and losses, showing an improved performance over existing methods.
Authors: Torsten Krau{\ss}, Hamid Dashtbani, Alexandra Dmitrienko
Abstract: Machine learning is advancing rapidly, with applications bringing notable benefits, such as improvements in translation and code generation. Models like ChatGPT, powered by Large Language Models (LLMs), are increasingly integrated into daily life. However, alongside these benefits, LLMs also introduce social risks. Malicious users can exploit LLMs by submitting harmful prompts, such as requesting instructions for illegal activities. To mitigate this, models often include a security mechanism that automatically rejects such harmful prompts. However, they can be bypassed through LLM jailbreaks. Current jailbreaks often require significant manual effort, high computational costs, or result in excessive model modifications that may degrade regular utility. We introduce TwinBreak, an innovative safety alignment removal method. Building on the idea that the safety mechanism operates like an embedded backdoor, TwinBreak identifies and prunes parameters responsible for this functionality. By focusing on the most relevant model layers, TwinBreak performs fine-grained analysis of parameters essential to model utility and safety. TwinBreak is the first method to analyze intermediate outputs from prompts with high structural and content similarity to isolate safety parameters. We present the TwinPrompt dataset containing 100 such twin prompts. Experiments confirm TwinBreak's effectiveness, achieving 89% to 98% success rates with minimal computational requirements across 16 LLMs from five vendors.
Authors: Zhixin Geng, Xu Fan, Xiqiao Lu, Yan Zhang, Guangyuan Yu, Cheng Huang, Qian Wang, Yuewu Li, Weichun Ma, Qi Yu, Libo Wu, Hao Li
Abstract: Air pollution has emerged as a major public health challenge in megacities. Numerical simulations and single-site machine learning approaches have been widely applied in air quality forecasting tasks. However, these methods face multiple limitations, including high computational costs, low operational efficiency, and limited integration with observational data. With the rapid advancement of artificial intelligence, there is an urgent need to develop a low-cost, efficient air quality forecasting model for smart urban management. An air quality forecasting model, named FuXi-Air, has been constructed in this study based on multimodal data fusion to support high-precision air quality forecasting and operated in typical megacities. The model integrates meteorological forecasts, emission inventories, and pollutant monitoring data under the guidance of air pollution mechanism. By combining an autoregressive prediction framework with a frame interpolation strategy, the model successfully completes 72-hour forecasts for six major air pollutants at an hourly resolution across multiple monitoring sites within 25-30 seconds. In terms of both computational efficiency and forecasting accuracy, it outperforms the mainstream numerical air quality models in operational forecasting work. Ablation experiments concerning key influencing factors show that although meteorological data contribute more to model accuracy than emission inventories do, the integration of multimodal data significantly improves forecasting precision and ensures that reliable predictions are obtained under differing pollution mechanisms across megacities. This study provides both a technical reference and a practical example for applying multimodal data-driven models to air quality forecasting and offers new insights into building hybrid forecasting systems to support air pollution risk warning in smart city management.
Authors: Toby Boyne, Juan S. Campos, Becky D. Langdon, Jixiang Qing, Yilin Xie, Shiqiang Zhang, Calvin Tsay, Ruth Misener, Daniel W. Davies, Kim E. Jelfs, Sarah Boyall, Thomas M. Dixon, Linden Schrecker, Jose Pablo Folch
Abstract: Machine learning has promised to change the landscape of laboratory chemistry, with impressive results in molecular property prediction and reaction retro-synthesis. However, chemical datasets are often inaccessible to the machine learning community as they tend to require cleaning, thorough understanding of the chemistry, or are simply not available. In this paper, we introduce a novel dataset for yield prediction, providing the first-ever transient flow dataset for machine learning benchmarking, covering over 1200 process conditions. While previous datasets focus on discrete parameters, our experimental set-up allow us to sample a large number of continuous process conditions, generating new challenges for machine learning models. We focus on solvent selection, a task that is particularly difficult to model theoretically and therefore ripe for machine learning applications. We showcase benchmarking for regression algorithms, transfer-learning approaches, feature engineering, and active learning, with important applications towards solvent replacement and sustainable manufacturing.
Authors: Ali Hariri, \'Alvaro Arroyo, Alessio Gravina, Moshe Eliasof, Carola-Bibiane Sch\"onlieb, Davide Bacciu, Kamyar Azizzadenesheli, Xiaowen Dong, Pierre Vandergheynst
Abstract: ChebNet, one of the earliest spectral GNNs, has largely been overshadowed by Message Passing Neural Networks (MPNNs), which gained popularity for their simplicity and effectiveness in capturing local graph structure. Despite their success, MPNNs are limited in their ability to capture long-range dependencies between nodes. This has led researchers to adapt MPNNs through rewiring or make use of Graph Transformers, which compromises the computational efficiency that characterized early spatial message-passing architectures, and typically disregards the graph structure. Almost a decade after its original introduction, we revisit ChebNet to shed light on its ability to model distant node interactions. We find that out-of-box, ChebNet already shows competitive advantages relative to classical MPNNs and GTs on long-range benchmarks, while maintaining good scalability properties for high-order polynomials. However, we uncover that this polynomial expansion leads ChebNet to an unstable regime during training. To address this limitation, we cast ChebNet as a stable and non-dissipative dynamical system, which we coin Stable-ChebNet. Our Stable-ChebNet model allows for stable information propagation, and has controllable dynamics which do not require the use of eigendecompositions, positional encodings, or graph rewiring. Across several benchmarks, Stable-ChebNet achieves near state-of-the-art performance.
Authors: Meir Feder, Ruediger Urbanke, Yaniv Fogel
Abstract: A fundamental question in modern machine learning is why large, over-parameterized models, such as deep neural networks and transformers, tend to generalize well, even when their number of parameters far exceeds the number of training samples. We investigate this phenomenon through the lens of information theory, grounded in universal learning theory. Specifically, we study a Bayesian mixture learner with log-loss and (almost) uniform prior over an expansive hypothesis class. Our key result shows that the learner's regret is not determined by the overall size of the hypothesis class, but rather by the cumulative probability of all models that are close, in Kullback-Leibler divergence distance, to the true data-generating process. We refer to this cumulative probability as the weight of the hypothesis. This leads to a natural notion of model simplicity: simple models are those with large weight and thus require fewer samples to generalize, while complex models have small weight and need more data. This perspective provides a rigorous and intuitive explanation for why over-parameterized models often avoid overfitting: the presence of simple hypotheses allows the posterior to concentrate on them when supported by the data. We further bridge theory and practice by recalling that stochastic gradient descent with Langevin dynamics samples from the correct posterior distribution, enabling our theoretical learner to be approximated using standard machine learning methods combined with ensemble learning. Our analysis yields non-uniform regret bounds and aligns with key practical concepts such as flat minima and model distillation. The results apply broadly across online, batch, and supervised learning settings, offering a unified and principled understanding of the generalization behavior of modern AI systems.
Authors: Seyedhamidreza Mousavi, Seyedali Mousavi, Masoud Daneshtalab
Abstract: Adversarial Robustness Distillation (ARD) has emerged as an effective method to enhance the robustness of lightweight deep neural networks against adversarial attacks. Current ARD approaches have leveraged a large robust teacher network to train one robust lightweight student. However, due to the diverse range of edge devices and resource constraints, current approaches require training a new student network from scratch to meet specific constraints, leading to substantial computational costs and increased CO2 emissions. This paper proposes Progressive Adversarial Robustness Distillation (ProARD), enabling the efficient one-time training of a dynamic network that supports a diverse range of accurate and robust student networks without requiring retraining. We first make a dynamic deep neural network based on dynamic layers by encompassing variations in width, depth, and expansion in each design stage to support a wide range of architectures. Then, we consider the student network with the largest size as the dynamic teacher network. ProARD trains this dynamic network using a weight-sharing mechanism to jointly optimize the dynamic teacher network and its internal student networks. However, due to the high computational cost of calculating exact gradients for all the students within the dynamic network, a sampling mechanism is required to select a subset of students. We show that random student sampling in each iteration fails to produce accurate and robust students.
Authors: Guanhua Zhang, Florian E. Dorner, Moritz Hardt
Abstract: Large language model (LLM) evaluation is increasingly costly, prompting interest in methods that speed up evaluation by shrinking benchmark datasets. Benchmark prediction (also called efficient LLM evaluation) aims to select a small subset of evaluation points and predict overall benchmark performance from that subset. In this paper, we systematically assess the strengths and limitations of 11 benchmark prediction methods across 19 diverse benchmarks. First, we identify a highly competitive baseline: Take a random sample and fit a regression model on the sample to predict missing entries. Outperforming most existing methods, this baseline challenges the assumption that careful subset selection is necessary for benchmark prediction. Second, we discover that all existing methods crucially depend on model similarity. They work best when interpolating scores among similar models. The effectiveness of benchmark prediction sharply declines when new models have higher accuracy than previously seen models. In this setting of extrapolation, none of the previous methods consistently beat a simple average over random samples. To improve over the sample average, we introduce a new method inspired by augmented inverse propensity weighting. This method consistently outperforms the random sample average even for extrapolation. However, its performance still relies on model similarity and the gains are modest in general. This shows that benchmark prediction fails just when it is most needed: at the evaluation frontier, where the goal is to evaluate new models of unknown capabilities.
Authors: Boris Martirosyan, Alexey Karmanov
Abstract: Latent diffusion models (LDMs) achieve state-of-the-art performance across various tasks, including image generation and video synthesis. However, they generally lack robustness, a limitation that remains not fully explored in current research. In this paper, we propose several methods to address this gap. First, we hypothesize that the robustness of LDMs primarily should be measured without their text encoder, because if we take and explore the whole architecture, the problems of image generator and text encoders wll be fused. Second, we introduce novel data augmentation techniques designed to reveal robustness shortcomings in LDMs when processing diverse textual prompts. We then fine-tune Stable Diffusion 3 and Stable Diffusion XL models using Dreambooth, incorporating these proposed augmentation methods across multiple tasks. Finally, we propose a novel evaluation pipeline specifically tailored to assess the robustness of LDMs fine-tuned via Dreambooth.
Authors: Haizhao Jing, Haokui Zhang, Zhenhao Shang, Rong Xiao, Peng Wang, Yanning Zhang
Abstract: Neural Architecture Representation Learning aims to transform network models into feature representations for predicting network attributes, playing a crucial role in deploying and designing networks for real-world applications. Recently, inspired by the success of transformers, transformer-based models integrated with Graph Neural Networks (GNNs) have achieved significant progress in representation learning. However, current methods still have some limitations. First, existing methods overlook hardware attribute information, which conflicts with the current trend of diversified deep learning hardware and limits the practical applicability of models. Second, current encoding approaches rely on static adjacency matrices to represent topological structures, failing to capture the structural differences between computational nodes, which ultimately compromises encoding effectiveness. In this paper, we introduce LeDG-Former, an innovative framework that addresses these limitations through the synergistic integration of language-based semantic embedding and dynamic graph representation learning. Specifically, inspired by large language models (LLMs), we propose a language embedding framework where both neural architectures and hardware platform specifications are projected into a unified semantic space through tokenization and LLM processing, enabling zero-shot prediction across different hardware platforms for the first time. Then, we propose a dynamic graph-based transformer for modeling neural architectures, resulting in improved neural architecture modeling performance. On the NNLQP benchmark, LeDG-Former surpasses previous methods, establishing a new SOTA while demonstrating the first successful cross-hardware latency prediction capability. Furthermore, our framework achieves superior performance on the cell-structured NAS-Bench-101 and NAS-Bench-201 datasets.
Authors: Seungho Baek, Taegeon Park, Jongchan Park, Seungjun Oh, Yusung Kim
Abstract: Existing offline hierarchical reinforcement learning methods rely on high-level policy learning to generate subgoal sequences. However, their efficiency degrades as task horizons increase, and they lack effective strategies for stitching useful state transitions across different trajectories. We propose Graph-Assisted Stitching (GAS), a novel framework that formulates subgoal selection as a graph search problem rather than learning an explicit high-level policy. By embedding states into a Temporal Distance Representation (TDR) space, GAS clusters semantically similar states from different trajectories into unified graph nodes, enabling efficient transition stitching. A shortest-path algorithm is then applied to select subgoal sequences within the graph, while a low-level policy learns to reach the subgoals. To improve graph quality, we introduce the Temporal Efficiency (TE) metric, which filters out noisy or inefficient transition states, significantly enhancing task performance. GAS outperforms prior offline HRL methods across locomotion, navigation, and manipulation tasks. Notably, in the most stitching-critical task, it achieves a score of 88.3, dramatically surpassing the previous state-of-the-art score of 1.0. Our source code is available at: https://github.com/qortmdgh4141/GAS.
Authors: Adam Breuer
Abstract: In this paper, we provide the first practical algorithms with provable guarantees for the problem of inferring the topics assigned to each document in an LDA topic model. This is the primary inference problem for many applications of topic models in social science, data exploration, and causal inference settings. We obtain this result by showing a novel non-gradient-based, combinatorial approach to estimating topic models. This yields algorithms that converge to near-optimal posterior probability in logarithmic parallel computation time (adaptivity) -- exponentially faster than any known LDA algorithm. We also show that our approach can provide interpretability guarantees such that each learned topic is formally associated with a known keyword. Finally, we show that unlike alternatives, our approach can maintain the independence assumptions necessary to use the learned topic model for downstream causal inference methods that allow researchers to study topics as treatments. In terms of practical performance, our approach consistently returns solutions of higher semantic quality than solutions from state-of-the-art LDA algorithms, neural topic models, and LLM-based topic models across a diverse range of text datasets and evaluation parameters.
Authors: Nicola Lavecchia, Sid Fadanelli, Federico Ricciuti, Gennaro Aloe, Enrico Bagli, Pietro Giuffrida, Daniele Vergari
Abstract: Generative AI technologies have demonstrated significant potential across diverse applications. This study provides a comparative analysis of credit score modeling techniques, contrasting traditional approaches with those leveraging generative AI. Our findings reveal that current generative AI models fall short of matching the performance of traditional methods, regardless of the integration strategy employed. These results highlight the limitations in the current capabilities of generative AI for credit risk scoring, emphasizing the need for further research and development before the possibility of applying generative AI for this specific task, or equivalent ones.
Authors: Dekai Zhang, Matthew Williams, Francesca Toni
Abstract: Federated learning (FL) is a widely used framework for machine learning in distributed data environments where clients hold data that cannot be easily centralised, such as for data protection reasons. FL, however, is known to be vulnerable to non-IID data. Clustered FL addresses this issue by finding more homogeneous clusters of clients. We propose a novel one-shot clustering method, EMD-CFL, using the Earth Mover's distance (EMD) between data distributions in embedding space. We theoretically motivate the use of EMDs using results from the domain adaptation literature and demonstrate empirically superior clustering performance in extensive comparisons against 16 baselines and on a range of challenging datasets.
Authors: Jie Bao, Chuangyin Dang, Rui Luo, Hanwei Zhang, Zhixin Zhou
Abstract: As deep learning models are increasingly deployed in high-risk applications, robust defenses against adversarial attacks and reliable performance guarantees become paramount. Moreover, accuracy alone does not provide sufficient assurance or reliable uncertainty estimates for these models. This study advances adversarial training by leveraging principles from Conformal Prediction. Specifically, we develop an adversarial attack method, termed OPSA (OPtimal Size Attack), designed to reduce the efficiency of conformal prediction at any significance level by maximizing model uncertainty without requiring coverage guarantees. Correspondingly, we introduce OPSA-AT (Adversarial Training), a defense strategy that integrates OPSA within a novel conformal training paradigm. Experimental evaluations demonstrate that our OPSA attack method induces greater uncertainty compared to baseline approaches for various defenses. Conversely, our OPSA-AT defensive model significantly enhances robustness not only against OPSA but also other adversarial attacks, and maintains reliable prediction. Our findings highlight the effectiveness of this integrated approach for developing trustworthy and resilient deep learning models for safety-critical domains. Our code is available at https://github.com/bjbbbb/Enhancing-Adversarial-Robustness-with-Conformal-Prediction.
URLs: https://github.com/bjbbbb/Enhancing-Adversarial-Robustness-with-Conformal-Prediction.
Authors: Avinash Kori, Francesca Toni, Ben Glocker
Abstract: Modular object-centric representations are essential for *human-like reasoning* but are challenging to obtain under spatial ambiguities, *e.g. due to occlusions and view ambiguities*. However, addressing challenges presents both theoretical and practical difficulties. We introduce a novel multi-view probabilistic approach that aggregates view-specific slots to capture *invariant content* information while simultaneously learning disentangled global *viewpoint-level* information. Unlike prior single-view methods, our approach resolves spatial ambiguities, provides theoretical guarantees for identifiability, and requires *no viewpoint annotations*. Extensive experiments on standard benchmarks and novel complex datasets validate our method's robustness and scalability.
Authors: Xintong Duan, Yutong He, Fahim Tajwar, Ruslan Salakhutdinov, J. Zico Kolter, Jeff Schneider
Abstract: Although diffusion models have achieved strong results in decision-making tasks, their slow inference speed remains a key limitation. While the consistency model offers a potential solution, its applications to decision-making often struggle with suboptimal demonstrations or rely on complex concurrent training of multiple networks. In this work, we propose a novel approach to consistency distillation for offline reinforcement learning that directly incorporates reward optimization into the distillation process. Our method enables single-step generation while maintaining higher performance and simpler training. Empirical evaluations on the Gym MuJoCo benchmarks and long horizon planning demonstrate that our approach can achieve an 8.7% improvement over previous state-of-the-art while offering up to 142x speedup over diffusion counterparts in inference time.
Authors: Jan Corazza, Hadi Partovi Aria, Hyohun Kim, Daniel Neider, Zhe Xu
Abstract: Reinforcement learning (RL) algorithms can find an optimal policy for a single agent to accomplish a particular task. However, many real-world problems require multiple agents to collaborate in order to achieve a common goal. For example, a robot executing a task in a warehouse may require the assistance of a drone to retrieve items from high shelves. In Decentralized Multi-Agent RL (DMARL), agents learn independently and then combine their policies at execution time, but often must satisfy constraints on compatibility of local policies to ensure that they can achieve the global task when combined. In this paper, we study how providing high-level symbolic knowledge to agents can help address unique challenges of this setting, such as privacy constraints, communication limitations, and performance concerns. In particular, we extend the formal tools used to check the compatibility of local policies with the team task, making decentralized training with theoretical guarantees usable in more scenarios. Furthermore, we empirically demonstrate that symbolic knowledge about the temporal evolution of events in the environment can significantly expedite the learning process in DMARL.
Authors: Michael K. Chen, Xikun Zhang, Jiaxing Huang, Dacheng Tao
Abstract: Large language models (LLMs) have become the cornerstone of modern AI. However, the existing paradigm of next-token prediction fundamentally limits their ability to form coherent, high-level concepts, making it a critical barrier to human-like understanding and reasoning. Take the phrase "ribonucleic acid" as an example: an LLM will first decompose it into tokens, i.e., artificial text fragments ("rib", "on", ...), then learn each token sequentially, rather than grasping the phrase as a unified, coherent semantic entity. This fragmented representation hinders deeper conceptual understanding and, ultimately, the development of truly intelligent systems. In response, we introduce Concept-Aware Fine-Tuning (CAFT), a novel multi-token training method that redefines how LLMs are fine-tuned. By enabling the learning of sequences that span multiple tokens, this method fosters stronger concept-aware learning. Our experiments demonstrate significant improvements compared to conventional next-token finetuning methods across diverse tasks, including traditional applications like text summarization and domain-specific ones like de novo protein design. Multi-token prediction was previously only possible in the prohibitively expensive pretraining phase; CAFT, to our knowledge, is the first to bring the multi-token setting to the post-training phase, thus effectively democratizing its benefits for the broader community of practitioners and researchers. Finally, the unexpected effectiveness of our proposed method suggests wider implications for the machine learning research community. All code and data are available at https://github.com/michaelchen-lab/caft-llm
Authors: Davide Carbone
Abstract: Energy-Based Models (EBMs) provide a flexible framework for generative modeling, but their training remains theoretically challenging due to the need to approximate normalization constants and efficiently sample from complex, multi-modal distributions. Traditional methods, such as contrastive divergence and score matching, introduce biases that can hinder accurate learning. In this work, we present a theoretical analysis of Jarzynski reweighting, a technique from non-equilibrium statistical mechanics, and its implications for training EBMs. We focus on the role of the choice of the kernel and we illustrate these theoretical considerations in two key generative frameworks: (i) flow-based diffusion models, where we reinterpret Jarzynski reweighting in the context of stochastic interpolants to mitigate discretization errors and improve sample quality, and (ii) Restricted Boltzmann Machines, where we analyze its role in correcting the biases of contrastive divergence. Our results provide insights into the interplay between kernel choice and model performance, highlighting the potential of Jarzynski reweighting as a principled tool for generative learning.
Authors: Zheng Zhang, Jie Bao, Zhixin Zhou, Nicolo Colombo, Lixin Cheng, Rui Luo
Abstract: Graph Neural Networks (GNNs) excel at modeling relational data but face significant challenges in high-stakes domains due to unquantified uncertainty. Conformal prediction (CP) offers statistical coverage guarantees, but existing methods often produce overly conservative prediction intervals that fail to account for graph heteroscedasticity and structural biases. While residual reweighting CP variants address some of these limitations, they neglect graph topology, cluster-specific uncertainties, and risk data leakage by reusing training sets. To address these issues, we propose Residual Reweighted GNN (RR-GNN), a framework designed to generate minimal prediction sets with provable marginal coverage guarantees. RR-GNN introduces three major innovations to enhance prediction performance. First, it employs Graph-Structured Mondrian CP to partition nodes or edges into communities based on topological features, ensuring cluster-conditional coverage that reflects heterogeneity. Second, it uses Residual-Adaptive Nonconformity Scores by training a secondary GNN on a held-out calibration set to estimate task-specific residuals, dynamically adjusting prediction intervals according to node or edge uncertainty. Third, it adopts a Cross-Training Protocol, which alternates the optimization of the primary GNN and the residual predictor to prevent information leakage while maintaining graph dependencies. We validate RR-GNN on 15 real-world graphs across diverse tasks, including node classification, regression, and edge weight prediction. Compared to CP baselines, RR-GNN achieves improved efficiency over state-of-the-art methods, with no loss of coverage.
Authors: Firas Laakom, Haobo Chen, J\"urgen Schmidhuber, Yuheng Bu
Abstract: Despite substantial progress in promoting fairness in high-stake applications using machine learning models, existing methods often modify the training process, such as through regularizers or other interventions, but lack formal guarantees that fairness achieved during training will generalize to unseen data. Although overfitting with respect to prediction performance has been extensively studied, overfitting in terms of fairness loss has received far less attention. This paper proposes a theoretical framework for analyzing fairness generalization error through an information-theoretic lens. Our novel bounding technique is based on Efron-Stein inequality, which allows us to derive tight information-theoretic fairness generalization bounds with both Mutual Information (MI) and Conditional Mutual Information (CMI). Our empirical results validate the tightness and practical relevance of these bounds across diverse fairness-aware learning algorithms. Our framework offers valuable insights to guide the design of algorithms improving fairness generalization.
Authors: Mirko Paolo Barbato, Giorgia Rigamonti, Davide Marelli, Paolo Napoletano
Abstract: Type 1 Diabetes (T1D) affects millions worldwide, requiring continuous monitoring to prevent severe hypo- and hyperglycemic events. While continuous glucose monitoring has improved blood glucose management, deploying predictive models on wearable devices remains challenging due to computational and memory constraints. To address this, we propose a novel Lightweight Sequential Transformer model designed for blood glucose prediction in T1D. By integrating the strengths of Transformers' attention mechanisms and the sequential processing of recurrent neural networks, our architecture captures long-term dependencies while maintaining computational efficiency. The model is optimized for deployment on resource-constrained edge devices and incorporates a balanced loss function to handle the inherent data imbalance in hypo- and hyperglycemic events. Experiments on two benchmark datasets, OhioT1DM and DiaTrend, demonstrate that the proposed model outperforms state-of-the-art methods in predicting glucose levels and detecting adverse events. This work fills the gap between high-performance modeling and practical deployment, providing a reliable and efficient T1D management solution.
Authors: Sigma Jahan, Mohammad Masudur Rahman
Abstract: As attention-based deep learning models scale in size and complexity, diagnosing their faults becomes increasingly challenging. In this work, we conduct an empirical study to evaluate the potential of Hessian-based analysis for diagnosing faults in attention-based models. Specifically, we use Hessian-derived insights to identify fragile regions (via curvature analysis) and parameter interdependencies (via parameter interaction analysis) within attention mechanisms. Through experiments on three diverse models (HAN, 3D-CNN, DistilBERT), we show that Hessian-based metrics can localize instability and pinpoint fault sources more effectively than gradients alone. Our empirical findings suggest that these metrics could significantly improve fault diagnosis in complex neural architectures, potentially improving software debugging practices.
Authors: Rajat Rasal, Avinash Kori, Fabio De Sousa Ribeiro, Tian Xia, Ben Glocker
Abstract: Counterfactual image generation presents significant challenges, including preserving identity, maintaining perceptual quality, and ensuring faithfulness to an underlying causal model. While existing auto-encoding frameworks admit semantic latent spaces which can be manipulated for causal control, they struggle with scalability and fidelity. Advancements in diffusion models present opportunities for improving counterfactual image editing, having demonstrated state-of-the-art visual quality, human-aligned perception and representation learning capabilities. Here, we present a suite of diffusion-based causal mechanisms, introducing the notions of spatial, semantic and dynamic abduction. We propose a general framework that integrates semantic representations into diffusion models through the lens of Pearlian causality to edit images via a counterfactual reasoning process. To our knowledge, this is the first work to consider high-level semantic identity preservation for diffusion counterfactuals and to demonstrate how semantic control enables principled trade-offs between faithful causal control and identity preservation.
Authors: Anand Ganesh, Babhrubahan Bose, Anand Rajagopalan
Abstract: We construct four Schauder bases for the space $C[0,1]$, one using ReLU functions, another using Softplus functions, and two more using sigmoidal versions of the ReLU and Softplus functions. This establishes the existence of a basis using these functions for the first time, and improves on the universal approximation property associated with them.
Authors: Sifan Wang, Zehao Dou, Tong-Rui Liu, Lu Lu
Abstract: Recent advances in generative modeling -- particularly diffusion models and flow matching -- have achieved remarkable success in synthesizing discrete data such as images and videos. However, adapting these models to physical applications remains challenging, as the quantities of interest are continuous functions governed by complex physical laws. Here, we introduce $\textbf{FunDiff}$, a novel framework for generative modeling in function spaces. FunDiff combines a latent diffusion process with a function autoencoder architecture to handle input functions with varying discretizations, generate continuous functions evaluable at arbitrary locations, and seamlessly incorporate physical priors. These priors are enforced through architectural constraints or physics-informed loss functions, ensuring that generated samples satisfy fundamental physical laws. We theoretically establish minimax optimality guarantees for density estimation in function spaces, showing that diffusion-based estimators achieve optimal convergence rates under suitable regularity conditions. We demonstrate the practical effectiveness of FunDiff across diverse applications in fluid dynamics and solid mechanics. Empirical results show that our method generates physically consistent samples with high fidelity to the target distribution and exhibits robustness to noisy and low-resolution data. Code and datasets are publicly available at https://github.com/sifanexisted/fundiff.
Authors: Kevin Rojas, Yuchen Zhu, Sichen Zhu, Felix X. -F. Ye, Molei Tao
Abstract: Diffusion models have demonstrated remarkable performance in generating unimodal data across various tasks, including image, video, and text generation. On the contrary, the joint generation of multimodal data through diffusion models is still in the early stages of exploration. Existing approaches heavily rely on external preprocessing protocols, such as tokenizers and variational autoencoders, to harmonize varied data representations into a unified, unimodal format. This process heavily demands the high accuracy of encoders and decoders, which can be problematic for applications with limited data. To lift this restriction, we propose a novel framework for building multimodal diffusion models on arbitrary state spaces, enabling native generation of coupled data across different modalities. By introducing an innovative decoupled noise schedule for each modality, we enable both unconditional and modality-conditioned generation within a single model simultaneously. We empirically validate our approach for text-image generation and mixed-type tabular data synthesis, demonstrating that it achieves competitive performance.
Authors: Vahid Balazadeh, Hamidreza Kamkari, Valentin Thomas, Benson Li, Junwei Ma, Jesse C. Cresswell, Rahul G. Krishnan
Abstract: Causal effect estimation from observational data is fundamental across various applications. However, selecting an appropriate estimator from dozens of specialized methods demands substantial manual effort and domain expertise. We present CausalPFN, a single transformer that amortizes this workflow: trained once on a large library of simulated data-generating processes that satisfy ignorability, it infers causal effects for new observational datasets out-of-the-box. CausalPFN combines ideas from Bayesian causal inference with the large-scale training protocol of prior-fitted networks (PFNs), learning to map raw observations directly to causal effects without any task-specific adjustment. Our approach achieves superior average performance on heterogeneous and average treatment effect estimation benchmarks (IHDP, Lalonde, ACIC). Moreover, it shows competitive performance for real-world policy making on uplift modeling tasks. CausalPFN provides calibrated uncertainty estimates to support reliable decision-making based on Bayesian principles. This ready-to-use model does not require any further training or tuning and takes a step toward automated causal inference (https://github.com/vdblm/CausalPFN).
Authors: Manuel Brenner, Georgia Koppe
Abstract: Memory and long-range temporal processing are core requirements for sequence modeling tasks across natural language processing, time-series forecasting, speech recognition, and control. While nonlinear recurrence has long been viewed as essential for enabling such mechanisms, recent work suggests that linear dynamics may often suffice. In this study, we go beyond performance comparisons to systematically dissect the functional role of nonlinearity in recurrent networks--identifying both when it is computationally necessary, and what mechanisms it enables. We use Almost Linear Recurrent Neural Networks (AL-RNNs), which allow fine-grained control over nonlinearity, as both a flexible modeling tool and a probe into the internal mechanisms of memory. Across a range of classic sequence modeling tasks and a real-world stimulus selection task, we find that minimal nonlinearity is not only sufficient but often optimal, yielding models that are simpler, more robust, and more interpretable than their fully nonlinear or linear counterparts. Our results provide a principled framework for selectively introducing nonlinearity, bridging dynamical systems theory with the functional demands of long-range memory and structured computation in recurrent neural networks, with implications for both artificial and biological neural systems.
Authors: Hossein Babaei, Mel White, Richard G. Baraniuk
Abstract: State Space Models (SSMs) have emerged as powerful components for sequence modeling, enabling efficient handling of long-range dependencies via linear recurrence and convolutional computation. However, their effectiveness depends heavily on the choice and initialization of the state matrix. In this work, we build on the SaFARi framework and existing WaLRUS SSMs to introduce a new variant, W4S4 (WaLRUS for S4), a new class of SSMs constructed from redundant wavelet frames. WaLRUS admits a stable diagonalization and supports fast kernel computation without requiring low-rank approximations, making it both theoretically grounded and computationally efficient. We show that WaLRUS retains information over long horizons significantly better than HiPPO-based SSMs, both in isolation and when integrated into deep architectures such as S4. Our experiments demonstrate consistent improvements across delay reconstruction tasks, classification benchmarks, and long-range sequence modeling, confirming that high-quality, structured initialization enabled by wavelet-based state dynamic offers substantial advantages over existing alternatives. WaLRUS provides a scalable and versatile foundation for the next generation of deep SSM-based models.
Authors: Amirreza Yasami, Mohammadali Tofigh, Mahdi Shahbakhti, Charles Robert Koch
Abstract: Accurate driving cycle construction is crucial for vehicle design, fuel economy analysis, and environmental impact assessments. A generative Physics-Informed Expected SARSA-Monte Carlo (PIESMC) approach that constructs representative driving cycles by capturing transient dynamics, acceleration, deceleration, idling, and road grade transitions while ensuring model fidelity is introduced. Leveraging a physics-informed reinforcement learning framework with Monte Carlo sampling, PIESMC delivers efficient cycle construction with reduced computational cost. Experimental evaluations on two real-world datasets demonstrate that PIESMC replicates key kinematic and energy metrics, achieving up to a 57.3% reduction in cumulative kinematic fragment errors compared to the Micro-trip-based (MTB) method and a 10.5% reduction relative to the Markov-chain-based (MCB) method. Moreover, it is nearly an order of magnitude faster than conventional techniques. Analyses of vehicle-specific power distributions and wavelet-transformed frequency content further confirm its ability to reproduce experimental central tendencies and variability.
Authors: Lev V. Utkin, Semen P. Khomets, Vlada A. Efremenko, Andrei V. Konstantinov, Natalya M. Verbova
Abstract: Survival analysis predicts the time until an event of interest, such as failure or death, but faces challenges due to censored data, where some events remain unobserved. Ensemble-based models, like random survival forests and gradient boosting, are widely used but can produce unstable predictions due to variations in bootstrap samples. To address this, we propose SurvBESA (Survival Beran Estimators Self-Attended), a novel ensemble model that combines Beran estimators with a self-attention mechanism. Unlike traditional methods, SurvBESA applies self-attention to predicted survival functions, smoothing out noise by adjusting each survival function based on its similarity to neighboring survival functions. We also explore a special case using Huber's contamination model to define attention weights, simplifying training to a quadratic or linear optimization problem. Numerical experiments show that SurvBESA outperforms state-of-the-art models. The implementation of SurvBESA is publicly available.
Authors: Kasimir Schulz, Kenneth Yeung, Kieran Evans
Abstract: Natural Language Processing (NLP) models are used for text-related tasks such as classification and generation. To complete these tasks, input data is first tokenized from human-readable text into a format the model can understand, enabling it to make inferences and understand context. Text classification models can be implemented to guard against threats such as prompt injection attacks against Large Language Models (LLMs), toxic input and cybersecurity risks such as spam emails. In this paper, we introduce TokenBreak: a novel attack that can bypass these protection models by taking advantage of the tokenization strategy they use. This attack technique manipulates input text in such a way that certain models give an incorrect classification. Importantly, the end target (LLM or email recipient) can still understand and respond to the manipulated text and therefore be vulnerable to the very attack the protection model was put in place to prevent. The tokenizer is tied to model architecture, meaning it is possible to predict whether or not a model is vulnerable to attack based on family. We also present a defensive strategy as an added layer of protection that can be implemented without having to retrain the defensive model.
Authors: Anastasios N. Angelopoulos, Jacob Eisenstein, Jonathan Berant, Alekh Agarwal, Adam Fisch
Abstract: The development lifecycle of generative AI systems requires continual evaluation, data acquisition, and annotation, which is costly in both resources and time. In practice, rapid iteration often makes it necessary to rely on synthetic annotation data because of the low cost, despite the potential for substantial bias. In this paper, we develop novel, cost-aware methods for actively balancing the use of a cheap, but often inaccurate, weak rater -- such as a model-based autorater that is designed to automatically assess the quality of generated content -- with a more expensive, but also more accurate, strong rater alternative such as a human. More specifically, the goal of our approach is to produce a low variance, unbiased estimate of the mean of the target "strong" rating, subject to some total annotation budget. Building on recent work in active and prediction-powered statistical inference, we derive a family of cost-optimal policies for allocating a given annotation budget between weak and strong raters so as to maximize statistical efficiency. Using synthetic and real-world data, we empirically characterize the conditions under which these policies yield improvements over prior methods. We find that, especially in tasks where there is high variability in the difficulty of examples, our policies can achieve the same estimation precision at a far lower total annotation budget than standard evaluation methods.
Authors: Salah A. Faroughi, Farinaz Mostajeran
Abstract: Physics-informed Kolmogorov-Arnold Networks (PIKANs), and in particular their Chebyshev-based variants (cPIKANs), have recently emerged as promising models for solving partial differential equations (PDEs). However, their training dynamics and convergence behavior remain largely unexplored both theoretically and numerically. In this work, we aim to advance the theoretical understanding of cPIKANs by analyzing them using Neural Tangent Kernel (NTK) theory. Our objective is to discern the evolution of kernel structure throughout gradient-based training and its subsequent impact on learning efficiency. We first derive the NTK of standard cKANs in a supervised setting, and then extend the analysis to the physics-informed context. We analyze the spectral properties of NTK matrices, specifically their eigenvalue distributions and spectral bias, for four representative PDEs: the steady-state Helmholtz equation, transient diffusion and Allen-Cahn equations, and forced vibrations governed by the Euler-Bernoulli beam equation. We also conduct an investigation into the impact of various optimization strategies, e.g., first-order, second-order, and hybrid approaches, on the evolution of the NTK and the resulting learning dynamics. Results indicate a tractable behavior for NTK in the context of cPIKANs, which exposes learning dynamics that standard physics-informed neural networks (PINNs) cannot capture. Spectral trends also reveal when domain decomposition improves training, directly linking kernel behavior to convergence rates under different setups. To the best of our knowledge, this is the first systematic NTK study of cPIKANs, providing theoretical insight that clarifies and predicts their empirical performance.
Authors: Jacob Helwig, Sai Sreeharsha Adavi, Xuan Zhang, Yuchao Lin, Felix S. Chim, Luke Takeshi Vizzini, Haiyang Yu, Muhammad Hasnain, Saykat Kumar Biswas, John J. Holloway, Narendra Singh, N. K. Anand, Swagnik Guhathakurta, Shuiwang Ji
Abstract: We consider the problem of modeling high-speed flows using machine learning methods. While most prior studies focus on low-speed fluid flows in which uniform time-stepping is practical, flows approaching and exceeding the speed of sound exhibit sudden changes such as shock waves. In such cases, it is essential to use adaptive time-stepping methods to allow a temporal resolution sufficient to resolve these phenomena while simultaneously balancing computational costs. Here, we propose a two-phase machine learning method, known as ShockCast, to model high-speed flows with adaptive time-stepping. In the first phase, we propose to employ a machine learning model to predict the timestep size. In the second phase, the predicted timestep is used as an input along with the current fluid fields to advance the system state by the predicted timestep. We explore several physically-motivated components for timestep prediction and introduce timestep conditioning strategies inspired by neural ODE and Mixture of Experts. As ShockCast is the first framework for learning high-speed flows, we evaluate our methods by generating two supersonic flow datasets, available at https://huggingface.co/datasets/divelab. Our code is publicly available as part of the AIRS library (https://github.com/divelab/AIRS).
URLs: https://huggingface.co/datasets/divelab., https://github.com/divelab/AIRS).
Authors: Hongzheng Chen, Yingheng Wang, Yaohui Cai, Hins Hu, Jiajie Li, Shirley Huang, Chenhui Deng, Rongjian Liang, Shufeng Kong, Haoxing Ren, Samitha Samaranayake, Carla P. Gomes, Zhiru Zhang
Abstract: While Large Language Models (LLMs) have demonstrated significant advancements in reasoning and agent-based problem-solving, current evaluation methodologies fail to adequately assess their capabilities: existing benchmarks either rely on closed-ended questions prone to saturation and memorization, or subjective comparisons that lack consistency and rigor. In this work, we introduce HeuriGym, an agentic framework designed for evaluating heuristic algorithms generated by LLMs for combinatorial optimization problems, characterized by clearly defined objectives and expansive solution spaces. HeuriGym empowers LLMs to propose heuristics, receive evaluative feedback via code execution, and iteratively refine their solutions. We evaluate nine state-of-the-art models on nine problems across domains such as computer systems, logistics, and biology, exposing persistent limitations in tool use, planning, and adaptive reasoning. To quantify performance, we propose the Quality-Yield Index (QYI), a metric that captures both solution pass rate and quality. Even top models like GPT-o4-mini-high and Gemini-2.5-Pro attain QYI scores of only 0.6, well below the expert baseline of 1. Our open-source benchmark aims to guide the development of LLMs toward more effective and realistic problem-solving in scientific and engineering domains.
Authors: Caleb Zheng, Eli Shlizerman
Abstract: A variety of pruning methods have been introduced for over-parameterized Recurrent Neural Networks to improve efficiency in terms of power consumption and storage utilization. These advances motivate a new paradigm, termed `hyperpruning', which seeks to identify the most suitable pruning strategy for a given network architecture and application. Unlike conventional hyperparameter search, where the optimal configuration's accuracy remains uncertain, in the context of network pruning, the accuracy of the dense model sets the target for the accuracy of the pruned one. The goal, therefore, is to discover pruned variants that match or even surpass this established accuracy. However, exhaustive search over pruning configurations is computationally expensive and lacks early performance guarantees. To address this challenge, we propose a novel Lyapunov Spectrum (LS)-based distance metric that enables early comparison between pruned and dense networks, allowing accurate prediction of post-training performance. By integrating this LS-based distance with standard hyperparameter optimization algorithms, we introduce an efficient hyperpruning framework, termed LS-based Hyperpruning (LSH). LSH reduces search time by an order of magnitude compared to conventional approaches relying on full training. Experiments on stacked LSTM and RHN architectures using the Penn Treebank dataset, and on AWD-LSTM-MoS using WikiText-2, demonstrate that under fixed training budgets and target pruning ratios, LSH consistently identifies superior pruned models. Remarkably, these pruned variants not only outperform those selected by loss-based baseline but also exceed the performance of their dense counterpart.
Authors: Junhong Shen, Hao Bai, Lunjun Zhang, Yifei Zhou, Amrith Setlur, Shengbang Tong, Diego Caples, Nan Jiang, Tong Zhang, Ameet Talwalkar, Aviral Kumar
Abstract: The current paradigm of test-time scaling relies on generating long reasoning traces ("thinking" more) before producing a response. In agent problems that require interaction, this can be done by generating thinking traces before acting in the world. However, this process does not allow agents to acquire new information from the environment or adapt their behavior over time. In this work, we propose to scale test-time interaction, an untapped dimension of test-time scaling that increases the agent's interaction horizon to enable running rich behaviors such as exploration, backtracking, and dynamic re-planning within a single rollout. To demonstrate the promise of this scaling dimension, we study the domain of web agents. We first show that even prompting-based interaction scaling without any training can improve task success on web benchmarks non-trivially. Building on this, we introduce TTI (Test-Time Interaction), a curriculum-based online reinforcement learning (RL) approach that trains agents by adaptively adjusting their rollout lengths. Using a Gemma 3 12B model, TTI produces state-of-the-art open-source, open-data web agents on WebVoyager and WebArena benchmarks. We further show that TTI enables agents to balance exploration and exploitation adaptively. Our results establish interaction scaling as a powerful, complementary axis to scaling per-step compute, offering new avenues for training adaptive agents.
Authors: Alberto Baz\'an-Guill\'en, Carlos Beis-Penedo, Diego Cajaraville-Aboy, Pablo Barbecho-Bautista, Rebeca P. D\'iaz-Redondo, Luis J. de la Cruz Llopis, Ana Fern\'andez-Vilas, M\'onica Aguilar Igartua, Manuel Fern\'andez-Veiga
Abstract: Realistic urban traffic simulation is essential for sustainable urban planning and the development of intelligent transportation systems. However, generating high-fidelity, time-varying traffic profiles that accurately reflect real-world conditions, especially in large-scale scenarios, remains a major challenge. Existing methods often suffer from limitations in accuracy, scalability, or raise privacy concerns due to centralized data processing. This work introduces DesRUTGe (Decentralized Realistic Urban Traffic Generator), a novel framework that integrates Deep Reinforcement Learning (DRL) agents with the SUMO simulator to generate realistic 24-hour traffic patterns. A key innovation of DesRUTGe is its use of Decentralized Federated Learning (DFL), wherein each traffic detector and its corresponding urban zone function as an independent learning node. These nodes train local DRL models using minimal historical data and collaboratively refine their performance by exchanging model parameters with selected peers (e.g., geographically adjacent zones), without requiring a central coordinator. Evaluated using real-world data from the city of Barcelona, DesRUTGe outperforms standard SUMO-based tools such as RouteSampler, as well as other centralized learning approaches, by delivering more accurate and privacy-preserving traffic pattern generation.
Authors: Boya Zeng, Yida Yin, Zhiqiu Xu, Zhuang Liu
Abstract: Generative models, with their success in image and video generation, have recently been explored for synthesizing effective neural network weights. These approaches take trained neural network checkpoints as training data, and aim to generate high-performing neural network weights during inference. In this work, we examine four representative methods on their ability to generate novel model weights, i.e., weights that are different from the checkpoints seen during training. Surprisingly, we find that these methods synthesize weights largely by memorization: they produce either replicas, or at best simple interpolations, of the training checkpoints. Current methods fail to outperform simple baselines, such as adding noise to the weights or taking a simple weight ensemble, in obtaining different and simultaneously high-performing models. We further show that this memorization cannot be effectively mitigated by modifying modeling factors commonly associated with memorization in image diffusion models, or applying data augmentations. Our findings provide a realistic assessment of what types of data current generative models can model, and highlight the need for more careful evaluation of generative models in new domains. Our code is available at https://github.com/boyazeng/weight_memorization.
Authors: Zeju Qiu, Simon Buchholz, Tim Z. Xiao, Maximilian Dax, Bernhard Sch\"olkopf, Weiyang Liu
Abstract: While large language models (LLMs) are driving the rapid advancement of artificial intelligence, effectively and reliably training these large models remains one of the field's most significant challenges. To address this challenge, we propose POET, a novel reParameterized training algorithm that uses Orthogonal Equivalence Transformation to optimize neurons. Specifically, POET reparameterizes each neuron with two learnable orthogonal matrices and a fixed random weight matrix. Because of its provable preservation of spectral properties of weight matrices, POET can stably optimize the objective function with improved generalization. We further develop efficient approximations that make POET flexible and scalable for training large-scale neural networks. Extensive experiments validate the effectiveness and scalability of POET in training LLMs.
Authors: Marco Di Gennaro, Francesco Panebianco, Marco Pianta, Stefano Zanero, Michele Carminati
Abstract: Money laundering is a financial crime that poses a serious threat to financial integrity and social security. The growing number of transactions makes it necessary to use automatic tools that help law enforcement agencies detect such criminal activity. In this work, we present Amatriciana, a novel approach based on Graph Neural Networks to detect money launderers inside a graph of transactions by considering temporal information. Amatriciana uses the whole graph of transactions without splitting it into several time-based subgraphs, exploiting all relational information in the dataset. Our experiments on a public dataset reveal that the model can learn from a limited amount of data. Furthermore, when more data is available, the model outperforms other State-of-the-art approaches; in particular, Amatriciana decreases the number of False Positives (FPs) while detecting many launderers. In summary, Amatriciana achieves an F1 score of 0.76. In addition, it lowers the FPs by 55% with respect to other State-of-the-art models.
Authors: Jiatao Gu, Tianrong Chen, David Berthelot, Huangjie Zheng, Yuyang Wang, Ruixiang Zhang, Laurent Dinh, Miguel Angel Bautista, Josh Susskind, Shuangfei Zhai
Abstract: We present STARFlow, a scalable generative model based on normalizing flows that achieves strong performance in high-resolution image synthesis. The core of STARFlow is Transformer Autoregressive Flow (TARFlow), which combines the expressive power of normalizing flows with the structured modeling capabilities of Autoregressive Transformers. We first establish the theoretical universality of TARFlow for modeling continuous distributions. Building on this foundation, we introduce several key architectural and algorithmic innovations to significantly enhance scalability: (1) a deep-shallow design, wherein a deep Transformer block captures most of the model representational capacity, complemented by a few shallow Transformer blocks that are computationally efficient yet substantially beneficial; (2) modeling in the latent space of pretrained autoencoders, which proves more effective than direct pixel-level modeling; and (3) a novel guidance algorithm that significantly boosts sample quality. Crucially, our model remains an end-to-end normalizing flow, enabling exact maximum likelihood training in continuous spaces without discretization. STARFlow achieves competitive performance in both class-conditional and text-conditional image generation tasks, approaching state-of-the-art diffusion models in sample quality. To our knowledge, this work is the first successful demonstration of normalizing flows operating effectively at this scale and resolution.
Authors: Xueying Ding, Aakriti Mittal, Achintya Gopal
Abstract: Time-series data is a vital modality within data science communities. This is particularly valuable in financial applications, where it helps in detecting patterns, understanding market behavior, and making informed decisions based on historical data. Recent advances in language modeling have led to the rise of time-series pre-trained models that are trained on vast collections of datasets and applied to diverse tasks across financial domains. However, across financial applications, existing time-series pre-trained models have not shown boosts in performance over simple finance benchmarks in both zero-shot and fine-tuning settings. This phenomenon occurs because of a i) lack of financial data within the pre-training stage, and ii) the negative transfer effect due to inherently different time-series patterns across domains. Furthermore, time-series data is continuous, noisy, and can be collected at varying frequencies and with varying lags across different variables, making this data more challenging to model than languages. To address the above problems, we introduce a Pre-trained MoDEL for FINance TimE-series (Delphyne). Delphyne achieves competitive performance to existing foundation and full-shot models with few fine-tuning steps on publicly available datasets, and also shows superior performances on various financial tasks.
Authors: Daniel Thilo Schroeder, Meeyoung Cha, Andrea Baronchelli, Nick Bostrom, Nicholas A. Christakis, David Garcia, Amit Goldenberg, Yara Kyrychenko, Kevin Leyton-Brown, Nina Lutz, Gary Marcus, Filippo Menczer, Gordon Pennycook, David G. Rand, Frank Schweitzer, Christopher Summerfield, Audrey Tang, Jay Van Bavel, Sander van der Linden, Dawn Song, Jonas R. Kunst
Abstract: Advances in AI portend a new era of sophisticated disinformation operations. While individual AI systems already create convincing -- and at times misleading -- information, an imminent development is the emergence of malicious AI swarms. These systems can coordinate covertly, infiltrate communities, evade traditional detectors, and run continuous A/B tests, with round-the-clock persistence. The result can include fabricated grassroots consensus, fragmented shared reality, mass harassment, voter micro-suppression or mobilization, contamination of AI training data, and erosion of institutional trust. With democratic processes worldwide increasingly vulnerable, we urge a three-pronged response: (1) platform-side defenses -- always-on swarm-detection dashboards, pre-election high-fidelity swarm-simulation stress-tests, transparency audits, and optional client-side "AI shields" for users; (2) model-side safeguards -- standardized persuasion-risk tests, provenance-authenticating passkeys, and watermarking; and (3) system-level oversight -- a UN-backed AI Influence Observatory.
Authors: No\'emie Bergues, Arthur Carr\'e, Paul Join-Lambert, Brice Hoffmann, Arnaud Blondel, Hamza Tajmouati
Abstract: Predicting the 3D conformation of small molecules within protein binding sites is a key challenge in drug design. When a crystallized reference ligand (template) is available, it provides geometric priors that can guide 3D pose prediction. We present a two-stage method for ligand conformation generation guided by such templates. In the first stage, we introduce a molecular alignment approach based on flow-matching to generate 3D coordinates for the ligand, using the template structure as a reference. In the second stage, a differentiable pose optimization procedure refines this conformation based on shape and pharmacophore similarities, internal energy, and, optionally, the protein binding pocket. We evaluate our approach on a new benchmark of ligand pairs co-crystallized with the same target and show that it outperforms standard docking tools and open-access alignment methods, especially in cases involving low similarity to the template or high ligand flexibility.
Authors: Ali Abedi, Charlene H. Chu, Shehroz S. Khan
Abstract: Agitation is one of the most common responsive behaviors in people living with dementia, particularly among those residing in community settings without continuous clinical supervision. Timely prediction of agitation can enable early intervention, reduce caregiver burden, and improve the quality of life for both patients and caregivers. This study aimed to develop and benchmark machine learning approaches for the early prediction of agitation in community-dwelling older adults with dementia using multimodal sensor data. A new set of agitation-related contextual features derived from activity data was introduced and employed for agitation prediction. A wide range of machine learning and deep learning models was evaluated across multiple problem formulations, including binary classification for single-timestamp tabular sensor data and multi-timestamp sequential sensor data, as well as anomaly detection for single-timestamp tabular sensor data. The study utilized the Technology Integrated Health Management (TIHM) dataset, the largest publicly available dataset for remote monitoring of people living with dementia, comprising 2,803 days of in-home activity, physiology, and sleep data. The most effective setting involved binary classification of sensor data using the current 6-hour timestamp to predict agitation at the subsequent timestamp. Incorporating additional information, such as time of day and agitation history, further improved model performance, with the highest AUC-ROC of 0.9720 and AUC-PR of 0.4320 achieved by the light gradient boosting machine. This work presents the first comprehensive benchmarking of state-of-the-art techniques for agitation prediction in community-based dementia care using privacy-preserving sensor data. The approach enables accurate, explainable, and efficient agitation prediction, supporting proactive dementia care and aging in place.
Authors: Adoubi Vincent De Paul Adombi
Abstract: Scientific machine learning (SciML) provides a structured approach to integrating physical knowledge into data-driven modeling, offering significant potential for advancing hydrological research. In recent years, multiple methodological families have emerged, including physics-informed machine learning, physics-guided machine learning, hybrid physics-machine learning, and data-driven physics discovery. Within each of these families, a proliferation of heterogeneous approaches has developed independently, often without conceptual coordination. This fragmentation complicates the assessment of methodological novelty and makes it difficult to identify where meaningful advances can still be made in the absence of a unified conceptual framework. This review, the first focused overview of SciML in hydrology, addresses these limitations by proposing a unified methodological framework for each SciML family, bringing together representative contributions into a coherent structure that fosters conceptual clarity and supports cumulative progress in hydrological modeling. Finally, we highlight the limitations and future opportunities of each unified family to guide systematic research in hydrology, where these methods remain underutilized.
Authors: Mohamed Kefi, Tien Dat Pham, Thin Nguyen, Mark G. Tjoelker, Viola Devasirvatham, Kenichi Kashiwagi
Abstract: Olive production is an important tree crop in Mediterranean climates. However, olive yield varies significantly due to climate change. Accurately estimating yield using remote sensing and machine learning remains a complex challenge. In this study, we developed a streamlined pipeline for olive yield estimation in the Kairouan and Sousse governorates of Tunisia. We extracted features from multispectral reflectance bands, vegetation indices derived from Landsat-8 OLI and Landsat-9 OLI-2 satellite imagery, along with digital elevation model data. These spatial features were combined with ground-based field survey data to form a structured tabular dataset. We then developed an automated ensemble learning framework, implemented using AutoGluon to train and evaluate multiple machine learning models, select optimal combinations through stacking, and generate robust yield predictions using five-fold cross-validation. The results demonstrate strong predictive performance from both sensors, with Landsat-8 OLI achieving R2 = 0.8635 and RMSE = 1.17 tons per ha, and Landsat-9 OLI-2 achieving R2 = 0.8378 and RMSE = 1.32 tons per ha. This study highlights a scalable, cost-effective, and accurate method for olive yield estimation, with potential applicability across diverse agricultural regions globally.
Authors: Xiaoyu Sun, Yang Yang, Xunde Dong
Abstract: In the field of automatic Electrocardiogram (ECG) diagnosis, due to the relatively limited amount of labeled data, how to build a robust ECG pretrained model based on unlabeled data is a key area of focus for researchers. Recent advancements in contrastive learning-based ECG pretrained models highlight the potential of exploiting the additional patient-level self-supervisory signals inherent in ECG. They are referred to as patient contrastive learning. Its rationale is that multiple physical recordings from the same patient may share commonalities, termed patient consistency, so redefining positive and negative pairs in contrastive learning as intrapatient and inter-patient samples provides more shared context to learn an effective representation. However, these methods still fail to efficiently exploit patient consistency due to the insufficient amount of intra-inter patient samples existing in a batch. Hence, we propose a contrastive learning-based ECG pretrained model enhanced by the Patient Memory Queue (PMQ), which incorporates a large patient memory queue to mitigate model degeneration that can arise from insufficient intra-inter patient samples. In order to further enhance the performance of the pretrained model, we introduce two extra data augmentation methods to provide more perspectives of positive and negative pairs for pretraining. Extensive experiments were conducted on three public datasets with three different data ratios. The experimental results show that the comprehensive performance of our method outperforms previous contrastive learning methods and exhibits greater robustness in scenarios with limited labeled data. The code is available at https://github.com/3hiuwoo/PMQ.
Authors: Meiyan Kang, Shizuo Kaji, Sang-Yun Lee, Taegon Kim, Hee-Hwan Ryu, Suyoung Choi
Abstract: Ground Penetrating Radar (GPR) is a widely used Non-Destructive Testing (NDT) technique for subsurface exploration, particularly in infrastructure inspection and maintenance. However, conventional interpretation methods are often limited by noise sensitivity and a lack of structural awareness. This study presents a novel framework that enhances the detection of underground utilities, especially pipelines, by integrating shape-aware topological features derived from B-scan GPR images using Topological Data Analysis (TDA), with the spatial detection capabilities of the YOLOv5 deep neural network (DNN). We propose a novel shape-aware topological representation that amplifies structural features in the input data, thereby improving the model's responsiveness to the geometrical features of buried objects. To address the scarcity of annotated real-world data, we employ a Sim2Real strategy that generates diverse and realistic synthetic datasets, effectively bridging the gap between simulated and real-world domains. Experimental results demonstrate significant improvements in mean Average Precision (mAP), validating the robustness and efficacy of our approach. This approach underscores the potential of TDA-enhanced learning in achieving reliable, real-time subsurface object detection, with broad applications in urban planning, safety inspection, and infrastructure management.
Authors: Masoud Rahimi, Reza Karbasi, Abdol-Hossein Vahabie
Abstract: We introduce an open-source Python framework for generating synthetic ECG image datasets to advance critical deep learning-based tasks in ECG analysis, including ECG digitization, lead region and lead name detection, and pixel-level waveform segmentation. Using the PTB-XL signal dataset, our proposed framework produces four open-access datasets: (1) ECG images in various lead configurations paired with time-series signals for ECG digitization, (2) ECG images annotated with YOLO-format bounding boxes for detection of lead region and lead name, (3)-(4) cropped single-lead images with segmentation masks compatible with U-Net-based models in normal and overlapping versions. In the overlapping case, waveforms from neighboring leads are superimposed onto the target lead image, while the segmentation masks remain clean. The open-source Python framework and datasets are publicly available at https://github.com/rezakarbasi/ecg-image-and-signal-dataset and https://doi.org/10.5281/zenodo.15484519, respectively.
URLs: https://github.com/rezakarbasi/ecg-image-and-signal-dataset, https://doi.org/10.5281/zenodo.15484519,
Authors: Abdullah Burkan Bereketoglu
Abstract: Model-free and reinforcement learning-based adaptive filtering methods are gaining traction for denoising in dynamic, non-stationary environments such as wireless signal channels. Traditional filters like LMS, RLS, Wiener, and Kalman are limited by assumptions of stationary or requiring complex fine-tuning or exact noise statistics or fixed models. This letter proposes an adaptive filtering framework using Proximal Policy Optimization (PPO), guided by a composite reward that balances SNR improvement, MSE reduction, and residual smoothness. Experiments on synthetic signals with various noise types show that our PPO agent generalizes beyond its training distribution, achieving real-time performance and outperforming classical filters. This work demonstrates the viability of policy-gradient reinforcement learning for robust, low-latency adaptive signal filtering.
Authors: Mikko Stenlund
Abstract: Predictive coding networks (PCNs) constitute a biologically inspired framework for understanding hierarchical computation in the brain, and offer an alternative to traditional feedforward neural networks in ML. This note serves as a quick, onboarding introduction to PCNs for machine learning practitioners. We cover the foundational network architecture, inference and learning update rules, and algorithmic implementation. A concrete image-classification task (CIFAR-10) is provided as a benchmark-smashing application, together with an accompanying Python notebook containing the PyTorch implementation.
Authors: Alexandre Bouras, Audrey Durand, Richard Khoury
Abstract: This study explores strategies for optimizing news headline recommendations through preference-based learning. Using real-world data of user interactions with French-language online news posts, we learn a headline recommender agent under a contextual bandit setting. This allows us to explore the impact of translation on engagement predictions, as well as the benefits of different interactive strategies on user engagement during data collection. Our results show that explicit exploration may not be required in the presence of noisy contexts, opening the door to simpler but efficient strategies in practice.
Authors: Mohd Ashhad, Sana Rahmani, Mohammed Fayiz, Ali Etemad, Javad Hashemi
Abstract: We propose a deep neural architecture that performs uncertainty-aware multi-view classification of arrhythmia from ECG. Our method learns two different views (1D and 2D) of single-lead ECG to capture different types of information. We use a fusion technique to reduce the conflict between the different views caused by noise and artifacts in ECG data, thus incorporating uncertainty to obtain stronger final predictions. Our framework contains the following three modules (1) a time-series module to learn the morphological features from ECG; (2) an image-space learning module to learn the spatiotemporal features; and (3) the uncertainty-aware fusion module to fuse the information from the two different views. Experimental results on two real-world datasets demonstrate that our framework not only improves the performance on arrhythmia classification compared to the state-of-the-art but also shows better robustness to noise and artifacts present in ECG.
Authors: Taesoo Kim, Jong Hwan Ko
Abstract: Recent advances in speech-enabled language models have shown promising results in building intelligent voice assistants. However, most existing approaches rely on large-scale paired speech-text data and extensive computational resources, which pose challenges in terms of scalability and accessibility. In this paper, we present \textbf{TESU-LLM}, a novel framework that enables training speech-capable language models using only text data. Our key insight is to leverage a unified encoder that maps semantically equivalent text and speech inputs to a shared latent space. By aligning the encoder output with the embedding space of a LLM via a lightweight projection network, we enable the model to generalize from text-only supervision to speech-based inference. Despite being trained exclusively on text, TESU-LLM achieves strong performance on various speech-related benchmarks, comparable to baseline methods trained with large-scale multimodal datasets and substantial computational resources. These results highlight the effectiveness and efficiency of our approach, offering a scalable path toward building speech LLMs without speech data.
Authors: Alex Pierron, Michel Barbeau, Luca De Cicco, Jose Rubio-Hernan, Joaquin Garcia-Alfaro
Abstract: Reconfigurable Intelligent Surfaces (RISs) are composed of physical elements that can dynamically alter electromagnetic wave properties to enhance beamforming and leading to improvements in areas with low coverage properties. They have the potential to be combined with Reinforcement Learning (RL) techniques to achieve network performance and energy efficiency via optimization techniques. In addition to performance and energy improvements, it is also crucial to consider the concept of fair communications. RISs must ensure that User Equipment (UE) units receive their signals with adequate strength, without other UE being deprived of service due to insufficient power. In this paper, we address such a problem. We explore the fairness properties of previous work and propose a novel method that aims at obtaining an efficient and fair duplex RIS-RL system for multiple legitimate UE units. We report and discuss our experimental work and simulation results. We also release our code and datasets to foster further research in the topic.
Authors: Sukru Selim Calik, Andac Akyuz, Zeynep Hilal Kilimci, Kerem Colak
Abstract: Financial literacy is increasingly dependent on the ability to interpret complex financial data and utilize advanced forecasting tools. In this context, this study proposes a novel approach that combines transformer-based time series models with explainable artificial intelligence (XAI) to enhance the interpretability and accuracy of stock price predictions. The analysis focuses on the daily stock prices of the five highest-volume banks listed in the BIST100 index, along with XBANK and XU100 indices, covering the period from January 2015 to March 2025. Models including DLinear, LTSNet, Vanilla Transformer, and Time Series Transformer are employed, with input features enriched by technical indicators. SHAP and LIME techniques are used to provide transparency into the influence of individual features on model outputs. The results demonstrate the strong predictive capabilities of transformer models and highlight the potential of interpretable machine learning to empower individuals in making informed investment decisions and actively engaging in financial markets.
Authors: Wei Li, Xiaochun Wu, Xiaoxi Hu, Yuxuan Zhang, Sebastian Bader, Yuhan Huang
Abstract: Near-sensor diagnosis has become increasingly prevalent in industry. This study proposes a lightweight model named LD-RPMNet that integrates Transformers and Convolutional Neural Networks, leveraging both local and global feature extraction to optimize computational efficiency for a practical railway application. The LD-RPMNet introduces a Multi-scale Depthwise Separable Convolution (MDSC) module, which decomposes cross-channel convolutions into pointwise and depthwise convolutions while employing multi-scale kernels to enhance feature extraction. Meanwhile, a Broadcast Self-Attention (BSA) mechanism is incorporated to simplify complex matrix multiplications and improve computational efficiency. Experimental results based on collected sound signals during the operation of railway point machines demonstrate that the optimized model reduces parameter count and computational complexity by 50% while improving diagnostic accuracy by nearly 3%, ultimately achieving an accuracy of 98.86%. This demonstrates the possibility of near-sensor fault diagnosis applications in railway point machines.
Authors: Vassiliki Mancoridis, Brian Bue, Jake H. Lee, Andrew K. Thorpe, Daniel Cusworth, Alana Ayasse, Philip G. Brodrick, Riley Duren
Abstract: Prioritizing methane for near-term climate action is crucial due to its significant impact on global warming. Previous work used columnwise matched filter products from the airborne AVIRIS-NG imaging spectrometer to detect methane plume sources; convolutional neural networks (CNNs) discerned anthropogenic methane plumes from false positive enhancements. However, as an increasing number of remote sensing platforms are used for methane plume detection, there is a growing need to address cross-platform alignment. In this work, we describe model- and data-driven machine learning approaches that leverage airborne observations to improve spaceborne methane plume detection, reconciling the distributional shifts inherent with performing the same task across platforms. We develop a spaceborne methane plume classifier using data from the EMIT imaging spectroscopy mission. We refine classifiers trained on airborne imagery from AVIRIS-NG campaigns using transfer learning, outperforming the standalone spaceborne model. Finally, we use CycleGAN, an unsupervised image-to-image translation technique, to align the data distributions between airborne and spaceborne contexts. Translating spaceborne EMIT data to the airborne AVIRIS-NG domain using CycleGAN and applying airborne classifiers directly yields the best plume detection results. This methodology is useful not only for data simulation, but also for direct data alignment. Though demonstrated on the task of methane plume detection, our work more broadly demonstrates a data-driven approach to align related products obtained from distinct remote sensing instruments.
Authors: Thien Nhan Vo, Thanh Xuan Truong
Abstract: This study addresses the classification of heartbeats from ECG signals through two distinct approaches: traditional machine learning utilizing hand-crafted features and deep learning via transformed images of ECG beats. The dataset underwent preprocessing steps, including downsampling, filtering, and normalization, to ensure consistency and relevance for subsequent analysis. In the first approach, features such as heart rate variability (HRV), mean, variance, and RR intervals were extracted to train various classifiers, including SVM, Random Forest, AdaBoost, LSTM, Bi-directional LSTM, and LightGBM. The second approach involved transforming ECG signals into images using Gramian Angular Field (GAF), Markov Transition Field (MTF), and Recurrence Plots (RP), with these images subsequently classified using CNN architectures like VGG and Inception. Experimental results demonstrate that the LightGBM model achieved the highest performance, with an accuracy of 99% and an F1 score of 0.94, outperforming the image-based CNN approach (F1 score of 0.85). Models such as SVM and AdaBoost yielded significantly lower scores, indicating limited suitability for this task. The findings underscore the superior ability of hand-crafted features to capture temporal and morphological variations in ECG signals compared to image-based representations of individual beats. Future investigations may benefit from incorporating multi-lead ECG signals and temporal dependencies across successive beats to enhance classification accuracy further.
Authors: Alexis Le Pichon, Alice Janela Cameijo, Samir Aknine, Youcef Sklab, Souhila Arib, Quentin Brissaud, Sven Peter Naesholm
Abstract: Accurate modeling of infrasound transmission losses (TLs) is essential to assess the performance of the global International Monitoring System infrasound network. Among existing propagation modeling tools, parabolic equation (PE) method enables TLs to be finely modeled, but its computational cost does not allow exploration of a large parameter space for operational monitoring applications. To reduce computation times, Brissaud et al. 2023 explored the potential of convolutional neural networks trained on a large set of regionally simulated wavefields (< 1000 km from the source) to predict TLs with negligible computation times compared to PE simulations. However, this method struggles in unfavorable initial wind conditions, especially at high frequencies, and causal issues with winds at large distances from the source affecting ground TLs close to the source. In this study, we have developed an optimized convolutional network designed to minimize prediction errors while predicting TLs from globally simulated combined temperature and wind fields spanning over propagation ranges of 4000 km. Our approach enhances the previously proposed one by implementing key optimizations that improve the overall architecture performance. The implemented model predicts TLs with an average error of 8.6 dB in the whole frequency band (0.1-3.2 Hz) and explored realistic atmospheric scenarios.
Authors: Naseem Babu, Jimson Mathew, A. P. Vinod
Abstract: The growing convergence between Large Language Models (LLMs) and electroencephalography (EEG) research is enabling new directions in neural decoding, brain-computer interfaces (BCIs), and affective computing. This survey offers a systematic review and structured taxonomy of recent advancements that utilize LLMs for EEG-based analysis and applications. We organize the literature into four domains: (1) LLM-inspired foundation models for EEG representation learning, (2) EEG-to-language decoding, (3) cross-modal generation including image and 3D object synthesis, and (4) clinical applications and dataset management tools. The survey highlights how transformer-based architectures adapted through fine-tuning, few-shot, and zero-shot learning have enabled EEG-based models to perform complex tasks such as natural language generation, semantic interpretation, and diagnostic assistance. By offering a structured overview of modeling strategies, system designs, and application areas, this work serves as a foundational resource for future work to bridge natural language processing and neural signal analysis through language models.
Authors: Yimin Du
Abstract: This paper presents a sophisticated multi-day turnover quantitative trading algorithm that integrates advanced deep learning techniques with comprehensive cross-sectional stock prediction for the Chinese A-share market. Our framework combines five interconnected modules: initial stock selection through deep cross-sectional prediction networks, opening signal distribution analysis using mixture models for arbitrage identification, market capitalization and liquidity-based dynamic position sizing, grid-search optimized profit-taking and stop-loss mechanisms, and multi-granularity volatility-based market timing models. The algorithm employs a novel approach to balance capital efficiency with risk management through adaptive holding periods and sophisticated entry/exit timing. Trained on comprehensive A-share data from 2010-2020 and rigorously backtested on 2021-2024 data, our method achieves remarkable performance with 15.2\% annualized returns, maximum drawdown constrained below 5\%, and a Sharpe ratio of 1.87. The strategy demonstrates exceptional scalability by maintaining 50-100 daily positions with a 9-day maximum holding period, incorporating dynamic profit-taking and stop-loss mechanisms that enhance capital turnover efficiency while preserving risk-adjusted returns. Our approach exhibits robust performance across various market regimes while maintaining high capital capacity suitable for institutional deployment.
Authors: Jiyao Wang, Suzan Ayas, Jiahao Zhang, Xiao Wen, Dengbo He, Birsen Donmez
Abstract: Accurately detecting drowsiness is vital to driving safety. Among all measures, physiological-signal-based drowsiness monitoring can be more privacy-preserving than a camera-based approach. However, conflicts exist regarding how physiological metrics are associated with different drowsiness labels across datasets. Thus, we analyzed key features from electrocardiograms (ECG), electrodermal activity (EDA), and respiratory (RESP) signals across four datasets, where different drowsiness inducers (such as fatigue and low arousal) and assessment methods (subjective vs. objective) were used. Binary logistic regression models were built to identify the physiological metrics that are associated with drowsiness. Findings indicate that distinct different drowsiness inducers can lead to different physiological responses, and objective assessments were more sensitive than subjective ones in detecting drowsiness. Further, the increased heart rate stability, reduced respiratory amplitude, and decreased tonic EDA are robustly associated with increased drowsiness. The results enhance understanding of drowsiness detection and can inform future generalizable monitoring designs.
Authors: Tim Schneider, Guillaume Duret, Cristiana de Farias, Roberto Calandra, Liming Chen, Jan Peters
Abstract: Tactile perception has the potential to significantly enhance dexterous robotic manipulation by providing rich local information that can complement or substitute for other sensory modalities such as vision. However, because tactile sensing is inherently local, it is not well-suited for tasks that require broad spatial awareness or global scene understanding on its own. A human-inspired strategy to address this issue is to consider active perception techniques instead. That is, to actively guide sensors toward regions with more informative or significant features and integrate such information over time in order to understand a scene or complete a task. Both active perception and different methods for tactile sensing have received significant attention recently. Yet, despite advancements, both fields lack standardized benchmarks. To bridge this gap, we introduce the Tactile MNIST Benchmark Suite, an open-source, Gymnasium-compatible benchmark specifically designed for active tactile perception tasks, including localization, classification, and volume estimation. Our benchmark suite offers diverse simulation scenarios, from simple toy environments all the way to complex tactile perception tasks using vision-based tactile sensors. Furthermore, we also offer a comprehensive dataset comprising 13,500 synthetic 3D MNIST digit models and 153,600 real-world tactile samples collected from 600 3D printed digits. Using this dataset, we train a CycleGAN for realistic tactile simulation rendering. By providing standardized protocols and reproducible evaluation frameworks, our benchmark suite facilitates systematic progress in the fields of tactile sensing and active perception.
Authors: Dejun Xu, Jijia Chen, Gary G. Yen, Min Jiang
Abstract: Bilevel optimization poses a significant computational challenge due to its nested structure, where each upper-level candidate solution requires solving a corresponding lower-level problem. While evolutionary algorithms (EAs) are effective at navigating such complex landscapes, their high resource demands remain a key bottleneck -- particularly the redundant evaluation of numerous unpromising lower-level tasks. Despite recent advances in multitasking and transfer learning, resource waste persists. To address this issue, we propose a novel resource allocation framework for bilevel EAs that selectively identifies and focuses on promising lower-level tasks. Central to our approach is a contrastive ranking network that learns relational patterns between paired upper- and lower-level solutions online. This knowledge guides a reference-based ranking strategy that prioritizes tasks for optimization and adaptively controls resampling based on estimated population quality. Comprehensive experiments across five state-of-the-art bilevel algorithms show that our framework significantly reduces computational cost while preserving -- or even enhancing -- solution accuracy. This work offers a generalizable strategy to improve the efficiency of bilevel EAs, paving the way for more scalable bilevel optimization.
Authors: Thang D. Pham, Aditya Tanikanti, Murat Ke\c{c}eli
Abstract: Atomistic simulations are essential tools in chemistry and materials science, accelerating the discovery of novel catalysts, energy storage materials, and pharmaceuticals. However, running these simulations remains challenging due to the wide range of computational methods, diverse software ecosystems, and the need for expert knowledge and manual effort for the setup, execution, and validation stages. In this work, we present ChemGraph, an agentic framework powered by artificial intelligence and state-of-the-art simulation tools to streamline and automate computational chemistry and materials science workflows. ChemGraph leverages graph neural network-based foundation models for accurate yet computationally efficient calculations and large language models (LLMs) for natural language understanding, task planning, and scientific reasoning to provide an intuitive and interactive interface. Users can perform tasks such as molecular structure generation, single-point energy, geometry optimization, vibrational analysis, and thermochemistry calculations with methods ranging from tight-binding and machine learning interatomic potentials to density functional theory or wave function theory-based methods. We evaluate ChemGraph across 13 benchmark tasks and demonstrate that smaller LLMs (GPT-4o-mini, Claude-3.5-haiku, Qwen2.5-14B) perform well on simple workflows, while more complex tasks benefit from using larger models like GPT-4o. Importantly, we show that decomposing complex tasks into smaller subtasks through a multi-agent framework enables smaller LLM models to match or exceed GPT-4o's performance in specific scenarios.
Authors: Th\'eo Guyard, C\'edric Herzet, Cl\'ement Elvira
Abstract: This paper presents El0ps, a Python toolbox providing several utilities to handle L0-regularized problems related to applications in machine learning, statistics, and signal processing, among other fields. In contrast to existing toolboxes, El0ps allows users to define custom instances of these problems through a flexible framework, provides a dedicated solver achieving state-of-the-art performance, and offers several built-in machine learning pipelines. Our aim with El0ps is to provide a comprehensive tool which opens new perspectives for the integration of L0-regularized problems in practical applications.
Authors: Giuseppe Arbia, Luca Morandini, Vincenzo Nardelli
Abstract: This paper investigates Large Language Models (LLMs) ability to assess the economic soundness and theoretical consistency of empirical findings in spatial econometrics. We created original and deliberately altered "counterfactual" summaries from 28 published papers (2005-2024), which were evaluated by a diverse set of LLMs. The LLMs provided qualitative assessments and structured binary classifications on variable choice, coefficient plausibility, and publication suitability. The results indicate that while LLMs can expertly assess the coherence of variable choices (with top models like GPT-4o achieving an overall F1 score of 0.87), their performance varies significantly when evaluating deeper aspects such as coefficient plausibility and overall publication suitability. The results further revealed that the choice of LLM, the specific characteristics of the paper and the interaction between these two factors significantly influence the accuracy of the assessment, particularly for nuanced judgments. These findings highlight LLMs' current strengths in assisting with initial, more surface-level checks and their limitations in performing comprehensive, deep economic reasoning, suggesting a potential assistive role in peer review that still necessitates robust human oversight.
Authors: Charalampos Tsirmpas, Stasinos Konstantopoulos, Dimitris Andrikopoulos, Konstantina Kyriakouli, Panagiotis Fatouros
Abstract: Decomposing Electrodermal Activity (EDA) into phasic (short-term, stimulus-linked responses) and tonic (longer-term baseline) components is essential for extracting meaningful emotional and physiological biomarkers. This study presents a comparative analysis of knowledge-driven, statistical, and deep learning-based methods for EDA signal decomposition, with a focus on in-the-wild data collected from wearable devices. In particular, the authors introduce the Feel Transformer, a novel Transformer-based model adapted from the Autoformer architecture, designed to separate phasic and tonic components without explicit supervision. The model leverages pooling and trend-removal mechanisms to enforce physiologically meaningful decompositions. Comparative experiments against methods such as Ledalab, cvxEDA, and conventional detrending show that the Feel Transformer achieves a balance between feature fidelity (SCR frequency, amplitude, and tonic slope) and robustness to noisy, real-world data. The model demonstrates potential for real-time biosignal analysis and future applications in stress prediction, digital mental health interventions, and physiological forecasting.
Authors: Micha{\l} P. Karpowicz
Abstract: This paper explains \textbf{why it is impossible to create large language models that do not hallucinate and what are the trade-offs we should be looking for}. It presents a formal \textbf{impossibility theorem} demonstrating that no inference mechanism can simultaneously satisfy four fundamental properties: \textbf{truthful (non-hallucinatory) generation, semantic information conservation, relevant knowledge revelation, and knowledge-constrained optimality}. By modeling LLM inference as an \textbf{auction of ideas} where neural components compete to contribute to responses, we prove the impossibility using the Green-Laffont theorem. That mathematical framework provides a rigorous foundation for understanding the nature of inference process, with implications for model architecture, training objectives, and evaluation methods.
Authors: Baptiste Chatelier (IETR, INSA Rennes, MERCE-France), Vincent Corlay (MERCE-France), Musa Furkan Keskin (INSA Rennes, IETR), Matthieu Crussi\`ere (INSA Rennes, IETR), Henk Wymeersch (INSA Rennes, IETR), Luc Le Magoarou (INSA Rennes, IETR)
Abstract: The increasing deployment of large antenna arrays at base stations has significantly improved the spatial resolution and localization accuracy of radio-localization methods. However, traditional signal processing techniques struggle in complex radio environments, particularly in scenarios dominated by non line of sight (NLoS) propagation paths, resulting in degraded localization accuracy. Recent developments in machine learning have facilitated the development of machine learning-assisted localization techniques, enhancing localization accuracy in complex radio environments. However, these methods often involve substantial computational complexity during both the training and inference phases. This work extends the well-established fingerprinting-based localization framework by simultaneously reducing its memory requirements and improving its accuracy. Specifically, a model-based neural network is used to learn the location-to-channel mapping, and then serves as a generative neural channel model. This generative model augments the fingerprinting comparison dictionary while reducing the memory requirements. The proposed method outperforms fingerprinting baselines by achieving sub-wavelength localization accuracy, even in NLoS environments. Remarkably, it offers an improvement by several orders of magnitude in localization accuracy, while simultaneously reducing memory requirements by an order of magnitude compared to classical fingerprinting methods.
Authors: Rifat Sadik, Tanvir Rahman, Arpan Bhattacharjee, Bikash Chandra Halder, Ismail Hossain
Abstract: Deep learning models have shown remarkable success in dermatological image analysis, offering potential for automated skin disease diagnosis. Previously, convolutional neural network(CNN) based architectures have achieved immense popularity and success in computer vision (CV) based task like skin image recognition, generation and video analysis. But with the emergence of transformer based models, CV tasks are now are nowadays carrying out using these models. Vision Transformers (ViTs) is such a transformer-based models that have shown success in computer vision. It uses self-attention mechanisms to achieve state-of-the-art performance across various tasks. However, their reliance on global attention mechanisms makes them susceptible to adversarial perturbations. This paper aims to investigate the susceptibility of ViTs for medical images to adversarial watermarking-a method that adds so-called imperceptible perturbations in order to fool models. By generating adversarial watermarks through Projected Gradient Descent (PGD), we examine the transferability of such attacks to CNNs and analyze the performance defense mechanism -- adversarial training. Results indicate that while performance is not compromised for clean images, ViTs certainly become much more vulnerable to adversarial attacks: an accuracy drop of as low as 27.6%. Nevertheless, adversarial training raises it up to 90.0%.
Authors: Pengyi Li, Matvey Skripkin, Alexander Zubrey, Andrey Kuznetsov, Ivan Oseledets
Abstract: Large language models (LLMs) excel at reasoning, yet post-training remains critical for aligning their behavior with task goals. Existing reinforcement learning (RL) methods often depend on costly human annotations or external reward models. We propose Reinforcement Learning via Self-Confidence (RLSC), which uses the model's own confidence as reward signals-eliminating the need for labels, preference models, or reward engineering. Applied to Qwen2.5-Math-7B with only 16 samples per question and 10 or 20 training steps, RLSC improves accuracy by +13.4% on AIME2024, +21.2% on MATH500, +21.7% on Minerva Math, +20.8% on Olympiadbench, and +9.7% on AMC23. RLSC provides a simple, scalable post-training method for inference models, requiring only a small number of samples and unlabelled supervision.
Authors: Hongming Yang, Shi Lin, Jun Shao, Changting Lin, Donghai Zhu, Meng Han, Qinglei Kong
Abstract: Lightweight Large Language Models (LwLLMs) are reduced-parameter, optimized models designed to run efficiently on consumer-grade hardware, offering significant advantages in resource efficiency, cost-effectiveness, and data privacy. However, these models often struggle with limited inference and reasoning capabilities, which restrict their performance on complex tasks and limit their practical applicability. Moreover, existing prompt optimization methods typically rely on extensive manual effort or the meta-cognitive abilities of state-of-the-art LLMs, making them less effective for LwLLMs. To address these challenges, we introduce DeBoP, a new Direct Behavior Optimization Paradigm, original from the Chain-of-Thought (CoT) prompting technique. Unlike CoT Prompting, DeBoP is an automatic optimization method, which focuses on the optimization directly on the behavior of LwLLMs. In particular, DeBoP transforms the optimization of complex prompts into the optimization of discrete, quantifiable execution sequences using a gradient-free Monte Carlo Tree Search. We evaluate DeBoP on seven challenging tasks where state-of-the-art LLMs excel but LwLLMs generally underperform. Experimental results demonstrate that DeBoP significantly outperforms recent prompt optimization methods on most tasks. In particular, DeBoP-optimized LwLLMs surpass GPT-3.5 on most tasks while reducing computational time by approximately 60% compared to other automatic prompt optimization methods.
Authors: Sooyung Choi, Jaehyeok Lee, Xiaoyuan Yi, Jing Yao, Xing Xie, JinYeong Bak
Abstract: The application scope of Large Language Models (LLMs) continues to expand, leading to increasing interest in personalized LLMs that align with human values. However, aligning these models with individual values raises significant safety concerns, as certain values may correlate with harmful information. In this paper, we identify specific safety risks associated with value-aligned LLMs and investigate the psychological principles behind these challenges. Our findings reveal two key insights. (1) Value-aligned LLMs are more prone to harmful behavior compared to non-fine-tuned models and exhibit slightly higher risks in traditional safety evaluations than other fine-tuned models. (2) These safety issues arise because value-aligned LLMs genuinely generate text according to the aligned values, which can amplify harmful outcomes. Using a dataset with detailed safety categories, we find significant correlations between value alignment and safety risks, supported by psychological hypotheses. This study offers insights into the "black box" of value alignment and proposes in-context alignment methods to enhance the safety of value-aligned LLMs.
Authors: Zhi Wen Soi, Chaoyi Zhu, Fouad Abiad, Aditya Shankar, Jeroen M. Galjaard, Huijuan Wang, Lydia Y. Chen
Abstract: Synthetic time series generated by diffusion models enable sharing privacy-sensitive datasets, such as patients' functional MRI records. Key criteria for synthetic data include high data utility and traceability to verify the data source. Recent watermarking methods embed in homogeneous latent spaces, but state-of-the-art time series generators operate in real space, making latent-based watermarking incompatible. This creates the challenge of watermarking directly in real space while handling feature heterogeneity and temporal dependencies. We propose TimeWak, the first watermarking algorithm for multivariate time series diffusion models. To handle temporal dependence and spatial heterogeneity, TimeWak embeds a temporal chained-hashing watermark directly within the real temporal-feature space. The other unique feature is the $\epsilon$-exact inversion, which addresses the non-uniform reconstruction error distribution across features from inverting the diffusion process to detect watermarks. We derive the error bound of inverting multivariate time series and further maintain high watermark detectability. We extensively evaluate TimeWak on its impact on synthetic data quality, watermark detectability, and robustness under various post-editing attacks, against 5 datasets and baselines of different temporal lengths. Our results show that TimeWak achieves improvements of 61.96% in context-FID score, and 8.44% in correlational scores against the state-of-the-art baseline, while remaining consistently detectable.
Authors: Dor Tsur, Carol Xuan Long, Claudio Mayrink Verdun, Hsiang Hsu, Chen-Fu Chen, Haim Permuter, Sajani Vithana, Flavio P. Calmon
Abstract: Large language model (LLM) watermarks enable authentication of text provenance, curb misuse of machine-generated text, and promote trust in AI systems. Current watermarks operate by changing the next-token predictions output by an LLM. The updated (i.e., watermarked) predictions depend on random side information produced, for example, by hashing previously generated tokens. LLM watermarking is particularly challenging in low-entropy generation tasks - such as coding - where next-token predictions are near-deterministic. In this paper, we propose an optimization framework for watermark design. Our goal is to understand how to most effectively use random side information in order to maximize the likelihood of watermark detection and minimize the distortion of generated text. Our analysis informs the design of two new watermarks: HeavyWater and SimplexWater. Both watermarks are tunable, gracefully trading-off between detection accuracy and text distortion. They can also be applied to any LLM and are agnostic to side information generation. We examine the performance of HeavyWater and SimplexWater through several benchmarks, demonstrating that they can achieve high watermark detection accuracy with minimal compromise of text generation quality, particularly in the low-entropy regime. Our theoretical analysis also reveals surprising new connections between LLM watermarking and coding theory. The code implementation can be found in https://github.com/DorTsur/HeavyWater_SimplexWater
Authors: Gabriel Nova, Sander van Cranenburgh, Stephane Hess
Abstract: Discrete choice modelling is a theory-driven modelling framework for understanding and forecasting choice behaviour. To obtain behavioural insights, modellers test several competing model specifications in their attempts to discover the 'true' data generation process. This trial-and-error process requires expertise, is time-consuming, and relies on subjective theoretical assumptions. Although metaheuristics have been proposed to assist choice modellers, they treat model specification as a classic optimisation problem, relying on static strategies, applying predefined rules, and neglecting outcomes from previous estimated models. As a result, current metaheuristics struggle to prioritise promising search regions, adapt exploration dynamically, and transfer knowledge to other modelling tasks. To address these limitations, we introduce a deep reinforcement learning-based framework where an 'agent' specifies models by estimating them and receiving rewards based on goodness-of-fit and parsimony. Results demonstrate the agent dynamically adapts its strategies to identify promising specifications across data generation processes, showing robustness and potential transferability, without prior domain knowledge.
Authors: Ivi Chatzi, Nina Corvelo Benz, Stratis Tsirtsis, Manuel Gomez-Rodriguez
Abstract: State of the art large language models are trained using large amounts of tokens derived from raw text using what is called a tokenizer. Crucially, the tokenizer determines the (token) vocabulary a model will use during inference as well as, in principle, the (token) language. This is because, while the token vocabulary may allow for different tokenizations of a string, the tokenizer always maps the string to only one of these tokenizations--the canonical tokenization. However, multiple lines of empirical evidence suggest that large language models do not always generate canonical token sequences, and this comes with several negative consequences. In this work, we first show that, to generate a canonical token sequence, a model needs to generate (partial) canonical token sequences at each step of the autoregressive generation process underpinning its functioning. Building upon this theoretical result, we introduce canonical sampling, a simple and efficient sampling method that precludes a given model from generating non-canonical token sequences. Further, we also show that, in comparison with standard sampling, the distribution of token sequences generated using canonical sampling is provably closer to the true distribution of token sequences used during training.
Authors: Yanwei Ren, Haotian Zhang, Fuxiang Wu, Jiayan Qiu, Jiaxing Huang, Baosheng Yu, Liu Liu
Abstract: Enhancing large language models by simply scaling up datasets has begun to yield diminishing returns, shifting the spotlight to data quality. Monte Carlo Tree Search (MCTS) has emerged as a powerful technique for generating high-quality chain-of-thought data, yet conventional approaches typically retain only the top-scoring trajectory from the search tree, discarding sibling nodes that often contain valuable partial insights, recurrent error patterns, and alternative reasoning strategies. This unconditional rejection of non-optimal reasoning branches may waste vast amounts of informative data in the whole search tree. We propose SIGMA (Sibling Guided Monte Carlo Augmentation), a novel framework that reintegrates these discarded sibling nodes to refine LLM reasoning. SIGMA forges semantic links among sibling nodes along each search path and applies a two-stage refinement: a critique model identifies overlooked strengths and weaknesses across the sibling set, and a revision model conducts text-based backpropagation to refine the top-scoring trajectory in light of this comparative feedback. By recovering and amplifying the underutilized but valuable signals from non-optimal reasoning branches, SIGMA substantially improves reasoning trajectories. On the challenging MATH benchmark, our SIGMA-tuned 7B model achieves 54.92% accuracy using only 30K samples, outperforming state-of-the-art models trained on 590K samples. This result highlights that our sibling-guided optimization not only significantly reduces data usage but also significantly boosts LLM reasoning.
Authors: Ziqi Yuan, Haoyang Zhang, Yirui Eric Zhou, Apoorve Mohan, I-Hsin Chung, Seetharami Seelam, Jian Huang
Abstract: We present the design and implementation of a new lifetime-aware tensor offloading framework for GPU memory expansion using low-cost PCIe-based solid-state drives (SSDs). Our framework, TERAIO, is developed explicitly for large language model (LLM) training with multiple GPUs and multiple SSDs. Its design is driven by our observation that the active tensors take only a small fraction (1.7% on average) of allocated GPU memory in each LLM training iteration, the inactive tensors are usually large and will not be used for a long period of time, creating ample opportunities for offloading/prefetching tensors to/from slow SSDs without stalling the GPU training process. TERAIO accurately estimates the lifetime (active period of time in GPU memory) of each tensor with the profiling of the first few iterations in the training process. With the tensor lifetime analysis, TERAIO will generate an optimized tensor offloading/prefetching plan and integrate it into the compiled LLM program via PyTorch. TERAIO has a runtime tensor migration engine to execute the offloading/prefetching plan via GPUDirect storage, which allows direct tensor migration between GPUs and SSDs for alleviating the CPU bottleneck and maximizing the SSD bandwidth utilization. In comparison with state-of-the-art studies such as ZeRO-Offload and ZeRO-Infinity, we show that TERAIO improves the training performance of various LLMs by 1.47x on average, and achieves 80.7% of the ideal performance assuming unlimited GPU memory.
Authors: Yao Ni, Song Wen, Piotr Koniusz, Anoop Cherian
Abstract: Fine-tuning Stable Diffusion enables subject-driven image synthesis by adapting the model to generate images containing specific subjects. However, existing fine-tuning methods suffer from two key issues: underfitting, where the model fails to reliably capture subject identity, and overfitting, where it memorizes the subject image and reduces background diversity. To address these challenges, we propose two auxiliary consistency losses for diffusion fine-tuning. First, a prior consistency regularization loss ensures that the predicted diffusion noise for prior (non-subject) images remains consistent with that of the pretrained model, improving fidelity. Second, a subject consistency regularization loss enhances the fine-tuned model's robustness to multiplicative noise modulated latent code, helping to preserve subject identity while improving diversity. Our experimental results demonstrate that incorporating these losses into fine-tuning not only preserves subject identity but also enhances image diversity, outperforming DreamBooth in terms of CLIP scores, background variation, and overall visual quality.
Authors: Manuel Sage, Khalil Al Handawi, Yaoyao Fiona Zhao
Abstract: Power-to-Gas (P2G) technologies gain recognition for enabling the integration of intermittent renewables, such as wind and solar, into electricity grids. However, determining the most cost-effective operation of these systems is complex due to the volatile nature of renewable energy, electricity prices, and loads. Additionally, P2G systems are less efficient in converting and storing energy compared to battery energy storage systems (BESs), and the benefits of converting electricity into gas are not immediately apparent. Deep Reinforcement Learning (DRL) has shown promise in managing the operation of energy systems amidst these uncertainties. Yet, DRL techniques face difficulties with the delayed reward characteristic of P2G system operation. Previous research has mostly focused on short-term studies that look at the energy conversion process, neglecting the long-term storage capabilities of P2G. This study presents a new method by thoroughly examining how DRL can be applied to the economic operation of P2G systems, in combination with BESs and gas turbines, over extended periods. Through three progressively more complex case studies, we assess the performance of DRL algorithms, specifically Deep Q-Networks and Proximal Policy Optimization, and introduce modifications to enhance their effectiveness. These modifications include integrating forecasts, implementing penalties on the reward function, and applying strategic cost calculations, all aimed at addressing the issue of delayed rewards. Our findings indicate that while DRL initially struggles with the complex decision-making required for P2G system operation, the adjustments we propose significantly improve its capability to devise cost-effective operation strategies, thereby unlocking the potential for long-term energy storage in P2G technologies.
Authors: Neil Fendley, Edward W. Staley, Joshua Carney, William Redman, Marie Chau, Nathan Drenkow
Abstract: With the widespread availability of pretrained Large Language Models (LLMs) and their training datasets, concerns about the security risks associated with their usage has increased significantly. One of these security risks is the threat of LLM poisoning attacks where an attacker modifies some part of the LLM training process to cause the LLM to behave in a malicious way. As an emerging area of research, the current frameworks and terminology for LLM poisoning attacks are derived from earlier classification poisoning literature and are not fully equipped for generative LLM settings. We conduct a systematic review of published LLM poisoning attacks to clarify the security implications and address inconsistencies in terminology across the literature. We propose a comprehensive poisoning threat model applicable to categorize a wide range of LLM poisoning attacks. The poisoning threat model includes four poisoning attack specifications that define the logistics and manipulation strategies of an attack as well as six poisoning metrics used to measure key characteristics of an attack. Under our proposed framework, we organize our discussion of published LLM poisoning literature along four critical dimensions of LLM poisoning attacks: concept poisons, stealthy poisons, persistent poisons, and poisons for unique tasks, to better understand the current landscape of security risks.
Authors: Sherman Khoo, Yakun Wang, Song Liu, Mark Beaumont
Abstract: We study the problem of likelihood maximization when the likelihood function is intractable but model simulations are readily available. We propose a sequential, gradient-based optimization method that directly models the Fisher score based on a local score matching technique which uses simulations from a localized region around each parameter iterate. By employing a linear parameterization to the surrogate score model, our technique admits a closed-form, least-squares solution. This approach yields a fast, flexible, and efficient approximation to the Fisher score, effectively smoothing the likelihood objective and mitigating the challenges posed by complex likelihood landscapes. We provide theoretical guarantees for our score estimator, including bounds on the bias introduced by the smoothing. Empirical results on a range of synthetic and real-world problems demonstrate the superior performance of our method compared to existing benchmarks.
Authors: Antonio Pariente, Ignacio Hounie, Santiago Segarra, Alejandro Ribeiro
Abstract: Despite the ubiquity of vector search applications, prevailing search algorithms overlook the metric structure of vector embeddings, treating it as a constraint rather than exploiting its underlying properties. In this paper, we demonstrate that in $q$-metric spaces, metric trees can leverage a stronger version of the triangle inequality to reduce comparisons for exact search. Notably, as $q$ approaches infinity, the search complexity becomes logarithmic. Therefore, we propose a novel projection method that embeds vector datasets with arbitrary dissimilarity measures into $q$-metric spaces while preserving the nearest neighbor. We propose to learn an approximation of this projection to efficiently transform query points to a space where euclidean distances satisfy the desired properties. Our experimental results with text and image vector embeddings show that learning $q$-metric approximations enables classic metric tree algorithms -- which typically underperform with high-dimensional data -- to achieve competitive performance against state-of-the-art search methods.
Authors: Thushari Hapuarachchi, Long Dang, Kaiqi Xiong
Abstract: Deep Neural Networks (DNNs) are widely used for traffic sign recognition because they can automatically extract high-level features from images. These DNNs are trained on large-scale datasets obtained from unknown sources. Therefore, it is important to ensure that the models remain secure and are not compromised or poisoned during training. In this paper, we investigate the robustness of DNNs trained for traffic sign recognition. First, we perform the error-minimizing attacks on DNNs used for traffic sign recognition by adding imperceptible perturbations on training data. Then, we propose a data augmentation-based training method to mitigate the error-minimizing attacks. The proposed training method utilizes nonlinear transformations to disrupt the perturbations and improve the model robustness. We experiment with two well-known traffic sign datasets to demonstrate the severity of the attack and the effectiveness of our mitigation scheme. The error-minimizing attacks reduce the prediction accuracy of the DNNs from 99.90% to 10.6%. However, our mitigation scheme successfully restores the prediction accuracy to 96.05%. Moreover, our approach outperforms adversarial training in mitigating the error-minimizing attacks. Furthermore, we propose a detection model capable of identifying poisoned data even when the perturbations are imperceptible to human inspection. Our detection model achieves a success rate of over 99% in identifying the attack. This research highlights the need to employ advanced training methods for DNNs in traffic sign recognition systems to mitigate the effects of data poisoning attacks.
Authors: Yijia Shao, Humishka Zope, Yucheng Jiang, Jiaxin Pei, David Nguyen, Erik Brynjolfsson, Diyi Yang
Abstract: The rapid rise of compound AI systems (a.k.a., AI agents) is reshaping the labor market, raising concerns about job displacement, diminished human agency, and overreliance on automation. Yet, we lack a systematic understanding of the evolving landscape. In this paper, we address this gap by introducing a novel auditing framework to assess which occupational tasks workers want AI agents to automate or augment, and how those desires align with the current technological capabilities. Our framework features an audio-enhanced mini-interview to capture nuanced worker desires and introduces the Human Agency Scale (HAS) as a shared language to quantify the preferred level of human involvement. Using this framework, we construct the WORKBank database, building on the U.S. Department of Labor's O*NET database, to capture preferences from 1,500 domain workers and capability assessments from AI experts across over 844 tasks spanning 104 occupations. Jointly considering the desire and technological capability divides tasks in WORKBank into four zones: Automation "Green Light" Zone, Automation "Red Light" Zone, R&D Opportunity Zone, Low Priority Zone. This highlights critical mismatches and opportunities for AI agent development. Moving beyond a simple automate-or-not dichotomy, our results reveal diverse HAS profiles across occupations, reflecting heterogeneous expectations for human involvement. Moreover, our study offers early signals of how AI agent integration may reshape the core human competencies, shifting from information-focused skills to interpersonal ones. These findings underscore the importance of aligning AI agent development with human desires and preparing workers for evolving workplace dynamics.
Authors: Daniel Leite, Igor \v{S}krjanc, Fernando Gomide
Abstract: Evolving fuzzy systems build and adapt fuzzy models - such as predictors and controllers - by incrementally updating their rule-base structure from data streams. On the occasion of the 60-year anniversary of fuzzy set theory, commemorated during the Fuzz-IEEE 2025 event, this brief paper revisits the historical development and core contributions of classical fuzzy and adaptive modeling and control frameworks. It then highlights the emergence and significance of evolving intelligent systems in fuzzy modeling and control, emphasizing their advantages in handling nonstationary environments. Key challenges and future directions are discussed, including safety, interpretability, and principled structural evolution.
Authors: Armin Sarabi, Manish Karir, Mingyan Liu
Abstract: In this paper we present a study on using novel data types to perform cyber risk quantification by estimating the likelihood of a data breach. We demonstrate that it is feasible to build a highly accurate cyber risk assessment model using public and readily available technology signatures obtained from crawling an organization's website. This approach overcomes the limitations of previous similar approaches that relied on large-scale IP address based scanning data, which suffers from incomplete/missing IP address mappings as well as the lack of such data for large numbers of small and medium-sized organizations (SMEs). In comparison to scan data, technology digital signature data is more readily available for millions of SMEs. Our study shows that there is a strong relationship between these technology signatures and an organization's cybersecurity posture. In cross-validating our model using different cyber incident datasets, we also highlight the key differences between ransomware attack victims and the larger population of cyber incident and data breach victims.
Authors: Charles Goddard, Fernando Fernandes Neto
Abstract: We present a training-free method to transplant tokenizers in pretrained large language models (LLMs) by reconstructing unseen token embeddings via Orthogonal Matching Pursuit (OMP). Specifically, we approximate each out-of-vocabulary token as a sparse linear combination of shared tokens, in two phases: first, compute each new token's representation in the donor embedding space with a small dictionary of shared anchor tokens, then transfer these same sparse coefficients back into the base model's embedding space. On two challenging cross-tokenizer tasks--Llama$\to$Mistral NeMo (12B) and Qwen$\to$Llama (1B)--we show that OMP achieves best zero-shot preservation of the base model's performance across multiple benchmarks, while other zero-shot approaches degrade significantly. Compared to baselines (zero-init, mean-init, and existing approaches like WECHSEL, FOCUS, ZETT), OMP consistently achieves the best overall performance, effectively bridging large tokenizer discrepancies without gradient updates. Our analysis further identifies mismatched numerical tokenization schemes as a critical challenge for preserving mathematical reasoning capabilities. This technique enables direct reuse of pretrained model weights with new tokenizers, facilitating cross-tokenizer knowledge distillation, speculative decoding, ensembling, merging, and domain-specific vocabulary adaptations. We integrate our method into the open-source mergekit-tokensurgeon tool for post hoc vocabulary realignment.
Authors: Alan Chen, Jack Merullo, Alessandro Stolfo, Ellie Pavlick
Abstract: In this work, we demonstrate that affine mappings between residual streams of language models is a cheap way to effectively transfer represented features between models. We apply this technique to transfer the weights of Sparse Autoencoders (SAEs) between models of different sizes to compare their representations. We find that small and large models learn highly similar representation spaces, which motivates training expensive components like SAEs on a smaller model and transferring to a larger model at a FLOPs savings. For example, using a small-to-large transferred SAE as initialization can lead to 50% cheaper training runs when training SAEs on larger models. Next, we show that transferred probes and steering vectors can effectively recover ground truth performance. Finally, we dive deeper into feature-level transferability, finding that semantic and structural features transfer noticeably differently while specific classes of functional features have their roles faithfully mapped. Overall, our findings illustrate similarities and differences in the linear representation spaces of small and large models and demonstrate a method for improving the training efficiency of SAEs.
Authors: Arefe Boushehrian, Amir Najafi
Abstract: Learning distribution families over $\mathbb{R}^d$ is a fundamental problem in unsupervised learning and statistics. A central question in this setting is whether a given family of distributions possesses sufficient structure to be (at least) information-theoretically learnable and, if so, to characterize its sample complexity. In 2018, Ashtiani et al. reframed \emph{sample compressibility}, originally due to Littlestone and Warmuth (1986), as a structural property of distribution classes, proving that it guarantees PAC-learnability. This discovery subsequently enabled a series of recent advancements in deriving nearly tight sample complexity bounds for various high-dimensional open problems. It has been further conjectured that the converse also holds: every learnable class admits a tight sample compression scheme. In this work, we establish that sample compressible families remain learnable even from perturbed samples, subject to a set of necessary and sufficient conditions. We analyze two models of data perturbation: (i) an additive independent noise model, and (ii) an adversarial corruption model, where an adversary manipulates a limited subset of the samples unknown to the learner. Our results are general and rely on as minimal assumptions as possible. We develop a perturbation-quantization framework that interfaces naturally with the compression scheme and leads to sample complexity bounds that scale gracefully with the noise level and corruption budget. As concrete applications, we establish new sample complexity bounds for learning finite mixtures of high-dimensional uniform distributions under both noise and adversarial perturbations, as well as for learning Gaussian mixture models from adversarially corrupted samples, resolving two open problems in the literature.
Authors: Dangxing Chen
Abstract: In recent years, machine learning models have achieved great success at the expense of highly complex black-box structures. By using axiomatic attribution methods, we can fairly allocate the contributions of each feature, thus allowing us to interpret the model predictions. In high-risk sectors such as finance, risk is just as important as mean predictions. Throughout this work, we address the following risk attribution problem: how to fairly allocate the risk given a model with data? We demonstrate with analysis and empirical examples that risk can be well allocated by extending the Shapley value framework.
Authors: Yu-Hsuan Ho, Ali Mostafavi
Abstract: Most post-disaster damage classifiers succeed only when destructive forces leave clear spectral or structural signatures -- conditions rarely present after inundation. Consequently, existing models perform poorly at identifying flood-related building damages. The model presented in this study, Flood-DamageSense, addresses this gap as the first deep-learning framework purpose-built for building-level flood-damage assessment. The architecture fuses pre- and post-event SAR/InSAR scenes with very-high-resolution optical basemaps and an inherent flood-risk layer that encodes long-term exposure probabilities, guiding the network toward plausibly affected structures even when compositional change is minimal. A multimodal Mamba backbone with a semi-Siamese encoder and task-specific decoders jointly predicts (1) graded building-damage states, (2) floodwater extent, and (3) building footprints. Training and evaluation on Hurricane Harvey (2017) imagery from Harris County, Texas -- supported by insurance-derived property-damage extents -- show a mean F1 improvement of up to 19 percentage points over state-of-the-art baselines, with the largest gains in the frequently misclassified "minor" and "moderate" damage categories. Ablation studies identify the inherent-risk feature as the single most significant contributor to this performance boost. An end-to-end post-processing pipeline converts pixel-level outputs to actionable, building-scale damage maps within minutes of image acquisition. By combining risk-aware modeling with SAR's all-weather capability, Flood-DamageSense delivers faster, finer-grained, and more reliable flood-damage intelligence to support post-disaster decision-making and resource allocation.
Authors: Radha Kodali, Venkata Rao Dhulipalla, Venkata Siva Kishor Tatavarty, Madhavi Nadakuditi, Bharadwaj Thiruveedhula, Suryanarayana Gunnam, Durga Prasad Bavirisetti
Abstract: Infertility has a considerable impact on individuals' quality of life, affecting them socially and psychologically, with projections indicating a rise in the upcoming years. In vitro fertilization (IVF) emerges as one of the primary techniques within economically developed nations, employed to address the rising problem of low fertility. Expert embryologists conventionally grade embryos by reviewing blastocyst images to select the most optimal for transfer, yet this process is time-consuming and lacks efficiency. Blastocyst images provide a valuable resource for assessing embryo viability. In this study, we introduce an explainable artificial intelligence (XAI) framework for classifying embryos, employing a fusion of convolutional neural network (CNN) and long short-term memory (LSTM) architecture, referred to as CNN-LSTM. Utilizing deep learning, our model achieves high accuracy in embryo classification while maintaining interpretability through XAI.
Authors: Yitao Liu, Chenglei Si, Karthik Narasimhan, Shunyu Yao
Abstract: Large language model (LLM) agents have been applied to sequential decision-making tasks such as web navigation, but without any environment-specific experiences, they often fail in these complex tasks. Moreover, current LLM agents are not designed to continually learn from past experiences during inference time, which could be crucial for them to gain these environment-specific experiences. To address this, we propose Contextual Experience Replay (CER), a training-free framework to enable efficient self-improvement for language agents in their context window. Specifically, CER accumulates and synthesizes past experiences into a dynamic memory buffer. These experiences encompass environment dynamics and common decision-making patterns, allowing the agents to retrieve and augment themselves with relevant knowledge in new tasks, enhancing their adaptability in complex environments. We evaluate CER on the challenging WebArena and VisualWebArena benchmarks. On VisualWebArena, CER achieves a competitive performance of 31.9%. On WebArena, CER also gets a competitive average success rate of 36.7%, relatively improving the success rate of the GPT-4o agent baseline by 51.0%. We also conduct a comprehensive analysis on it to prove its efficiency, validity and understand it better.
Authors: Omar Mashaal, Hatem Abou-Zeid
Abstract: Foundational models have shown remarkable potential in natural language processing and computer vision, yet remain in their infancy in wireless communications. While a few efforts have explored image-based modalities such as channel state information (CSI) and frequency spectrograms, foundational models that operate directly on raw IQ data remain largely unexplored. This paper presents, IQFM, the first I/Q signal foundational model for wireless communications. IQFM supporting diverse tasks: modulation classification, angle-of-arrival (AoA), beam prediction, and RF fingerprinting, without heavy preprocessing or handcrafted features. We also introduce a task-aware augmentation strategy that categorizes transformations into core augmentations, such as cyclic time shifting, and task-specific augmentations. This strategy forms the basis for structured, task-dependent representation learning within a contrastive self-supervised learning (SSL) framework. Using this strategy, the lightweight encoder, pre-trained via SSL on over-the-air multi-antenna IQ data, achieves up to 99.67% and 65.45% accuracy on modulation and AoA classification, respectively, using only one labeled sample per class, outperforming supervised baselines by up to 7x and 145x. The model also generalizes to out-of-distribution tasks; when adapted to new tasks using only 500 samples per class and minimal parameter updates via LoRA, the same frozen encoder achieves 94.15% on beam prediction (vs. 89.53% supervised), 50.00% on RML2016a modulation classification (vs. 49.30%), and 96.05% on RF fingerprinting (vs. 96.64%). These results demonstrate the potential of raw IQ-based foundational models as efficient, reusable encoders for multi-task learning in AI-native 6G systems.
Authors: Guillaume Levy, Cedric Colas, Pierre-Yves Oudeyer, Thomas Carta, Clement Romac
Abstract: Large Language Models (LLMs) possess general world knowledge but often struggle to generate precise predictions in structured, domain-specific contexts such as simulations. These limitations arise from their inability to ground their broad, unstructured understanding in specific environments. To address this, we present WorldLLM, a framework that enhances LLM-based world modeling by combining Bayesian inference and autonomous active exploration with reinforcement learning. WorldLLM leverages the in-context learning abilities of LLMs to guide an LLM-based world model's predictions using natural language hypotheses given in its prompt. These hypotheses are iteratively refined through a Bayesian inference framework that leverages a second LLM as the proposal distribution given collected evidence. This evidence is collected using a curiosity-driven reinforcement learning policy that explores the environment to find transitions with a low log-likelihood under our LLM-based predictive model using the current hypotheses. By alternating between refining hypotheses and collecting new evidence, our framework autonomously drives continual improvement of the predictions. Our experiments demonstrate the effectiveness of WorldLLM in a textual game environment that requires agents to manipulate and combine objects. The framework not only enhances predictive accuracy, but also generates human-interpretable theories of environment dynamics.
Authors: Andrew Cropper, Filipe Gouveia, David M. Cerna
Abstract: Inductive logic programming (ILP) is a form of logical machine learning. The goal is to search a hypothesis space for a hypothesis that generalises training examples and background knowledge. We introduce an approach that 'shrinks' the hypothesis space before an ILP system searches it. Our approach uses background knowledge to find rules that cannot be in an optimal hypothesis regardless of the training examples. For instance, our approach discovers relationships such as "even numbers cannot be odd" and "prime numbers greater than 2 are odd". It then removes violating rules from the hypothesis space. We implement our approach using answer set programming and use it to shrink the hypothesis space of a constraint-based ILP system. Our experiments on multiple domains, including visual reasoning and game playing, show that our approach can substantially reduce learning times whilst maintaining predictive accuracies. For instance, given just 10 seconds of preprocessing time, our approach can reduce learning times from over 10 hours to only 2 seconds.
Authors: Zofia Rudnicka, Janusz Szczepanski, Agnieszka Pregowska
Abstract: Training of Spiking Neural Networks (SNN) is challenging due to their unique properties, including temporal dynamics, non-differentiability of spike events, and sparse event-driven activations. In this paper, we widely consider the influence of the type of chosen learning algorithm, including bioinspired learning rules on the accuracy of classification. We proposed a bioinspired classifier based on the combination of SNN and Lempel-Ziv complexity (LZC). This approach synergizes the strengths of SNNs in temporal precision and biological realism with LZC's structural complexity analysis, facilitating efficient and interpretable classification of spatiotemporal neural data. It turned out that the classic backpropagation algorithm achieves excellent classification accuracy, but at extremely high computational cost, which makes it impractical for real-time applications. Biologically inspired learning algorithms such as tempotron and Spikprop provide increased computational efficiency while maintaining competitive classification performance, making them suitable for time-sensitive tasks. The results obtained indicate that the selection of the most appropriate learning algorithm depends on the trade-off between classification accuracy and computational cost as well as application constraints.
Authors: Emet Behrendt, Shing Wai Pun, Prashant J. Nair
Abstract: Branch prediction is key to the performance of out-of-order processors. While the CBP-2016 winner TAGE-SC-L combines geometric-history tables, a statistical corrector, and a loop predictor, over half of its remaining mispredictions stem from a small set of hard-to-predict (H2P) branches. These branches occur under diverse global histories, causing repeated thrashing in TAGE and eviction before usefulness counters can mature. Prior work shows that simply enlarging the tables offers only marginal improvement. We augment a 159 KB TAGE-SC-L predictor with a 28 KB H2P-targeted subsystem called the Bullseye predictor. It identifies problematic PCs using a set-associative H2P Identification Table (HIT) and steers them to one of two branch-specific perceptrons, one indexed by hashed local history and the other by folded global history. A short trial phase tracks head-to-head accuracy in an H2P cache. A branch becomes perceptron-resident only if the perceptron's sustained accuracy and output magnitude exceed dynamic thresholds, after which TAGE updates for that PC are suppressed to reduce pollution. The HIT, cache, and perceptron operate fully in parallel with TAGE-SC-L, providing higher fidelity on the H2P tail. This achieves an average MPKI of 3.4045 and CycWpPKI of 145.09.
Authors: Longlin Yu, Jiajun Zha, Tong Yang, Tianyu Xie, Xiangyu Zhang, S. -H. Gary Chan, Cheng Zhang
Abstract: Semi-implicit distributions have shown great promise in variational inference and generative modeling. Hierarchical semi-implicit models, which stack multiple semi-implicit layers, enhance the expressiveness of semi-implicit distributions and can be used to accelerate diffusion models given pretrained score networks. However, their sequential training often suffers from slow convergence. In this paper, we introduce CoSIM, a continuous semi-implicit model that extends hierarchical semi-implicit models into a continuous framework. By incorporating a continuous transition kernel, CoSIM enables efficient, simulation-free training. Furthermore, we show that CoSIM achieves consistency with a carefully designed transition kernel, offering a novel approach for multistep distillation of generative models at the distributional level. Extensive experiments on image generation demonstrate that CoSIM performs on par or better than existing diffusion model acceleration methods, achieving superior performance on FD-DINOv2.
Authors: Lennart Bastian, Mohammad Rashed, Nassir Navab, Tolga Birdal
Abstract: Tracking and forecasting the rotation of objects is fundamental in computer vision and robotics, yet SO(3) extrapolation remains challenging as (1) sensor observations can be noisy and sparse, (2) motion patterns can be governed by complex dynamics, and (3) application settings can demand long-term forecasting. This work proposes modeling continuous-time rotational object dynamics on $SO(3)$ using Neural Controlled Differential Equations guided by Savitzky-Golay paths. Unlike existing methods that rely on simplified motion assumptions, our method learns a general latent dynamical system of the underlying object trajectory while respecting the geometric structure of rotations. Experimental results on real-world data demonstrate compelling forecasting capabilities compared to existing approaches.
Authors: Subhendu Khatuya, Shashwat Naidu, Saptarshi Ghosh, Pawan Goyal, Niloy Ganguly
Abstract: The explosion of textual data has made manual document classification increasingly challenging. To address this, we introduce a robust, efficient domain-agnostic generative model framework for multi-label text classification. Instead of treating labels as mere atomic symbols, our approach utilizes predefined label descriptions and is trained to generate these descriptions based on the input text. During inference, the generated descriptions are matched to the pre-defined labels using a finetuned sentence transformer. We integrate this with a dual-objective loss function, combining cross-entropy loss and cosine similarity of the generated sentences with the predefined target descriptions, ensuring both semantic alignment and accuracy. Our proposed model LAGAMC stands out for its parameter efficiency and versatility across diverse datasets, making it well-suited for practical applications. We demonstrate the effectiveness of our proposed model by achieving new state-of-the-art performances across all evaluated datasets, surpassing several strong baselines. We achieve improvements of 13.94% in Micro-F1 and 24.85% in Macro-F1 compared to the closest baseline across all datasets.
Authors: Haoran Wu, Ce Guo, Wayne Luk, Robert Mullins
Abstract: Bayesian Optimization (BO) has shown promise in tuning processor design parameters. However, standard BO does not support constraints involving categorical parameters such as types of branch predictors and division circuits. In addition, optimization time of BO grows with processor complexity, which becomes increasingly significant especially for FPGA-based soft processors. This paper introduces ASPO, an approach that leverages disjunctive form to enable BO to handle constraints involving categorical parameters. Unlike existing methods that directly apply standard BO, the proposed ASPO method, for the first time, customizes the mathematical mechanism of BO to address challenges faced by soft-processor designs on FPGAs. Specifically, ASPO supports categorical parameters using a novel customized BO covariance kernel. It also accelerates the design evaluation procedure by penalizing the BO acquisition function with potential evaluation time and by reusing FPGA synthesis checkpoints from previously evaluated configurations. ASPO targets three soft processors: RocketChip, BOOM, and EL2 VeeR. The approach is evaluated based on seven RISC-V benchmarks. Results show that ASPO can reduce execution time for the ``multiply'' benchmark on the BOOM processor by up to 35\% compared to the default configuration. Furthermore, it reduces design time for the BOOM processor by up to 74\% compared to Boomerang, a state-of-the-art hardware-oriented BO approach.
Authors: Simon P. von der Maase
Abstract: I present a novel approach to estimating the temporal and spatial patterns of violent conflict. I show how we can use highly temporally and spatially disaggregated data on conflict events in tandem with Gaussian processes to estimate temporospatial conflict trends. These trends can be studied to gain insight into conflict traps, diffusion and tempo-spatial conflict exposure in general; they can also be used to control for such phenomenons given other estimation tasks; lastly, the approach allow us to extrapolate the estimated tempo-spatial conflict patterns into future temporal units, thus facilitating powerful, stat-of-the-art, conflict forecasts. Importantly, these results are achieved via a relatively parsimonious framework using only one data source: past conflict patterns.
Authors: Zelin He, Sarah Alnegheimish, Matthew Reimherr
Abstract: Time-series anomaly detection (TSAD) has played a vital role in a variety of fields, including healthcare, finance, and industrial monitoring. Prior methods, which mainly focus on training domain-specific models on numerical data, lack the visual-temporal reasoning capacity that human experts have to identify contextual anomalies. To fill this gap, we explore a solution based on vision language models (VLMs). Recent studies have shown the ability of VLMs for visual reasoning tasks, yet their direct application to time series has fallen short on both accuracy and efficiency. To harness the power of VLMs for TSAD, we propose a two-stage solution, with (1) ViT4TS, a vision-screening stage built on a relatively lightweight pretrained vision encoder, which leverages 2-D time-series representations to accurately localize candidate anomalies; (2) VLM4TS, a VLM-based stage that integrates global temporal context and VLM reasoning capacity to refine the detection upon the candidates provided by ViT4TS. We show that without any time-series training, VLM4TS outperforms time-series pretrained and from-scratch baselines in most cases, yielding a 24.6 percent improvement in F1-max score over the best baseline. Moreover, VLM4TS also consistently outperforms existing language-model-based TSAD methods and is on average 36 times more efficient in token usage.
Authors: Fahad Mostafa
Abstract: Long Short-Term Memory (LSTM) neural network models have become the cornerstone for sequential data modeling in numerous applications, ranging from natural language processing to time series forecasting. Despite their success, the problem of model selection, including hyperparameter tuning, architecture specification, and regularization choice remains largely heuristic and computationally expensive. In this paper, we propose a unified statistical framework for systematic model selection in LSTM networks. Our framework extends classical model selection ideas, such as information criteria and shrinkage estimation, to sequential neural networks. We define penalized likelihoods adapted to temporal structures, propose a generalized threshold approach for hidden state dynamics, and provide efficient estimation strategies using variational Bayes and approximate marginal likelihood methods. Several biomedical data centric examples demonstrate the flexibility and improved performance of the proposed framework.
Authors: Chenguang Huang, Oier Mees, Andy Zeng, Wolfram Burgard
Abstract: Grounding language to a navigating agent's observations can leverage pretrained multimodal foundation models to match perceptions to object or event descriptions. However, previous approaches remain disconnected from environment mapping, lack the spatial precision of geometric maps, or neglect additional modality information beyond vision. To address this, we propose multimodal spatial language maps as a spatial map representation that fuses pretrained multimodal features with a 3D reconstruction of the environment. We build these maps autonomously using standard exploration. We present two instances of our maps, which are visual-language maps (VLMaps) and their extension to audio-visual-language maps (AVLMaps) obtained by adding audio information. When combined with large language models (LLMs), VLMaps can (i) translate natural language commands into open-vocabulary spatial goals (e.g., "in between the sofa and TV") directly localized in the map, and (ii) be shared across different robot embodiments to generate tailored obstacle maps on demand. Building upon the capabilities above, AVLMaps extend VLMaps by introducing a unified 3D spatial representation integrating audio, visual, and language cues through the fusion of features from pretrained multimodal foundation models. This enables robots to ground multimodal goal queries (e.g., text, images, or audio snippets) to spatial locations for navigation. Additionally, the incorporation of diverse sensory inputs significantly enhances goal disambiguation in ambiguous environments. Experiments in simulation and real-world settings demonstrate that our multimodal spatial language maps enable zero-shot spatial and multimodal goal navigation and improve recall by 50% in ambiguous scenarios. These capabilities extend to mobile robots and tabletop manipulators, supporting navigation and interaction guided by visual, audio, and spatial cues.
Authors: Reese Kneeland, Paul S. Scotti, Ghislain St-Yves, Jesse Breedlove, Kendrick Kay, Thomas Naselaris
Abstract: We release NSD-Imagery, a benchmark dataset of human fMRI activity paired with mental images, to complement the existing Natural Scenes Dataset (NSD), a large-scale dataset of fMRI activity paired with seen images that enabled unprecedented improvements in fMRI-to-image reconstruction efforts. Recent models trained on NSD have been evaluated only on seen image reconstruction. Using NSD-Imagery, it is possible to assess how well these models perform on mental image reconstruction. This is a challenging generalization requirement because mental images are encoded in human brain activity with relatively lower signal-to-noise and spatial resolution; however, generalization from seen to mental imagery is critical for real-world applications in medical domains and brain-computer interfaces, where the desired information is always internally generated. We provide benchmarks for a suite of recent NSD-trained open-source visual decoding models (MindEye1, MindEye2, Brain Diffuser, iCNN, Takagi et al.) on NSD-Imagery, and show that the performance of decoding methods on mental images is largely decoupled from performance on vision reconstruction. We further demonstrate that architectural choices significantly impact cross-decoding performance: models employing simple linear decoding architectures and multimodal feature decoding generalize better to mental imagery, while complex architectures tend to overfit visual training data. Our findings indicate that mental imagery datasets are critical for the development of practical applications, and establish NSD-Imagery as a useful resource for better aligning visual decoding methods with this goal.
Authors: Odin Zhang, Haitao Lin, Xujun Zhang, Xiaorui Wang, Zhenxing Wu, Qing Ye, Weibo Zhao, Jike Wang, Kejun Ying, Yu Kang, Chang-yu Hsieh, Tingjun Hou
Abstract: Graph neural networks (GNNs), as topology/structure-aware models within deep learning, have emerged as powerful tools for AI-aided drug discovery (AIDD). By directly operating on molecular graphs, GNNs offer an intuitive and expressive framework for learning the complex topological and geometric features of drug-like molecules, cementing their role in modern molecular modeling. This review provides a comprehensive overview of the methodological foundations and representative applications of GNNs in drug discovery, spanning tasks such as molecular property prediction, virtual screening, molecular generation, biomedical knowledge graph construction, and synthesis planning. Particular attention is given to recent methodological advances, including geometric GNNs, interpretable models, uncertainty quantification, scalable graph architectures, and graph generative frameworks. We also discuss how these models integrate with modern deep learning approaches, such as self-supervised learning, multi-task learning, meta-learning and pre-training. Throughout this review, we highlight the practical challenges and methodological bottlenecks encountered when applying GNNs to real-world drug discovery pipelines, and conclude with a discussion on future directions.
Authors: Parshin Shojaee, Iman Mirzadeh, Keivan Alizadeh, Maxwell Horton, Samy Bengio, Mehrdad Farajtabar
Abstract: Recent generations of language models have introduced Large Reasoning Models (LRMs) that generate detailed thinking processes before providing answers. While these models demonstrate improved performance on reasoning benchmarks, their fundamental capabilities, scaling properties, and limitations remain insufficiently understood. Current evaluations primarily focus on established math and coding benchmarks, emphasizing final answer accuracy. However, this evaluation paradigm often suffers from contamination and does not provide insights into the reasoning traces. In this work, we systematically investigate these gaps with the help of controllable puzzle environments that allow precise manipulation of complexity while maintaining consistent logical structures. This setup enables the analysis of not only final answers but also the internal reasoning traces, offering insights into how LRMs think. Through extensive experiments, we show that LRMs face a complete accuracy collapse beyond certain complexities. Moreover, they exhibit a counterintuitive scaling limit: their reasoning effort increases with problem complexity up to a point, then declines despite having remaining token budget. By comparing LRMs with their standard LLM counterparts under same inference compute, we identify three performance regimes: (1) low-complexity tasks where standard models outperform LRMs, (2) medium-complexity tasks where LRMs demonstrates advantage, and (3) high-complexity tasks where both models face complete collapse. We found that LRMs have limitations in exact computation: they fail to use explicit algorithms and reason inconsistently across scales. We also investigate the reasoning traces in more depth, studying the patterns of explored solutions and analyzing the models' computational behavior, shedding light on their strengths, limitations, and raising questions about their reasoning capabilities.
Authors: Mohammad Farzanullah, Han Zhang, Akram Bin Sediq, Ali Afana, Melike Erol-Kantarci
Abstract: Cell-free Integrated Sensing and Communication (ISAC) aims to revolutionize 6th Generation (6G) networks. By combining distributed access points with ISAC capabilities, it boosts spectral efficiency, situational awareness, and communication reliability. Channel estimation is a critical step in cell-free ISAC systems to ensure reliable communication, but its performance is usually limited by challenges such as pilot contamination and noisy channel estimates. This paper presents a novel framework leveraging sensing information as a key input within a Conditional Denoising Diffusion Model (CDDM). In this framework, we integrate CDDM with a Multimodal Transformer (MMT) to enhance channel estimation in ISAC-enabled cell-free systems. The MMT encoder effectively captures inter-modal relationships between sensing and location data, enabling the CDDM to iteratively denoise and refine channel estimates. Simulation results demonstrate that the proposed approach achieves significant performance gains. As compared with Least Squares (LS) and Minimum Mean Squared Error (MMSE) estimators, the proposed model achieves normalized mean squared error (NMSE) improvements of 8 dB and 9 dB, respectively. Moreover, we achieve a 27.8% NMSE improvement compared to the traditional denoising diffusion model (TDDM), which does not incorporate sensing channel information. Additionally, the model exhibits higher robustness against pilot contamination and maintains high accuracy under challenging conditions, such as low signal-to-noise ratios (SNRs). According to the simulation results, the model performs well for users near sensing targets by leveraging the correlation between sensing and communication channels.
Authors: Mellon M. Zhang, Glen Chou, Saibal Mukhopadhyay
Abstract: Accurate and efficient object detection is essential for autonomous vehicles, where real-time perception requires low latency and high throughput. LiDAR sensors provide robust depth information, but conventional methods process full 360{\deg} scans in a single pass, introducing significant delay. Streaming approaches address this by sequentially processing partial scans in the native polar coordinate system, yet they rely on translation-invariant convolutions that are misaligned with polar geometry -- resulting in degraded performance or requiring complex distortion mitigation. Recent Mamba-based state space models (SSMs) have shown promise for LiDAR perception, but only in the full-scan setting, relying on geometric serialization and positional embeddings that are memory-intensive and ill-suited to streaming. We propose Polar Hierarchical Mamba (PHiM), a novel SSM architecture designed for polar-coordinate streaming LiDAR. PHiM uses local bidirectional Mamba blocks for intra-sector spatial encoding and a global forward Mamba for inter-sector temporal modeling, replacing convolutions and positional encodings with distortion-aware, dimensionally-decomposed operations. PHiM sets a new state-of-the-art among streaming detectors on the Waymo Open Dataset, outperforming the previous best by 10\% and matching full-scan baselines at twice the throughput. Code will be available at https://github.com/meilongzhang/Polar-Hierarchical-Mamba .
URLs: https://github.com/meilongzhang/Polar-Hierarchical-Mamba
Authors: Daniel Lawand (University of S\~ao Paulo), Lucas Quaresma (University of S\~ao Paulo), Roberto Bolgheroni (University of S\~ao Paulo), Alfredo Goldman (University of S\~ao Paulo), Renato Cordeiro Ferreira (University of S\~ao Paulo, Jheronimus Academy of Data Science, Technical University of Eindhoven, Tilburg University)
Abstract: Deploying a Machine Learning (ML) training pipeline into production requires robust software engineering practices. This differs significantly from experimental workflows. This experience report investigates this challenge in SPIRA, a project whose goal is to create an ML-Enabled System (MLES) to pre-diagnose insufficiency respiratory via speech analysis. The first version of SPIRA's training pipeline lacked critical software quality attributes. This paper presents an overview of the MLES, then compares three versions of the architecture of the Continuous Training subsystem, which evolved from a Big Ball of Mud, to a Modular Monolith, towards Microservices. By adopting different design principles and patterns to enhance its maintainability, robustness, and extensibility. In this way, the paper seeks to offer insights for both ML Engineers tasked to productionize ML training pipelines and Data Scientists seeking to adopt MLOps practices.
Authors: Subhojyoti Mukherjee, Viet Dac Lai, Raghavendra Addanki, Ryan Rossi, Seunghyun Yoon, Trung Bui, Anup Rao, Jayakumar Subramanian, Branislav Kveton
Abstract: Question answering (QA) agents automatically answer questions posed in natural language. In this work, we learn to ask clarifying questions in QA agents. The key idea in our method is to simulate conversations that contain clarifying questions and learn from them using reinforcement learning (RL). To make RL practical, we propose and analyze offline RL objectives that can be viewed as reward-weighted supervised fine-tuning (SFT) and easily optimized in large language models. Our work stands in a stark contrast to recently proposed methods, based on SFT and direct preference optimization, which have additional hyper-parameters and do not directly optimize rewards. We compare to these methods empirically and report gains in both optimized rewards and language quality.
Authors: Riley Simmons-Edler, Ryan P. Badman, Felix Baastad Berg, Raymond Chua, John J. Vastola, Joshua Lunger, William Qian, Kanaka Rajan
Abstract: Understanding the behavior of deep reinforcement learning (DRL) agents -- particularly as task and agent sophistication increase -- requires more than simple comparison of reward curves, yet standard methods for behavioral analysis remain underdeveloped in DRL. We apply tools from neuroscience and ethology to study DRL agents in a novel, complex, partially observable environment, ForageWorld, designed to capture key aspects of real-world animal foraging -- including sparse, depleting resource patches, predator threats, and spatially extended arenas. We use this environment as a platform for applying joint behavioral and neural analysis to agents, revealing detailed, quantitatively grounded insights into agent strategies, memory, and planning. Contrary to common assumptions, we find that model-free RNN-based DRL agents can exhibit structured, planning-like behavior purely through emergent dynamics -- without requiring explicit memory modules or world models. Our results show that studying DRL agents like animals -- analyzing them with neuroethology-inspired tools that reveal structure in both behavior and neural dynamics -- uncovers rich structure in their learning dynamics that would otherwise remain invisible. We distill these tools into a general analysis framework linking core behavioral and representational features to diagnostic methods, which can be reused for a wide range of tasks and agents. As agents grow more complex and autonomous, bridging neuroscience, cognitive science, and AI will be essential -- not just for understanding their behavior, but for ensuring safe alignment and maximizing desirable behaviors that are hard to measure via reward. We show how this can be done by drawing on lessons from how biological intelligence is studied.
Authors: Md Aminul Islam, Kathryn Vasilaky, Elena Zheleva
Abstract: Implicit feedback data, such as user clicks, is commonly used in learning-to-rank (LTR) systems because it is easy to collect and it often reflects user preferences. However, this data is prone to various biases, and training an LTR system directly on biased data can result in suboptimal ranking performance. One of the most prominent and well-studied biases in implicit feedback data is position bias, which occurs because users are more likely to interact with higher-ranked documents regardless of their true relevance. In this paper, we propose a novel control function-based method that accounts for position bias in a two-stage process. The first stage uses exogenous variation from the residuals of the ranking process to correct for position bias in the second stage click equation. Unlike previous position bias correction methods, our method does not require knowledge of the click or propensity model and allows for nonlinearity in the underlying ranking model. Moreover, our method is general and allows for debiasing any state-of-the-art ranking algorithm by plugging it into the second stage. We also introduce a technique to debias validation clicks for hyperparameter tuning to select the optimal model in the absence of unbiased validation data. Experimental results demonstrate that our method outperforms state-of-the-art approaches in correcting for position bias.
Authors: Ming Li, Zhengyuan Yang, Xiyao Wang, Dianqi Li, Kevin Lin, Tianyi Zhou, Lijuan Wang
Abstract: Large reasoning models (LRMs) achieve strong reasoning performance by emitting long chains of thought. Yet, these verbose traces slow down inference and often drift into unnecessary detail, known as the overthinking phenomenon. To better understand LRMs' behavior, we systematically analyze the token-level misalignment between reasoning and non-reasoning models. While it is expected that their primary difference lies in the stylistic "thinking cues", LRMs uniquely exhibit two pivotal, previously under-explored phenomena: a Global Misalignment Rebound, where their divergence from non-reasoning models persists or even grows as response length increases, and more critically, a Local Misalignment Diminish, where the misalignment concentrates at the "thinking cues" each sentence starts with but rapidly declines in the remaining of the sentence. Motivated by the Local Misalignment Diminish, we propose FoReaL-Decoding, a collaborative fast-slow thinking decoding method for cost-quality trade-off. In FoReaL-Decoding, a Leading model leads the first few tokens for each sentence, and then a weaker draft model completes the following tokens to the end of each sentence. FoReaL-Decoding adopts a stochastic gate to smoothly interpolate between the small and the large model. On four popular math-reasoning benchmarks (AIME24, GPQA-Diamond, MATH500, AMC23), FoReaL-Decoding reduces theoretical FLOPs by 30 to 50% and trims CoT length by up to 40%, while preserving 86 to 100% of model performance. These results establish FoReaL-Decoding as a simple, plug-and-play route to controllable cost-quality trade-offs in reasoning-centric tasks.
Authors: Yuan-Hao Wei, Yan-Jie Sun
Abstract: This study advances the Variational Autoencoder (VAE) framework by addressing challenges in Independent Component Analysis (ICA) under both determined and underdetermined conditions, focusing on enhancing the independence and interpretability of latent variables. Traditional VAEs map observed data to latent variables and back via an encoder-decoder architecture, but struggle with underdetermined ICA where the number of latent variables exceeds observed signals. The proposed Half Adversarial VAE (Half-AVAE) builds on the encoder-free Half-VAE framework, eliminating explicit inverse mapping to tackle underdetermined scenarios. By integrating adversarial networks and External Enhancement (EE) terms, Half-AVAE promotes mutual independence among latent dimensions, achieving factorized and interpretable representations. Experiments with synthetic signals demonstrate that Half-AVAE outperforms baseline models, including GP-AVAE and Half-VAE, in recovering independent components under underdetermined conditions, as evidenced by lower root mean square errors. The study highlights the flexibility of VAEs in variational inference, showing that encoder omission, combined with adversarial training and structured priors, enables effective solutions for complex ICA tasks, advancing applications in disentanglement, causal inference, and generative modeling.
Authors: Qiyu Hou, Jun Wang
Abstract: To address the challenges of table structure recognition, we propose a novel Split-Merge-based top-down model optimized for large, densely populated tables. Our approach formulates row and column splitting as sequence labeling tasks, utilizing dual Transformer encoders to capture feature interactions. The merging process is framed as a grid cell classification task, leveraging an additional Transformer encoder to ensure accurate and coherent merging. By eliminating unstable bounding box predictions, our method reduces resolution loss and computational complexity, achieving high accuracy while maintaining fast processing speed. Extensive experiments on FinTabNet and PubTabNet demonstrate the superiority of our model over existing approaches, particularly in real-world applications. Our method offers a robust, scalable, and efficient solution for large-scale table recognition, making it well-suited for industrial deployment.
Authors: Zhenyu Liu, Huizhi Liang, Rajiv Ranjan, Zhanxing Zhu, Vaclav Snasel, Varun Ojha
Abstract: The robustness of Deep Neural Network models is crucial for defending models against adversarial attacks. Recent defense methods have employed collaborative learning frameworks to enhance model robustness. Two key limitations of existing methods are (i) insufficient guidance of the target model via loss functions and (ii) non-collaborative adversarial generation. We, therefore, propose a dual regularization loss (D2R Loss) method and a collaborative adversarial generation (CAG) strategy for adversarial training. D2R loss includes two optimization steps. The adversarial distribution and clean distribution optimizations enhance the target model's robustness by leveraging the strengths of different loss functions obtained via a suitable function space exploration to focus more precisely on the target model's distribution. CAG generates adversarial samples using a gradient-based collaboration between guidance and target models. We conducted extensive experiments on three benchmark databases, including CIFAR-10, CIFAR-100, Tiny ImageNet, and two popular target models, WideResNet34-10 and PreActResNet18. Our results show that D2R loss with CAG produces highly robust models.
Authors: Zhican Wang, Guanghui He, Dantong Liu, Lingjun Gao, Shell Xu Hu, Chen Zhang, Zhuoran Song, Nicholas Lane, Wayne Luk, Hongxiang Fan
Abstract: 3D Gaussian Splatting (3DGS) has recently gained significant attention for high-quality and efficient view synthesis, making it widely adopted in fields such as AR/VR, robotics, and autonomous driving. Despite its impressive algorithmic performance, real-time rendering on resource-constrained devices remains a major challenge due to tight power and area budgets. This paper presents an architecture-algorithm co-design to address these inefficiencies. First, we reveal substantial redundancy caused by repeated computation of common terms/expressions during the conventional rasterization. To resolve this, we propose axis-oriented rasterization, which pre-computes and reuses shared terms along both the X and Y axes through a dedicated hardware design, effectively reducing multiply-and-add (MAC) operations by up to 63%. Second, by identifying the resource and performance inefficiency of the sorting process, we introduce a novel neural sorting approach that predicts order-independent blending weights using an efficient neural network, eliminating the need for costly hardware sorters. A dedicated training framework is also proposed to improve its algorithmic stability. Third, to uniformly support rasterization and neural network inference, we design an efficient reconfigurable processing array that maximizes hardware utilization and throughput. Furthermore, we introduce a $\pi$-trajectory tile schedule, inspired by Morton encoding and Hilbert curve, to optimize Gaussian reuse and reduce memory access overhead. Comprehensive experiments demonstrate that the proposed design preserves rendering quality while achieving a speedup of $23.4\sim27.8\times$ and energy savings of $28.8\sim51.4\times$ compared to edge GPUs for real-world scenes. We plan to open-source our design to foster further development in this field.
Authors: Mostafa Eslami, Maryam Babazadeh
Abstract: This paper introduces a hypothetical hybrid control framework for port-Hamiltonian (p$\mathcal{H}$) systems, employing a dynamic decomposition based on Data-Assisted Control (DAC). The system's evolution is split into two parts with fixed topology: Right-Hand Side (RHS)- an intrinsic Hamiltonian flow handling worst-case parametric uncertainties, and Left-Hand Side (LHS)- a dissipative/input flow addressing both structural and parametric uncertainties. A virtual port variable $\Pi$ serves as the interface between these two components. A nonlinear controller manages the intrinsic Hamiltonian flow, determining a desired port control value $\Pi_c$. Concurrently, Reinforcement Learning (RL) is applied to the dissipative/input flow to learn an agent for providing optimal policy in mapping $\Pi_c$ to the actual system input. This hybrid approach effectively manages RHS uncertainties while preserving the system's inherent structure. Key advantages include adjustable performance via LHS controller parameters, enhanced AI explainability and interpretability through the port variable $\Pi$, the ability to guarantee safety and state attainability with hard/soft constraints, reduced complexity in learning hypothesis classes compared to end-to-end solutions, and improved state/parameter estimation using LHS prior knowledge and system Hamiltonian to address partial observability. The paper details the p$\mathcal{H}$ formulation, derives the decomposition, and presents the modular controller architecture. Beyond design, crucial aspects of stability and robustness analysis and synthesis are investigated, paving the way for deeper theoretical investigations. An application example, a pendulum with nonlinear dynamics, is simulated to demonstrate the approach's empirical and phenomenological benefits for future research.
Authors: Jiawen Li, Jiang Guo, Yuanzhe Li, Zetian Mao, Jiaxing Shen, Tashi Xu, Diptesh Das, Jinming He, Run Hu, Yaerim Lee, Koji Tsuda, Junichiro Shiomi
Abstract: Metamaterials are artificially engineered structures that manipulate electromagnetic waves, having optical properties absent in natural materials. Recently, machine learning for the inverse design of metamaterials has drawn attention. However, the highly nonlinear relationship between the metamaterial structures and optical behaviour, coupled with fabrication difficulties, poses challenges for using machine learning to design and manufacture complex metamaterials. Herein, we propose a general framework that implements customised spectrum-to-shape and size parameters to address one-to-many metamaterial inverse design problems using conditional diffusion models. Our method exhibits superior spectral prediction accuracy, generates a diverse range of patterns compared to other typical generative models, and offers valuable prior knowledge for manufacturing through the subsequent analysis of the diverse generated results, thereby facilitating the experimental fabrication of metamaterial designs. We demonstrate the efficacy of the proposed method by successfully designing and fabricating a free-form metamaterial with a tailored selective emission spectrum for thermal camouflage applications.
Authors: Zhongren Chen, Siyu Chen, Zhengling Qi, Xiaohong Chen, Zhuoran Yang
Abstract: We study quantile-optimal policy learning where the goal is to find a policy whose reward distribution has the largest $\alpha$-quantile for some $\alpha \in (0, 1)$. We focus on the offline setting whose generating process involves unobserved confounders. Such a problem suffers from three main challenges: (i) nonlinearity of the quantile objective as a functional of the reward distribution, (ii) unobserved confounding issue, and (iii) insufficient coverage of the offline dataset. To address these challenges, we propose a suite of causal-assisted policy learning methods that provably enjoy strong theoretical guarantees under mild conditions. In particular, to address (i) and (ii), using causal inference tools such as instrumental variables and negative controls, we propose to estimate the quantile objectives by solving nonlinear functional integral equations. Then we adopt a minimax estimation approach with nonparametric models to solve these integral equations, and propose to construct conservative policy estimates that address (iii). The final policy is the one that maximizes these pessimistic estimates. In addition, we propose a novel regularized policy learning method that is more amenable to computation. Finally, we prove that the policies learned by these methods are $\tilde{\mathscr{O}}(n^{-1/2})$ quantile-optimal under a mild coverage assumption on the offline dataset. Here, $\tilde{\mathscr{O}}(\cdot)$ omits poly-logarithmic factors. To the best of our knowledge, we propose the first sample-efficient policy learning algorithms for estimating the quantile-optimal policy when there exist unmeasured confounding.
Authors: Vicky Xefteri, Tim Vieira, Ryan Cotterell, Afra Amini
Abstract: Controlling the syntactic structure of text generated by language models is valuable for applications requiring clarity, stylistic consistency, or interpretability, yet it remains a challenging task. In this paper, we argue that sampling algorithms based on the posterior inference can effectively enforce a target constituency structure during generation. Our approach combines sequential Monte Carlo, which estimates the posterior distribution by sampling from a proposal distribution, with a syntactic tagger that ensures that each generated token aligns with the desired syntactic structure. Our experiments with GPT2 and Llama3-8B models show that with an appropriate proposal distribution, we can improve syntactic accuracy, increasing the F1 score from $12.31$ (GPT2-large) and $35.33$ (Llama3-8B) to about $93$ in both cases without compromising the language model's fluency. These results underscore both the complexity of syntactic control and the effectiveness of sampling algorithms, offering a promising approach for applications where precise control over syntax is essential.
Authors: Mrinmay Sen, Chalavadi Krishna Mohan
Abstract: Personalized Federated Learning (PFL) enables clients to collaboratively train personalized models tailored to their individual objectives, addressing the challenge of model generalization in traditional Federated Learning (FL) due to high data heterogeneity. However, existing PFL methods often require increased communication rounds to achieve the desired performance, primarily due to slow training caused by the use of first-order optimization, which has linear convergence. Additionally, many of these methods increase local computation because of the additional data fed into the model during the search for personalized local models. One promising solution to this slow training is second-order optimization, known for its quadratic convergence. However, employing it in PFL is challenging due to the Hessian matrix and its inverse. In this paper, we propose pFedSOP, which efficiently utilizes second-order optimization in PFL to accelerate the training of personalized models and enhance performance with fewer communication rounds. Our approach first computes a personalized local gradient update using the Gompertz function-based normalized angle between local and global gradient updates, incorporating client-specific global information. We then use a regularized Fisher Information Matrix (FIM), computed from this personalized gradient update, as an approximation of the Hessian to update the personalized models. This FIM-based second-order optimization speeds up training with fewer communication rounds by tackling the challenges with exact Hessian and avoids additional data being fed into the model during the search for personalized local models. Extensive experiments on heterogeneously partitioned image classification datasets with partial client participation demonstrate that pFedSOP outperforms state-of-the-art FL and PFL algorithms.
Authors: Chenlong Zhang, Zhuoran Jin, Hongbang Yuan, Jiaheng Wei, Tong Zhou, Kang Liu, Jun Zhao, Yubo Chen
Abstract: The widespread deployment of Large Language Models (LLMs) trained on massive, uncurated corpora has raised growing concerns about the inclusion of sensitive, copyrighted, or illegal content. This has led to increasing interest in LLM unlearning: the task of selectively removing specific information from a model without retraining from scratch or degrading overall utility. However, existing methods often rely on large-scale forget and retain datasets, and suffer from unnatural responses, poor generalization, or catastrophic utility loss. In this work, we propose Reinforcement UnLearning (RULE), an efficient framework that formulates unlearning as a refusal boundary optimization problem. RULE is trained with a small portion of the forget set and synthesized boundary queries, using a verifiable reward function that encourages safe refusal on forget--related queries while preserving helpful responses on permissible inputs. We provide both theoretical and empirical evidence demonstrating the effectiveness of RULE in achieving targeted unlearning without compromising model utility. Experimental results show that, with only $12%$ forget set and $8%$ synthesized boundary data, RULE outperforms existing baselines by up to $17.5%$ forget quality and $16.3%$ naturalness response while maintaining general utility, achieving forget--retain Pareto optimality. Remarkably, we further observe that RULE improves the naturalness of model outputs, enhances training efficiency, and exhibits strong generalization ability, generalizing refusal behavior to semantically related but unseen queries.
Authors: Ben Hayes, Charalampos Saitis, Gy\"orgy Fazekas
Abstract: Many audio synthesizers can produce the same signal given different parameter configurations, meaning the inversion from sound to parameters is an inherently ill-posed problem. We show that this is largely due to intrinsic symmetries of the synthesizer, and focus in particular on permutation invariance. First, we demonstrate on a synthetic task that regressing point estimates under permutation symmetry degrades performance, even when using a permutation-invariant loss function or symmetry-breaking heuristics. Then, viewing equivalent solutions as modes of a probability distribution, we show that a conditional generative model substantially improves performance. Further, acknowledging the invariance of the implicit parameter distribution, we find that performance is further improved by using a permutation equivariant continuous normalizing flow. To accommodate intricate symmetries in real synthesizers, we also propose a relaxed equivariance strategy that adaptively discovers relevant symmetries from data. Applying our method to Surge XT, a full-featured open source synthesizer used in real world audio production, we find our method outperforms regression and generative baselines across audio reconstruction metrics.
Authors: Xinran Li, Chenjia Bai, Zijian Li, Jiakun Zheng, Ting Xiao, Jun Zhang
Abstract: Large language models (LLMs) possess extensive knowledge bases and strong reasoning capabilities, making them promising tools for complex, multi-agent planning in embodied environments. However, despite LLMs' advanced abilities and the sophisticated modular design of agentic methods, existing LLM-based planning algorithms remain limited by weak adaptation capabilities to multi-agent embodied scenarios. We address this limitation by introducing a framework that enables LLM agents to learn and evolve both before and during test time, equipping them with environment-relevant knowledge for better planning and enhanced communication for improved cooperation. Inspired by centralized training with decentralized execution in multi-agent reinforcement learning, we propose a \textit{Learn as Individuals, Evolve as a Team (LIET)} paradigm for multi-agent LLMs adaptation. At the individual level, LLM agents learn a local utility function from exploratory datasets to better comprehend the embodied environment, which is then queried during test time to support informed decision-making. At the team level, LLM agents collaboratively and iteratively maintain and update a shared cooperation knowledge list based on new experiences, using it to guide more effective communication. By combining individual learning with team evolution, LIET enables comprehensive and flexible adaptation for LLM agents. Our experiments on Communicative Watch-And-Help and ThreeD-World Multi-Agent Transport benchmarks demonstrate that LIET, instantiated with both LLaMA and GPT-4o, outperforms existing baselines and exhibits strong cooperative planning abilities.
Authors: Prathamesh Kokate, Mitali Sarnaik, Manavi Khopade, Raviraj Joshi
Abstract: Long document classification poses challenges due to the computational limitations of transformer-based models, particularly BERT, which are constrained by fixed input lengths and quadratic attention complexity. Moreover, using the full document for classification is often redundant, as only a subset of sentences typically carries the necessary information. To address this, we propose a TF-IDF-based sentence ranking method that improves efficiency by selecting the most informative content. Our approach explores fixed-count and percentage-based sentence selection, along with an enhanced scoring strategy combining normalized TF-IDF scores and sentence length. Evaluated on the MahaNews LDC dataset of long Marathi news articles, the method consistently outperforms baselines such as first, last, and random sentence selection. With MahaBERT-v2, we achieve near-identical classification accuracy with just a 0.33 percent drop compared to the full-context baseline, while reducing input size by over 50 percent and inference latency by 43 percent. This demonstrates that significant context reduction is possible without sacrificing performance, making the method practical for real-world long document classification tasks.
Authors: Daolang Huang, Xinyi Wen, Ayush Bharti, Samuel Kaski, Luigi Acerbi
Abstract: Many critical applications, from autonomous scientific discovery to personalized medicine, demand systems that can both strategically acquire the most informative data and instantaneously perform inference based upon it. While amortized methods for Bayesian inference and experimental design offer part of the solution, neither approach is optimal in the most general and challenging task, where new data needs to be collected for instant inference. To tackle this issue, we introduce the Amortized Active Learning and Inference Engine (ALINE), a unified framework for amortized Bayesian inference and active data acquisition. ALINE leverages a transformer architecture trained via reinforcement learning with a reward based on self-estimated information gain provided by its own integrated inference component. This allows it to strategically query informative data points while simultaneously refining its predictions. Moreover, ALINE can selectively direct its querying strategy towards specific subsets of model parameters or designated predictive tasks, optimizing for posterior estimation, data prediction, or a mixture thereof. Empirical results on regression-based active learning, classical Bayesian experimental design benchmarks, and a psychometric model with selectively targeted parameters demonstrate that ALINE delivers both instant and accurate inference along with efficient selection of informative points.
Authors: Amit Jaspal, Qian Dang, Ajantha Ramineni
Abstract: Modern large-scale recommender systems employ multi-stage ranking funnel (Retrieval, Pre-ranking, Ranking) to balance engagement and computational constraints (latency, CPU). However, the initial retrieval stage, often relying on efficient but less precise methods like K-Nearest Neighbors (KNN), struggles to effectively surface the most engaging items from billion-scale catalogs, particularly distinguishing highly relevant and engaging candidates from merely relevant ones. We introduce Recall Augmentation through Deferred Asynchronous Retrieval (RADAR), a novel framework that leverages asynchronous, offline computation to pre-rank a significantly larger candidate set for users using the full complexity ranking model. These top-ranked items are stored and utilized as a high-quality retrieval source during online inference, bypassing online retrieval and pre-ranking stages for these candidates. We demonstrate through offline experiments that RADAR significantly boosts recall (2X Recall@200 vs DNN retrieval baseline) by effectively combining a larger retrieved candidate set with a more powerful ranking model. Online A/B tests confirm a +0.8% lift in topline engagement metrics, validating RADAR as a practical and effective method to improve recommendation quality under strict online serving constraints.
Authors: Hikaru Sawafuji, Ryota Ozaki, Takuto Motomura, Toyohisa Matsuda, Masanori Tojima, Kento Uchida, Shinichi Shirakawa
Abstract: Self-localization is an important technology for automating bulldozers. Conventional bulldozer self-localization systems rely on RTK-GNSS (Real Time Kinematic-Global Navigation Satellite Systems). However, RTK-GNSS signals are sometimes lost in certain mining conditions. Therefore, self-localization methods that do not depend on RTK-GNSS are required. In this paper, we propose a machine learning-based self-localization method for bulldozers. The proposed method consists of two steps: estimating local velocities using a machine learning model from internal sensors, and incorporating these estimates into an Extended Kalman Filter (EKF) for global localization. We also created a novel dataset for bulldozer odometry and conducted experiments across various driving scenarios, including slalom, excavation, and driving on slopes. The result demonstrated that the proposed self-localization method suppressed the accumulation of position errors compared to kinematics-based methods, especially when slip occurred. Furthermore, this study showed that bulldozer-specific sensors, such as blade position sensors and hydraulic pressure sensors, contributed to improving self-localization accuracy.
Authors: Aditya Chakravarty
Abstract: Diffusion models have shown remarkable flexibility for solving inverse problems without task-specific retraining. However, existing approaches such as Manifold Preserving Guided Diffusion (MPGD) apply only a single gradient update per denoising step, limiting restoration fidelity and robustness, especially in embedded or out-of-distribution settings. In this work, we introduce a multistep optimization strategy within each denoising timestep, significantly enhancing image quality, perceptual accuracy, and generalization. Our experiments on super-resolution and Gaussian deblurring demonstrate that increasing the number of gradient updates per step improves LPIPS and PSNR with minimal latency overhead. Notably, we validate this approach on a Jetson Orin Nano using degraded ImageNet and a UAV dataset, showing that MPGD, originally trained on face datasets, generalizes effectively to natural and aerial scenes. Our findings highlight MPGD's potential as a lightweight, plug-and-play restoration module for real-time visual perception in embodied AI agents such as drones and mobile robots.
Authors: Xuanjun Chen, I-Ming Lin, Lin Zhang, Haibin Wu, Hung-yi Lee, Jyh-Shing Roger Jang
Abstract: Recent attempts at source tracing for codec-based deepfake speech (CodecFake), generated by neural audio codec-based speech generation (CoSG) models, have exhibited suboptimal performance. However, how to train source tracing models using simulated CoSG data while maintaining strong performance on real CoSG-generated audio remains an open challenge. In this paper, we show that models trained solely on codec-resynthesized data tend to overfit to non-speech regions and struggle to generalize to unseen content. To mitigate these challenges, we introduce the Semantic-Acoustic Source Tracing Network (SASTNet), which jointly leverages Whisper for semantic feature encoding and Wav2vec2 with AudioMAE for acoustic feature encoding. Our proposed SASTNet achieves state-of-the-art performance on the CoSG test set of the CodecFake+ dataset, demonstrating its effectiveness for reliable source tracing.
Authors: Hans Buehler, Blanka Horvath, Yannick Limmer, Thorsten Schmidt
Abstract: This paper addresses the challenge of model uncertainty in quantitative finance, where decisions in portfolio allocation, derivative pricing, and risk management rely on estimating stochastic models from limited data. In practice, the unavailability of the true probability measure forces reliance on an empirical approximation, and even small misestimations can lead to significant deviations in decision quality. Building on the framework of Klibanoff et al. (2005), we enhance the conventional objective - whether this is expected utility in an investing context or a hedging metric - by superimposing an outer "uncertainty measure", motivated by traditional monetary risk measures, on the space of models. In scenarios where a natural model distribution is lacking or Bayesian methods are impractical, we propose an ad hoc subsampling strategy, analogous to bootstrapping in statistical finance and related to mini-batch sampling in deep learning, to approximate model uncertainty. To address the quadratic memory demands of naive implementations, we also present an adapted stochastic gradient descent algorithm that enables efficient parallelization. Through analytical, simulated, and empirical studies - including multi-period, real data and high-dimensional examples - we demonstrate that uncertainty measures outperform traditional mixture of measures strategies and our model-agnostic subsampling-based approach not only enhances robustness against model risk but also achieves performance comparable to more elaborate Bayesian methods.
Authors: Brian Christian, Hannah Rose Kirk, Jessica A. F. Thompson, Christopher Summerfield, Tsvetomira Dumbalska
Abstract: Reward modeling has emerged as a crucial component in aligning large language models with human values. Significant attention has focused on using reward models as a means for fine-tuning generative models. However, the reward models themselves -- which directly encode human value judgments by turning prompt-response pairs into scalar rewards -- remain relatively understudied. We present a novel approach to reward model interpretability through exhaustive analysis of their responses across their entire vocabulary space. By examining how different reward models score every possible single-token response to value-laden prompts, we uncover several striking findings: (i) substantial heterogeneity between models trained on similar objectives, (ii) systematic asymmetries in how models encode high- vs low-scoring tokens, (iii) significant sensitivity to prompt framing that mirrors human cognitive biases, and (iv) overvaluation of more frequent tokens. We demonstrate these effects across ten recent open-source reward models of varying parameter counts and architectures. Our results challenge assumptions about the interchangeability of reward models, as well as their suitability as proxies of complex and context-dependent human values. We find that these models can encode concerning biases toward certain identity groups, which may emerge as unintended consequences of harmlessness training -- distortions that risk propagating through the downstream large language models now deployed to millions.
Authors: Dane Williamson, Yangfeng Ji, Matthew Dwyer
Abstract: Saliency methods are widely used to visualize which input features are deemed relevant to a model's prediction. However, their visual plausibility can obscure critical limitations. In this work, we propose a diagnostic test for class sensitivity: a method's ability to distinguish between competing class labels on the same input. Through extensive experiments, we show that many widely used saliency methods produce nearly identical explanations regardless of the class label, calling into question their reliability. We find that class-insensitive behavior persists across architectures and datasets, suggesting the failure mode is structural rather than model-specific. Motivated by these findings, we introduce CASE, a contrastive explanation method that isolates features uniquely discriminative for the predicted class. We evaluate CASE using the proposed diagnostic and a perturbation-based fidelity test, and show that it produces faithful and more class-specific explanations than existing methods.
Authors: Kevin Black, Manuel Y. Galliker, Sergey Levine
Abstract: Modern AI systems, especially those interacting with the physical world, increasingly require real-time performance. However, the high latency of state-of-the-art generalist models, including recent vision-language action models (VLAs), poses a significant challenge. While action chunking has enabled temporal consistency in high-frequency control tasks, it does not fully address the latency problem, leading to pauses or out-of-distribution jerky movements at chunk boundaries. This paper presents a novel inference-time algorithm that enables smooth asynchronous execution of action chunking policies. Our method, real-time chunking (RTC), is applicable to any diffusion- or flow-based VLA out of the box with no re-training. It generates the next action chunk while executing the current one, "freezing" actions guaranteed to execute and "inpainting" the rest. To test RTC, we introduce a new benchmark of 12 highly dynamic tasks in the Kinetix simulator, as well as evaluate 6 challenging real-world bimanual manipulation tasks. Results demonstrate that RTC is fast, performant, and uniquely robust to inference delay, significantly improving task throughput and enabling high success rates in precise tasks $\unicode{x2013}$ such as lighting a match $\unicode{x2013}$ even in the presence of significant latency. See https://pi.website/research/real_time_chunking for videos.
Authors: Armin Lederer, Erfaun Noorani, Andreas Krause
Abstract: Ensuring safety in multi-agent systems is a significant challenge, particularly in settings where centralized coordination is impractical. In this work, we propose a novel risk-sensitive safety filter for discrete-time multi-agent systems with uncertain dynamics that leverages control barrier functions (CBFs) defined through value functions. Our approach relies on centralized risk-sensitive safety conditions based on exponential risk operators to ensure robustness against model uncertainties. We introduce a distributed formulation of the safety filter by deriving two alternative strategies: one based on worst-case anticipation and another on proximity to a known safe policy. By allowing agents to switch between strategies, feasibility can be ensured. Through detailed numerical evaluations, we demonstrate the efficacy of our approach in maintaining safety without being overly conservative.
Authors: Jun Chen, Lina Liu, Tianyi Zhu, Yong Liu, Guang Dai, Yunliang Jiang, Ivor W. Tsang
Abstract: This paper considers the problem of decentralized optimization on compact submanifolds, where a finite sum of smooth (possibly non-convex) local functions is minimized by $n$ agents forming an undirected and connected graph. However, the efficiency of distributed optimization is often hindered by communication bottlenecks. To mitigate this, we propose the Quantized Riemannian Gradient Tracking (Q-RGT) algorithm, where agents update their local variables using quantized gradients. The introduction of quantization noise allows our algorithm to bypass the constraints of the accurate Riemannian projection operator (such as retraction), further improving iterative efficiency. To the best of our knowledge, this is the first algorithm to achieve an $\mathcal{O}(1/K)$ convergence rate in the presence of quantization, matching the convergence rate of methods without quantization. Additionally, we explicitly derive lower bounds on decentralized consensus associated with a function of quantization levels. Numerical experiments demonstrate that Q-RGT performs comparably to non-quantized methods while reducing communication bottlenecks and computational overhead.
Authors: Satvik Praveen, Yoonsung Jung
Abstract: Object detection is vital in precision agriculture for plant monitoring, disease detection, and yield estimation. However, models like YOLO struggle with occlusions, irregular structures, and background noise, reducing detection accuracy. While Spatial Transformer Networks (STNs) improve spatial invariance through learned transformations, affine mappings are insufficient for non-rigid deformations such as bent leaves and overlaps. We propose CBAM-STN-TPS-YOLO, a model integrating Thin-Plate Splines (TPS) into STNs for flexible, non-rigid spatial transformations that better align features. Performance is further enhanced by the Convolutional Block Attention Module (CBAM), which suppresses background noise and emphasizes relevant spatial and channel-wise features. On the occlusion-heavy Plant Growth and Phenotyping (PGP) dataset, our model outperforms STN-YOLO in precision, recall, and mAP. It achieves a 12% reduction in false positives, highlighting the benefits of improved spatial flexibility and attention-guided refinement. We also examine the impact of the TPS regularization parameter in balancing transformation smoothness and detection performance. This lightweight model improves spatial awareness and supports real-time edge deployment, making it ideal for smart farming applications requiring accurate and efficient monitoring.
Authors: Kuiyuan Zhang, Wenjie Pei, Rushi Lan, Yifang Guo, Zhongyun Hua
Abstract: Deepfakes are AI-synthesized multimedia data that may be abused for spreading misinformation. Deepfake generation involves both visual and audio manipulation. To detect audio-visual deepfakes, previous studies commonly employ two relatively independent sub-models to learn audio and visual features, respectively, and fuse them subsequently for deepfake detection. However, this may underutilize the inherent correlations between audio and visual features. Moreover, utilizing two isolated feature learning sub-models can result in redundant neural layers, making the overall model inefficient and impractical for resource-constrained environments. In this work, we design a lightweight network for audio-visual deepfake detection via a single-stream multi-modal learning framework. Specifically, we introduce a collaborative audio-visual learning block to efficiently integrate multi-modal information while learning the visual and audio features. By iteratively employing this block, our single-stream network achieves a continuous fusion of multi-modal features across its layers. Thus, our network efficiently captures visual and audio features without the need for excessive block stacking, resulting in a lightweight network design. Furthermore, we propose a multi-modal classification module that can boost the dependence of the visual and audio classifiers on modality content. It also enhances the whole resistance of the video classifier against the mismatches between audio and visual modalities. We conduct experiments on the DF-TIMIT, FakeAVCeleb, and DFDC benchmark datasets. Compared to state-of-the-art audio-visual joint detection methods, our method is significantly lightweight with only 0.48M parameters, yet it achieves superiority in both uni-modal and multi-modal deepfakes, as well as in unseen types of deepfakes.
Authors: Chengchao Shen, Dawei Liu, Jianxin Wang
Abstract: Contrastive learning for single object centric images has achieved remarkable progress on unsupervised representation, but suffering inferior performance on the widespread images with multiple objects. In this paper, we propose a simple but effective method, Multiple Object Stitching (MOS), to refine the unsupervised representation for multi-object images. Specifically, we construct the multi-object images by stitching the single object centric ones, where the objects in the synthesized multi-object images are predetermined. Hence, compared to the existing contrastive methods, our method provides additional object correspondences between multi-object images without human annotations. In this manner, our method pays more attention to the representations of each object in multi-object image, thus providing more detailed representations for complicated downstream tasks, such as object detection and semantic segmentation. Experimental results on ImageNet, CIFAR and COCO datasets demonstrate that our proposed method achieves the leading unsupervised representation performance on both single object centric images and multi-object ones. The source code is available at https://github.com/visresearch/MultipleObjectStitching.
URLs: https://github.com/visresearch/MultipleObjectStitching.
Authors: Jintao Tong, Ran Ma, Yixiong Zou, Guangyao Chen, Yuhua Li, Ruixuan Li
Abstract: Cross-domain few-shot segmentation (CD-FSS) is proposed to pre-train the model on a source-domain dataset with sufficient samples, and then transfer the model to target-domain datasets where only a few samples are available for efficient fine-tuning. There are majorly two challenges in this task: (1) the domain gap and (2) fine-tuning with scarce data. To solve these challenges, we revisit the adapter-based methods, and discover an intriguing insight not explored in previous works: the adapter not only helps the fine-tuning of downstream tasks but also naturally serves as a domain information decoupler. Then, we delve into this finding for an interpretation, and find the model's inherent structure could lead to a natural decoupling of domain information. Building upon this insight, we propose the Domain Feature Navigator (DFN), which is a structure-based decoupler instead of loss-based ones like current works, to capture domain-specific information, thereby directing the model's attention towards domain-agnostic knowledge. Moreover, to prevent the potential excessive overfitting of DFN during the source-domain training, we further design the SAM-SVN method to constrain DFN from learning sample-specific knowledge. On target domains, we freeze the model and fine-tune the DFN to learn target-specific knowledge specific. Extensive experiments demonstrate that our method surpasses the state-of-the-art method in CD-FSS significantly by 2.69% and 4.68% MIoU in 1-shot and 5-shot scenarios, respectively.
Authors: Yuyang Zhou, Guang Cheng, Kang Du, Zihan Chen, Tian Qin, Yuyu Zhao
Abstract: The proliferation of unmanned aerial vehicle (UAV) swarms has enabled a wide range of mission-critical applications, but also exposes UAV networks to severe Denial-of-Service (DoS) threats due to their open wireless environment, dynamic topology, and resource constraints. Traditional static or centralized defense mechanisms are often inadequate for such dynamic and distributed scenarios. To address these challenges, we propose a novel federated multi-agent deep reinforcement learning (FMADRL)-driven moving target defense (MTD) framework for proactive and adaptive DoS mitigation in UAV swarm networks. Specifically, we design three lightweight and coordinated MTD mechanisms, including leader switching, route mutation, and frequency hopping, that leverage the inherent flexibility of UAV swarms to disrupt attacker efforts and enhance network resilience. The defense problem is formulated as a multi-agent partially observable Markov decision process (POMDP), capturing the distributed, resource-constrained, and uncertain nature of UAV swarms under attack. Each UAV is equipped with a local policy agent that autonomously selects MTD actions based on partial observations and local experiences. By employing a policy gradient-based FMADRL algorithm, UAVs collaboratively optimize their defense policies via reward-weighted aggregation, enabling distributed learning without sharing raw data and thus reducing communication overhead. Extensive simulations demonstrate that our approach significantly outperforms state-of-the-art baselines, achieving up to a 34.6% improvement in attack mitigation rate, a reduction in average recovery time of up to 94.6%, and decreases in energy consumption and defense cost by as much as 29.3% and 98.3%, respectively, while maintaining robust mission continuity under various DoS attack strategies.
Authors: Guibin Zhang, Muxin Fu, Guancheng Wan, Miao Yu, Kun Wang, Shuicheng Yan
Abstract: Large language model (LLM)-powered multi-agent systems (MAS) have demonstrated cognitive and execution capabilities that far exceed those of single LLM agents, yet their capacity for self-evolution remains hampered by underdeveloped memory architectures. Upon close inspection, we are alarmed to discover that prevailing MAS memory mechanisms (1) are overly simplistic, completely disregarding the nuanced inter-agent collaboration trajectories, and (2) lack cross-trial and agent-specific customization, in stark contrast to the expressive memory developed for single agents. To bridge this gap, we introduce G-Memory, a hierarchical, agentic memory system for MAS inspired by organizational memory theory, which manages the lengthy MAS interaction via a three-tier graph hierarchy: insight, query, and interaction graphs. Upon receiving a new user query, G-Memory performs bi-directional memory traversal to retrieve both $\textit{high-level, generalizable insights}$ that enable the system to leverage cross-trial knowledge, and $\textit{fine-grained, condensed interaction trajectories}$ that compactly encode prior collaboration experiences. Upon task execution, the entire hierarchy evolves by assimilating new collaborative trajectories, nurturing the progressive evolution of agent teams. Extensive experiments across five benchmarks, three LLM backbones, and three popular MAS frameworks demonstrate that G-Memory improves success rates in embodied action and accuracy in knowledge QA by up to $20.89\%$ and $10.12\%$, respectively, without any modifications to the original frameworks. Our codes are available at https://github.com/bingreeky/GMemory.
Authors: Philip Liu, Sparsh Bansal, Jimmy Dinh, Aditya Pawar, Ramani Satishkumar, Shail Desai, Neeraj Gupta, Xin Wang, Shu Hu
Abstract: The integration of deep learning-based glaucoma detection with large language models (LLMs) presents an automated strategy to mitigate ophthalmologist shortages and improve clinical reporting efficiency. However, applying general LLMs to medical imaging remains challenging due to hallucinations, limited interpretability, and insufficient domain-specific medical knowledge, which can potentially reduce clinical accuracy. Although recent approaches combining imaging models with LLM reasoning have improved reporting, they typically rely on a single generalist agent, restricting their capacity to emulate the diverse and complex reasoning found in multidisciplinary medical teams. To address these limitations, we propose MedChat, a multi-agent diagnostic framework and platform that combines specialized vision models with multiple role-specific LLM agents, all coordinated by a director agent. This design enhances reliability, reduces hallucination risk, and enables interactive diagnostic reporting through an interface tailored for clinical review and educational use. Code available at https://github.com/Purdue-M2/MedChat.
Authors: Yuling Wang, Zihui Chen, Pengfei Jiao, Xiao Wang
Abstract: Heterogeneous Graph Neural Networks (HGNNs) are vulnerable, highlighting the need for tailored attacks to assess their robustness and ensure security. However, existing HGNN attacks often require complex retraining of parameters to generate specific perturbations for new scenarios. Recently, foundation models have opened new horizons for the generalization of graph neural networks by capturing shared semantics across various graph distributions. This leads us to ask:Can we design a foundation attack model for HGNNs that enables generalizable perturbations across different HGNNs, and quickly adapts to new heterogeneous graphs (HGs)? Empirical findings reveal that, despite significant differences in model design and parameter space, different HGNNs surprisingly share common vulnerability patterns from a relation-aware perspective. Therefore, we explore how to design foundation HGNN attack criteria by mining shared attack units. In this paper, we propose a novel relation-wise heterogeneous graph foundation attack model, HeTa. We introduce a foundation surrogate model to align heterogeneity and identify the importance of shared relation-aware attack units. Building on this, we implement a serialized relation-by-relation attack based on the identified relational weights. In this way, the perturbation can be transferred to various target HGNNs and easily fine-tuned for new HGs. Extensive experiments exhibit powerful attack performances and generalizability of our method.
Authors: Alexander Kolpakov, Igor Rivin
Abstract: Computing classical centrality measures such as betweenness and closeness is computationally expensive on large-scale graphs. In this work, we introduce an efficient force layout algorithm that embeds a graph into a low-dimensional space, where the radial distance from the origin serves as a proxy for various centrality measures. We evaluate our method on multiple graph families and demonstrate strong correlations with degree, PageRank, and paths-based centralities. As an application, it turns out that the proposed embedding allows to find high-influence nodes in a network, and provides a fast and scalable alternative to the standard greedy algorithm.
Authors: Yuxin Xiao, Shan Chen, Jack Gallifant, Danielle Bitterman, Thomas Hartvigsen, Marzyeh Ghassemi
Abstract: Characterizing a large language model's (LLM's) knowledge of a given question is challenging. As a result, prior work has primarily examined LLM behavior under knowledge conflicts, where the model's internal parametric memory contradicts information in the external context. However, this does not fully reflect how well the model knows the answer to the question. In this paper, we first introduce a taxonomy of five knowledge statuses based on the consistency and correctness of LLM knowledge modes. We then propose KScope, a hierarchical framework of statistical tests that progressively refines hypotheses about knowledge modes and characterizes LLM knowledge into one of these five statuses. We apply KScope to nine LLMs across four datasets and systematically establish: (1) Supporting context narrows knowledge gaps across models. (2) Context features related to difficulty, relevance, and familiarity drive successful knowledge updates. (3) LLMs exhibit similar feature preferences when partially correct or conflicted, but diverge sharply when consistently wrong. (4) Context summarization constrained by our feature analysis, together with enhanced credibility, further improves update effectiveness and generalizes across LLMs.
Authors: Dasol Hong, Wooju Lee, Hyun Myung
Abstract: Prompt tuning, which adapts vision-language models by freezing model parameters and optimizing only the prompt, has proven effective for task-specific adaptations. The core challenge in prompt tuning is improving specialization for a specific task and generalization for unseen domains. However, frozen encoders often produce misaligned features, leading to confusion between classes and limiting specialization. To overcome this issue, we propose a confusion-aware loss (CoA-loss) that improves specialization by refining the decision boundaries between confusing classes. Additionally, we mathematically demonstrate that a mixture model can enhance generalization without compromising specialization. This is achieved using confidence-aware weights (CoA-weights), which adjust the weights of each prediction in the mixture model based on its confidence within the class domains. Extensive experiments show that CoCoA-Mix, a mixture model with CoA-loss and CoA-weights, outperforms state-of-the-art methods by enhancing specialization and generalization. Our code is publicly available at https://github.com/url-kaist/CoCoA-Mix.
Authors: Lu Ma, Hao Liang, Meiyi Qiang, Lexiang Tang, Xiaochen Ma, Zhen Hao Wong, Junbo Niu, Chengyu Shen, Runming He, Bin Cui, Wentao Zhang
Abstract: Recent advances in large language model (LLM) reasoning have shown that sophisticated behaviors such as planning and self-reflection can emerge through reinforcement learning (RL). However, despite these successes, RL in its current form remains insufficient to induce capabilities that exceed the limitations of the base model, as it is primarily optimized based on existing knowledge of the model rather than facilitating the acquisition of new information. To address this limitation, we employ supervised fine-tuning (SFT) to learn what RL cannot, which enables the incorporation of new knowledge and reasoning patterns by leveraging high-quality demonstration data. We analyze the training dynamics of RL and SFT for LLM reasoning and find that RL excels at maintaining and improving performance on questions within the model's original capabilities, while SFT is more effective at enabling progress on questions beyond the current scope of the model. Motivated by the complementary strengths of RL and SFT, we introduce a novel training approach, \textbf{ReLIFT} (\textbf{Re}inforcement \textbf{L}earning \textbf{I}nterleaved with Online \textbf{F}ine-\textbf{T}uning). In ReLIFT, the model is primarily trained using RL, but when it encounters challenging questions, high-quality solutions are collected for fine-tuning, and the training process alternates between RL and fine-tuning to enhance the model's reasoning abilities. ReLIFT achieves an average improvement of over +5.2 points across five competition-level benchmarks and one out-of-distribution benchmark compared to other zero-RL models. Furthermore, we demonstrate that ReLIFT outperforms both RL and SFT while using only 13\% of the detailed demonstration data, highlighting its scalability. These results provide compelling evidence that ReLIFT overcomes the fundamental limitations of RL and underscores the significant potential.
Authors: Ruiyang Zhang, Hu Zhang, Hao Fei, Zhedong Zheng
Abstract: Large Multimodal Models (LMMs), harnessing the complementarity among diverse modalities, are often considered more robust than pure Language Large Models (LLMs); yet do LMMs know what they do not know? There are three key open questions remaining: (1) how to evaluate the uncertainty of diverse LMMs in a unified manner, (2) how to prompt LMMs to show its uncertainty, and (3) how to quantify uncertainty for downstream tasks. In an attempt to address these challenges, we introduce Uncertainty-o: (1) a model-agnostic framework designed to reveal uncertainty in LMMs regardless of their modalities, architectures, or capabilities, (2) an empirical exploration of multimodal prompt perturbations to uncover LMM uncertainty, offering insights and findings, and (3) derive the formulation of multimodal semantic uncertainty, which enables quantifying uncertainty from multimodal responses. Experiments across 18 benchmarks spanning various modalities and 10 LMMs (both open- and closed-source) demonstrate the effectiveness of Uncertainty-o in reliably estimating LMM uncertainty, thereby enhancing downstream tasks such as hallucination detection, hallucination mitigation, and uncertainty-aware Chain-of-Thought reasoning.
Authors: Jiacheng Shi, Yanfu Zhang, Huajie Shao, Ashley Gao
Abstract: Recent advancements in diffusion models have enabled high-fidelity and photorealistic image generation across diverse applications. However, these models also present security and privacy risks, including copyright violations, sensitive information leakage, and the creation of harmful or offensive content that could be exploited maliciously. In this study, we uncover a novel security threat where an attacker leverages diffusion model APIs to generate synthetic images, which are then used to train a high-performing substitute model. This enables the attacker to execute model extraction and transfer-based adversarial attacks on black-box classification models with minimal queries, without needing access to the original training data. The generated images are sufficiently high-resolution and diverse to train a substitute model whose outputs closely match those of the target model. Across the seven benchmarks, including CIFAR and ImageNet subsets, our method shows an average improvement of 27.37% over state-of-the-art methods while using just 0.01 times of the query budget, achieving a 98.68% success rate in adversarial attacks on the target model.
Authors: Marco Di Gennaro, Giovanni De Lucia, Stefano Longari, Stefano Zanero, Michele Carminati
Abstract: Federated Learning has emerged as a privacy-oriented alternative to centralized Machine Learning, enabling collaborative model training without direct data sharing. While extensively studied for neural networks, the security and privacy implications of tree-based models remain underexplored. This work introduces TimberStrike, an optimization-based dataset reconstruction attack targeting horizontally federated tree-based models. Our attack, carried out by a single client, exploits the discrete nature of decision trees by using split values and decision paths to infer sensitive training data from other clients. We evaluate TimberStrike on State-of-the-Art federated gradient boosting implementations across multiple frameworks, including Flower, NVFlare, and FedTree, demonstrating their vulnerability to privacy breaches. On a publicly available stroke prediction dataset, TimberStrike consistently reconstructs between 73.05% and 95.63% of the target dataset across all implementations. We further analyze Differential Privacy, showing that while it partially mitigates the attack, it also significantly degrades model performance. Our findings highlight the need for privacy-preserving mechanisms specifically designed for tree-based Federated Learning systems, and we provide preliminary insights into their design.
Authors: Rishikesh Srinivasan, Dheeraj Nagaraj
Abstract: We study the problem of sampling from strongly log-concave distributions over $\mathbb{R}^d$ using the Poisson midpoint discretization (a variant of the randomized midpoint method) for overdamped/underdamped Langevin dynamics. We prove its convergence in the 2-Wasserstein distance ($W_2$), achieving a cubic speedup in dependence on the target accuracy ($\epsilon$) over the Euler-Maruyama discretization, surpassing existing bounds for randomized midpoint methods. Notably, in the case of underdamped Langevin dynamics, we demonstrate the complexity of $W_2$ convergence is much smaller than the complexity lower bounds for convergence in $L^2$ strong error established in the literature.
Authors: Harsh Bihany, Shubham Patel, Ashutosh Modi
Abstract: Large Language Models have shown remarkable capabilities in the NLP domain. Their effectiveness can mainly be attributed to their ability to adapt to an array of downstream tasks. However, generally, full fine-tuning is a computationally expensive job. To mitigate this, many techniques have been developed that prime efficiency, a prominent one being Low-Rank Adaptation (LoRA). However, LoRA and its variants employ re-parametrized additive updates. In this paper, we propose Low-Rank Multiplicative Adaptation (LoRMA), which shifts the paradigm of additive updates to a richer space of matrix multiplicative transformations. We tackle challenges such as computational complexity and rank bottleneck of matrix multiplication by effectively re-ordering operations and introducing rank inflation strategies. We conduct extensive experiments to demonstrate the effectiveness of our approach in terms of various evaluation metrics.
Authors: Yuchong Long, Wen Sun, Ningxiao Sun, Wenxiao Wang, Chao Li, Shan Yin
Abstract: Automated pollen recognition is vital to paleoclimatology, biodiversity monitoring, and public health, yet conventional methods are hampered by inefficiency and subjectivity. Existing deep learning models often struggle to achieve the requisite localization accuracy for microscopic targets like pollen, which are characterized by their minute size, indistinct edges, and complex backgrounds. To overcome this limitation, we introduce HieraEdgeNet, a multi-scale edge-enhancement framework. The framework's core innovation is the introduction of three synergistic modules: the Hierarchical Edge Module (HEM), which explicitly extracts a multi-scale pyramid of edge features that corresponds to the semantic hierarchy at early network stages; the Synergistic Edge Fusion (SEF) module, for deeply fusing these edge priors with semantic information at each respective scale; and the Cross Stage Partial Omni-Kernel Module (CSPOKM), which maximally refines the most detail-rich feature layers using an Omni-Kernel operator - comprising anisotropic large-kernel convolutions and mixed-domain attention - all within a computationally efficient Cross-Stage Partial (CSP) framework. On a large-scale dataset comprising 120 pollen classes, HieraEdgeNet achieves a mean Average Precision (mAP@.5) of 0.9501, significantly outperforming state-of-the-art baseline models such as YOLOv12n and RT-DETR. Furthermore, qualitative analysis confirms that our approach generates feature representations that are more precisely focused on object boundaries. By systematically integrating edge information, HieraEdgeNet provides a robust and powerful solution for high-precision, high-efficiency automated detection of microscopic objects.
Authors: Prarabdh Shukla, Wei Yin Chong, Yash Patel, Brennan Schaffner, Danish Pruthi, Arjun Bhagoji
Abstract: To meet the demands of content moderation, online platforms have resorted to automated systems. Newer forms of real-time engagement($\textit{e.g.}$, users commenting on live streams) on platforms like Twitch exert additional pressures on the latency expected of such moderation systems. Despite their prevalence, relatively little is known about the effectiveness of these systems. In this paper, we conduct an audit of Twitch's automated moderation tool ($\texttt{AutoMod}$) to investigate its effectiveness in flagging hateful content. For our audit, we create streaming accounts to act as siloed test beds, and interface with the live chat using Twitch's APIs to send over $107,000$ comments collated from $4$ datasets. We measure $\texttt{AutoMod}$'s accuracy in flagging blatantly hateful content containing misogyny, racism, ableism and homophobia. Our experiments reveal that a large fraction of hateful messages, up to $94\%$ on some datasets, $\textit{bypass moderation}$. Contextual addition of slurs to these messages results in $100\%$ removal, revealing $\texttt{AutoMod}$'s reliance on slurs as a moderation signal. We also find that contrary to Twitch's community guidelines, $\texttt{AutoMod}$ blocks up to $89.5\%$ of benign examples that use sensitive words in pedagogical or empowering contexts. Overall, our audit points to large gaps in $\texttt{AutoMod}$'s capabilities and underscores the importance for such systems to understand context effectively.
Authors: Kevin Lam, Thang Bui, George Deligiannidis, Yee Whye Teh
Abstract: Latent Gaussian variables have been popularised in probabilistic machine learning. In turn, gradient estimators are the machinery that facilitates gradient-based optimisation for models with latent Gaussian variables. The reparameterisation trick is often used as the default estimator as it is simple to implement and yields low-variance gradients for variational inference. In this work, we propose the R2-G2 estimator as the Rao-Blackwellisation of the reparameterisation gradient estimator. Interestingly, we show that the local reparameterisation gradient estimator for Bayesian MLPs is an instance of the R2-G2 estimator and Rao-Blackwellisation. This lets us extend benefits of Rao-Blackwellised gradients to a suite of probabilistic models. We show that initial training with R2-G2 consistently yields better performance in models with multiple applications of the reparameterisation trick.
Authors: Jiaming Li, Haoran Ye, Yukun Chen, Xinyue Li, Lei Zhang, Hamid Alinejad-Rokny, Jimmy Chih-Hsien Peng, Min Yang
Abstract: As large language models (LLMs) grow in scale and capability, understanding their internal mechanisms becomes increasingly critical. Sparse autoencoders (SAEs) have emerged as a key tool in mechanistic interpretability, enabling the extraction of human-interpretable features from LLMs. However, existing SAE training methods are primarily designed for base models, resulting in reduced reconstruction quality and interpretability when applied to instruct models. To bridge this gap, we propose $\underline{\textbf{F}}$inetuning-$\underline{\textbf{a}}$ligned $\underline{\textbf{S}}$equential $\underline{\textbf{T}}$raining ($\textit{FAST}$), a novel training method specifically tailored for instruct models. $\textit{FAST}$ aligns the training process with the data distribution and activation patterns characteristic of instruct models, resulting in substantial improvements in both reconstruction and feature interpretability. On Qwen2.5-7B-Instruct, $\textit{FAST}$ achieves a mean squared error of 0.6468 in token reconstruction, significantly outperforming baseline methods with errors of 5.1985 and 1.5096. In feature interpretability, $\textit{FAST}$ yields a higher proportion of high-quality features, for Llama3.2-3B-Instruct, $21.1\%$ scored in the top range, compared to $7.0\%$ and $10.2\%$ for $\textit{BT(P)}$ and $\textit{BT(F)}$. Surprisingly, we discover that intervening on the activations of special tokens via the SAEs leads to improvements in output quality, suggesting new opportunities for fine-grained control of model behavior. Code, data, and 240 trained SAEs are available at https://github.com/Geaming2002/FAST.
Authors: Parsa Miraghaei, Sergio Moreschini, Antti Kolehmainen, David H\"astbacka
Abstract: Background: The growing demand for efficient and deployable language models has led to increased interest in Small Language Models (SLMs). However, existing research remains fragmented, lacking a unified lifecycle perspective. Objective: This study aims to define a comprehensive lifecycle framework for SLMs by synthesizing insights from academic literature and practitioner sources. Method: We conducted a comprehensive survey of 36 works, analyzing and categorizing lifecycle-relevant techniques. Results: We propose a modular lifecycle model structured into main, optional, and cross-cutting components. The model captures key interconnections across stages, supporting method reuse, co-adaptation, and lifecycle-awareness. Conclusion: Our framework provides a coherent foundation for developing and maintaining SLMs, bridging theory and practice, and guiding future research and tool development.
Authors: Francesco Marchiori, Denis Donadel, Alessandro Brighente, Mauro Conti
Abstract: Electric Vehicles (EVs) are rapidly gaining adoption as a sustainable alternative to fuel-powered vehicles, making secure charging infrastructure essential. Despite traditional authentication protocols, recent results showed that attackers may steal energy through tailored relay attacks. One countermeasure is leveraging the EV's fingerprint on the current exchanged during charging. However, existing methods focus on the final charging stage, allowing malicious actors to consume substantial energy before being detected and repudiated. This underscores the need for earlier and more effective authentication methods to prevent unauthorized charging. Meanwhile, profiling raises privacy concerns, as uniquely identifying EVs through charging patterns could enable user tracking. In this paper, we propose a framework for uniquely identifying EVs using physical measurements from the early charging stages. We hypothesize that voltage behavior early in the process exhibits similar characteristics to current behavior in later stages. By extracting features from early voltage measurements, we demonstrate the feasibility of EV profiling. Our approach improves existing methods by enabling faster and more reliable vehicle identification. We test our solution on a dataset of 7408 usable charges from 49 EVs, achieving up to 0.86 accuracy. Feature importance analysis shows that near-optimal performance is possible with just 10 key features, improving efficiency alongside our lightweight models. This research lays the foundation for a novel authentication factor while exposing potential privacy risks from unauthorized access to charging data.
Authors: Mark Burgess
Abstract: Some formal aspects of the Semantic Spacetime graph model are presented, with reference to its use for directed knowledge representations and process modelling. A finite $\gamma(3,4)$ representation is defined to form a closed set of operations that can scale to any degree of semantic complexity. The Semantic Spacetime postulates bring predictability with minimal constraints to pathways in graphs. The ubiquitous appearance of absorbing states in any partial graph means that a graph process leaks information. The issue is closely associated with the issue of division by zero, which signals a loss of closure and the need for manual injection of remedial information. The Semantic Spacetime model (and its Promise Theory) origins help to clarify how such absorbing states are associated with boundary information where intentionality can enter.
Authors: Haijie Xu, Chen Zhang
Abstract: We propose an algorithm for change point monitoring in linear causal models that accounts for interventions. Through a special centralization technique, we can concentrate the changes arising from causal propagation across nodes into a single dimension. Additionally, by selecting appropriate intervention nodes based on Kullback-Leibler divergence, we can amplify the change magnitude. We also present an algorithm for selecting the intervention values, which aids in the identification of the most effective intervention nodes. Two monitoring methods are proposed, each with an adaptive intervention policy to make a balance between exploration and exploitation. We theoretically demonstrate the first-order optimality of the proposed methods and validate their properties using simulation datasets and two real-world case studies.
Authors: Yang Wang, Yin Xu, Cixiao Zhang, Zhiyong Chen, Xiaowu Ou, Mingzeng Dai, Meixia Tao, Wenjun Zhang
Abstract: This letter proposes a channel estimation method for reconfigurable intelligent surface (RIS)-assisted systems through a novel diffusion model (DM) framework. We reformulate the channel estimation problem as a denoising process, which aligns with the reverse process of the DM. To overcome the inherent randomness in the reverse process of conventional DM approaches, we adopt a deterministic sampling strategy with a step alignment mechanism that ensures the accuracy of channel estimation while adapting to different signal-to-noise ratio (SNR). Furthermore, to reduce the number of parameters of the U-Net, we meticulously design a lightweight network that achieves comparable performance, thereby enhancing the practicality of our proposed method. Extensive simulations demonstrate superior performance over a wide range of SNRs compared to baselines. For instance, the proposed method achieves performance improvements of up to 13.5 dB in normalized mean square error (NMSE) at SNR = 0 dB. Notably, the proposed lightweight network exhibits almost no performance loss compared to the original U-Net, while requiring only 6.59\% of its parameters.
Authors: Mohamed Djilani, Nassim Ali Ousalah, Nidhal Eddine Chenni
Abstract: We introduce a trend-aware and visually-grounded fashion recommendation system that integrates deep visual representations, garment-aware segmentation, semantic category similarity and user behavior simulation. Our pipeline extracts focused visual embeddings by masking non-garment regions via semantic segmentation followed by feature extraction using pretrained CNN backbones (ResNet-50, DenseNet-121, VGG16). To simulate realistic shopping behavior, we generate synthetic purchase histories influenced by user-specific trendiness and item popularity. Recommendations are computed using a weighted scoring function that fuses visual similarity, semantic coherence and popularity alignment. Experiments on the DeepFashion dataset demonstrate consistent gender alignment and improved category relevance, with ResNet-50 achieving 64.95% category similarity and lowest popularity MAE. An ablation study confirms the complementary roles of visual and popularity cues. Our method provides a scalable framework for personalized fashion recommendations that balances individual style with emerging trends. Our implementation is available at https://github.com/meddjilani/FashionRecommender
Authors: Qi Yang, Chenghao Zhang, Lubin Fan, Kun Ding, Jieping Ye, Shiming Xiang
Abstract: Recent advancements in Large Vision Language Models (LVLMs) have significantly improved performance in Visual Question Answering (VQA) tasks through multimodal Retrieval-Augmented Generation (RAG). However, existing methods still face challenges, such as the scarcity of knowledge with reasoning examples and erratic responses from retrieved knowledge. To address these issues, in this study, we propose a multimodal RAG framework, termed RCTS, which enhances LVLMs by constructing a Reasoning Context-enriched knowledge base and a Tree Search re-ranking method. Specifically, we introduce a self-consistent evaluation mechanism to enrich the knowledge base with intrinsic reasoning patterns. We further propose a Monte Carlo Tree Search with Heuristic Rewards (MCTS-HR) to prioritize the most relevant examples. This ensures that LVLMs can leverage high-quality contextual reasoning for better and more consistent responses. Extensive experiments demonstrate that our framework achieves state-of-the-art performance on multiple VQA datasets, significantly outperforming In-Context Learning (ICL) and Vanilla-RAG methods. It highlights the effectiveness of our knowledge base and re-ranking method in improving LVLMs. Our code is available at https://github.com/yannqi/RCTS-RAG.
Authors: Xiaotian Ye, Mengqi Zhang, Shu Wu
Abstract: Large Language Model (LLM) unlearning aims to erase or suppress undesirable knowledge within the model, offering promise for controlling harmful or private information to prevent misuse. However, recent studies highlight its limited efficacy in real-world scenarios, hindering practical adoption. In this study, we identify a pervasive issue underlying many downstream failures: the effectiveness of existing unlearning methods heavily depends on the form of training samples and frequently fails to generalize to alternate expressions of the same knowledge. We formally characterize this problem as Form-Dependent Bias and systematically investigate its specific manifestation patterns across various downstream tasks. To quantify its prevalence and support future research, we introduce ORT, a novel benchmark designed to evaluate the robustness of unlearning methods against variations in knowledge expression. Results reveal that Form-Dependent Bias is both widespread and severe among current techniques. We argue that LLM unlearning should be form-independent to address the endless forms of downstream tasks encountered in real-world security-critical scenarios. Towards this goal, we introduce Rank-one Concept Redirection (ROCR), a novel training-free method, as a promising solution path. ROCR performs unlearning by targeting the invariants in downstream tasks, specifically the activated dangerous concepts. It is capable of modifying model parameters within seconds to redirect the model's perception of a specific unlearning target concept to another harmless concept. Extensive experiments demonstrate that ROCR significantly improves unlearning effectiveness compared to traditional methods while generating highly natural outputs.
Authors: Iustin Sirbu (National University of Science and Technology POLITEHNICA Bucharest), Robert-Adrian Popovici (National University of Science and Technology POLITEHNICA Bucharest), Cornelia Caragea (University of Illinois Chicago), Stefan Trausan-Matu (National University of Science and Technology POLITEHNICA Bucharest), Traian Rebedea (National University of Science and Technology POLITEHNICA Bucharest, NVIDIA)
Abstract: We introduce MultiMatch, a novel semi-supervised learning (SSL) algorithm combining the paradigms of co-training and consistency regularization with pseudo-labeling. At its core, MultiMatch features a three-fold pseudo-label weighting module designed for three key purposes: selecting and filtering pseudo-labels based on head agreement and model confidence, and weighting them according to the perceived classification difficulty. This novel module enhances and unifies three existing techniques -- heads agreement from Multihead Co-training, self-adaptive thresholds from FreeMatch, and Average Pseudo-Margins from MarginMatch -- resulting in a holistic approach that improves robustness and performance in SSL settings. Experimental results on benchmark datasets highlight the superior performance of MultiMatch, achieving state-of-the-art results on 9 out of 10 setups from 5 natural language processing datasets and ranking first according to the Friedman test among 19 methods. Furthermore, MultiMatch demonstrates exceptional robustness in highly imbalanced settings, outperforming the second-best approach by 3.26% -- and data imbalance is a key factor for many text classification tasks.
Authors: Emiliano Tolotti, Enrico Blanzieri, Davide Pastorello
Abstract: Ensemble methods in machine learning aim to improve prediction accuracy by combining multiple models. This is achieved by ensuring diversity among predictors to capture different data aspects. Homogeneous ensembles use identical models, achieving diversity through different data subsets, and weighted-average ensembles assign higher influence to more accurate models through a weight learning procedure. We propose a method to achieve a weighted homogeneous quantum ensemble using quantum classifiers with indexing registers for data encoding. This approach leverages instance-based quantum classifiers, enabling feature and training point subsampling through superposition and controlled unitaries, and allowing for a quantum-parallel execution of diverse internal classifiers with different data compositions in superposition. The method integrates a learning process involving circuit execution and classical weight optimization, for a trained ensemble execution with weights encoded in the circuit at test-time. Empirical evaluation demonstrate the effectiveness of the proposed method, offering insights into its performance.
Authors: Yingli Wang, Changwei Tu, Xiaoyu Wang, Lingjiong Zhu
Abstract: The problem of sampling a target probability distribution on a constrained domain arises in many applications including machine learning. For constrained sampling, various Langevin algorithms such as projected Langevin Monte Carlo (PLMC) based on the discretization of reflected Langevin dynamics (RLD) and more generally skew-reflected non-reversible Langevin Monte Carlo (SRNLMC) based on the discretization of skew-reflected non-reversible Langevin dynamics (SRNLD) have been proposed and studied in the literature. This work focuses on the long-time behavior of SRNLD, where a skew-symmetric matrix is added to RLD. Although the non-asymptotic convergence analysis for SRNLD (and SRNLMC) and the acceleration compared to RLD (and PMLC) have been studied in the literature, it is not clear how one should design the skew-symmetric matrix in the dynamics to achieve good performance in practice. We establish a large deviation principle (LDP) for the empirical measure of SRNLD when the skew-symmetric matrix is chosen such that its product with the inward unit normal vector field on the boundary is zero. By explicitly characterizing the rate functions, we show that SRNLD can accelerate the convergence to the target distribution compared to RLD with this choice of the skew-symmetric matrix. Numerical experiments for SRNLMC based on the proposed skew-symmetric matrix show superior performance which validate the theoretical findings from the large deviations theory.
Authors: William Ljungbergh, Bernardo Taveira, Wenzhao Zheng, Adam Tonderski, Chensheng Peng, Fredrik Kahl, Christoffer Petersson, Michael Felsberg, Kurt Keutzer, Masayoshi Tomizuka, Wei Zhan
Abstract: Validating autonomous driving (AD) systems requires diverse and safety-critical testing, making photorealistic virtual environments essential. Traditional simulation platforms, while controllable, are resource-intensive to scale and often suffer from a domain gap with real-world data. In contrast, neural reconstruction methods like 3D Gaussian Splatting (3DGS) offer a scalable solution for creating photorealistic digital twins of real-world driving scenes. However, they struggle with dynamic object manipulation and reusability as their per-scene optimization-based methodology tends to result in incomplete object models with integrated illumination effects. This paper introduces R3D2, a lightweight, one-step diffusion model designed to overcome these limitations and enable realistic insertion of complete 3D assets into existing scenes by generating plausible rendering effects-such as shadows and consistent lighting-in real time. This is achieved by training R3D2 on a novel dataset: 3DGS object assets are generated from in-the-wild AD data using an image-conditioned 3D generative model, and then synthetically placed into neural rendering-based virtual environments, allowing R3D2 to learn realistic integration. Quantitative and qualitative evaluations demonstrate that R3D2 significantly enhances the realism of inserted assets, enabling use-cases like text-to-3D asset insertion and cross-scene/dataset object transfer, allowing for true scalability in AD validation. To promote further research in scalable and realistic AD simulation, we will release our dataset and code, see https://research.zenseact.com/publications/R3D2/.
Authors: Mingzhou Liu, Xinwei Sun, Yizhou Wang
Abstract: In this note, we extend the conditional local independence testing theory developed in Christgau et al. (2024) to Ito processes. The result can be applied to causal discovery in dynamic systems.
Authors: Zihui Zhang, Weisheng Dai, Hongtao Wen, Bo Yang
Abstract: We study the problem of unsupervised 3D semantic segmentation on raw point clouds without needing human labels in training. Existing methods usually formulate this problem into learning per-point local features followed by a simple grouping strategy, lacking the ability to discover additional and possibly richer semantic priors beyond local features. In this paper, we introduce LogoSP to learn 3D semantics from both local and global point features. The key to our approach is to discover 3D semantic information by grouping superpoints according to their global patterns in the frequency domain, thus generating highly accurate semantic pseudo-labels for training a segmentation network. Extensive experiments on two indoor and an outdoor datasets show that our LogoSP surpasses all existing unsupervised methods by large margins, achieving the state-of-the-art performance for unsupervised 3D semantic segmentation. Notably, our investigation into the learned global patterns reveals that they truly represent meaningful 3D semantics in the absence of human labels during training.
Authors: Amanuel Anteneh, L\'eandre Brunel, Carlos Gonz\'alez-Arciniegas, Olivier Pfister
Abstract: Cubic-phase states are a sufficient resource for universal quantum computing over continuous variables. We present results from numerical experiments in which deep neural networks are trained via reinforcement learning to control a quantum optical circuit for generating cubic-phase states, with an average success rate of 96%. The only non-Gaussian resource required is photon-number-resolving measurements. We also show that the exact same resources enable the direct generation of a quartic-phase gate, with no need for a cubic gate decomposition.
Authors: Lev Novitskiy, Viacheslav Vasilev, Maria Kovaleva, Vladimir Arkhipkin, Denis Dimitrov
Abstract: Variational Autoencoders (VAEs) remain a cornerstone of generative computer vision, yet their training is often plagued by artifacts that degrade reconstruction and generation quality. This paper introduces VIVAT, a systematic approach to mitigating common artifacts in KL-VAE training without requiring radical architectural changes. We present a detailed taxonomy of five prevalent artifacts - color shift, grid patterns, blur, corner and droplet artifacts - and analyze their root causes. Through straightforward modifications, including adjustments to loss weights, padding strategies, and the integration of Spatially Conditional Normalization, we demonstrate significant improvements in VAE performance. Our method achieves state-of-the-art results in image reconstruction metrics (PSNR and SSIM) across multiple benchmarks and enhances text-to-image generation quality, as evidenced by superior CLIP scores. By preserving the simplicity of the KL-VAE framework while addressing its practical challenges, VIVAT offers actionable insights for researchers and practitioners aiming to optimize VAE training.
Authors: Jinxi Li, Ziyang Song, Siyuan Zhou, Bo Yang
Abstract: In this paper, we aim to model 3D scene geometry, appearance, and the underlying physics purely from multi-view videos. By applying various governing PDEs as PINN losses or incorporating physics simulation into neural networks, existing works often fail to learn complex physical motions at boundaries or require object priors such as masks or types. In this paper, we propose FreeGave to learn the physics of complex dynamic 3D scenes without needing any object priors. The key to our approach is to introduce a physics code followed by a carefully designed divergence-free module for estimating a per-Gaussian velocity field, without relying on the inefficient PINN losses. Extensive experiments on three public datasets and a newly collected challenging real-world dataset demonstrate the superior performance of our method for future frame extrapolation and motion segmentation. Most notably, our investigation into the learned physics codes reveals that they truly learn meaningful 3D physical motion patterns in the absence of any human labels in training.
Authors: Rui Wen, Yiyong Liu, Michael Backes, Yang Zhang
Abstract: Data reconstruction attacks, which aim to recover the training dataset of a target model with limited access, have gained increasing attention in recent years. However, there is currently no consensus on a formal definition of data reconstruction attacks or appropriate evaluation metrics for measuring their quality. This lack of rigorous definitions and universal metrics has hindered further advancement in this field. In this paper, we address this issue in the vision domain by proposing a unified attack taxonomy and formal definitions of data reconstruction attacks. We first propose a set of quantitative evaluation metrics that consider important criteria such as quantifiability, consistency, precision, and diversity. Additionally, we leverage large language models (LLMs) as a substitute for human judgment, enabling visual evaluation with an emphasis on high-quality reconstructions. Using our proposed taxonomy and metrics, we present a unified framework for systematically evaluating the strengths and limitations of existing attacks and establishing a benchmark for future research. Empirical results, primarily from a memorization perspective, not only validate the effectiveness of our metrics but also offer valuable insights for designing new attacks.
Authors: Shuja Khalid, Mohamed Ibrahim, Yang Liu
Abstract: We present a novel approach for enhancing the resolution and geometric fidelity of 3D Gaussian Splatting (3DGS) beyond native training resolution. Current 3DGS methods are fundamentally limited by their input resolution, producing reconstructions that cannot extrapolate finer details than are present in the training views. Our work breaks this limitation through a lightweight generative model that predicts and refines additional 3D Gaussians where needed most. The key innovation is our Hessian-assisted sampling strategy, which intelligently identifies regions that are likely to benefit from densification, ensuring computational efficiency. Unlike computationally intensive GANs or diffusion approaches, our method operates in real-time (0.015s per inference on a single consumer-grade GPU), making it practical for interactive applications. Comprehensive experiments demonstrate significant improvements in both geometric accuracy and rendering quality compared to state-of-the-art methods, establishing a new paradigm for resolution-free 3D scene enhancement.
Authors: Ke Wang, Yiming Qin, Nikolaos Dimitriadis, Alessandro Favero, Pascal Frossard
Abstract: Language models deployed in real-world systems often require post-hoc updates to incorporate new or corrected knowledge. However, editing such models efficiently and reliably - without retraining or forgetting previous information - remains a major challenge. Existing methods for lifelong model editing either compromise generalization, interfere with past edits, or fail to scale to long editing sequences. We propose MEMOIR, a novel scalable framework that injects knowledge through a residual memory, i.e., a dedicated parameter module, while preserving the core capabilities of the pre-trained model. By sparsifying input activations through sample-dependent masks, MEMOIR confines each edit to a distinct subset of the memory parameters, minimizing interference among edits. At inference, it identifies relevant edits by comparing the sparse activation patterns of new queries to those stored during editing. This enables generalization to rephrased queries by activating only the relevant knowledge while suppressing unnecessary memory activation for unrelated prompts. Experiments on question answering, hallucination correction, and out-of-distribution generalization benchmarks across LLaMA-3 and Mistral demonstrate that MEMOIR achieves state-of-the-art performance across reliability, generalization, and locality metrics, scaling to thousands of sequential edits with minimal forgetting.
Authors: Yaxita Amin, Naimisha S Trivedi, Rashmi Bhattad
Abstract: Remote sensing change detection is essential for monitoring the everchanging landscapes of the Earth. The U-Net architecture has gained popularity for its capability to capture spatial information and perform pixel-wise classification. However, their application in the Remote sensing field remains largely unexplored. Therefore, this paper fill the gap by conducting a comprehensive analysis of 34 papers. This study conducts a comparison and analysis of 18 different U-Net variations, assessing their potential for detecting changes in remote sensing. We evaluate both benefits along with drawbacks of each variation within the framework of this particular application. We emphasize variations that are explicitly built for change detection, such as Siamese Swin-U-Net, which utilizes a Siamese architecture. The analysis highlights the significance of aspects such as managing data from different time periods and collecting relationships over a long distance to enhance the precision of change detection. This study provides valuable insights for researchers and practitioners that choose U-Net versions for remote sensing change detection tasks.
Authors: Jiayi Sheng, Luna Lyu, Jikai Jin, Tony Xia, Alex Gu, James Zou, Pan Lu
Abstract: Inequality proving, crucial across diverse scientific and mathematical fields, tests advanced reasoning skills such as discovering tight bounds and strategic theorem application. This makes it a distinct, demanding frontier for large language models (LLMs), offering insights beyond general mathematical problem-solving. Progress in this area is hampered by existing datasets that are often scarce, synthetic, or rigidly formal. We address this by proposing an informal yet verifiable task formulation, recasting inequality proving into two automatically checkable subtasks: bound estimation and relation prediction. Building on this, we release IneqMath, an expert-curated dataset of Olympiad-level inequalities, including a test set and training corpus enriched with step-wise solutions and theorem annotations. We also develop a novel LLM-as-judge evaluation framework, combining a final-answer judge with four step-wise judges designed to detect common reasoning flaws. A systematic evaluation of 29 leading LLMs on IneqMath reveals a surprising reality: even top models like o1 achieve less than 10% overall accuracy under step-wise scrutiny; this is a drop of up to 65.5% from their accuracy considering only final answer equivalence. This discrepancy exposes fragile deductive chains and a critical gap for current LLMs between merely finding an answer and constructing a rigorous proof. Scaling model size and increasing test-time computation yield limited gains in overall proof correctness. Instead, our findings highlight promising research directions such as theorem-guided reasoning and self-refinement. Code and data are available at https://ineqmath.github.io/.
Authors: Rishit Dagli, Yushi Guan, Sankeerth Durvasula, Mohammadreza Mofayezi, Nandita Vijaykumar
Abstract: We propose Squeeze3D, a novel framework that leverages implicit prior knowledge learnt by existing pre-trained 3D generative models to compress 3D data at extremely high compression ratios. Our approach bridges the latent spaces between a pre-trained encoder and a pre-trained generation model through trainable mapping networks. Any 3D model represented as a mesh, point cloud, or a radiance field is first encoded by the pre-trained encoder and then transformed (i.e. compressed) into a highly compact latent code. This latent code can effectively be used as an extremely compressed representation of the mesh or point cloud. A mapping network transforms the compressed latent code into the latent space of a powerful generative model, which is then conditioned to recreate the original 3D model (i.e. decompression). Squeeze3D is trained entirely on generated synthetic data and does not require any 3D datasets. The Squeeze3D architecture can be flexibly used with existing pre-trained 3D encoders and existing generative models. It can flexibly support different formats, including meshes, point clouds, and radiance fields. Our experiments demonstrate that Squeeze3D achieves compression ratios of up to 2187x for textured meshes, 55x for point clouds, and 619x for radiance fields while maintaining visual quality comparable to many existing methods. Squeeze3D only incurs a small compression and decompression latency since it does not involve training object-specific networks to compress an object.
Authors: Chengyue Huang, Yuchen Zhu, Sichen Zhu, Jingyun Xiao, Moises Andrade, Shivang Chopra, Zsolt Kira
Abstract: Vision-language models (VLMs) are widely assumed to exhibit in-context learning (ICL), a property similar to that of their language-only counterparts. While recent work suggests VLMs can perform multimodal ICL (MM-ICL), studies show they often rely on shallow heuristics -- such as copying or majority voting -- rather than true task understanding. We revisit this assumption by evaluating VLMs under distribution shifts, where support examples come from a dataset different from the query. Surprisingly, performance often degrades with more demonstrations, and models tend to copy answers rather than learn from them. To investigate further, we propose a new MM-ICL with Reasoning pipeline that augments each demonstration with a generated rationale alongside the answer. We conduct extensive and comprehensive experiments on both perception- and reasoning-required datasets with open-source VLMs ranging from 3B to 72B and proprietary models such as Gemini 2.0. We conduct controlled studies varying shot count, retrieval method, rationale quality, and distribution. Our results show limited performance sensitivity across these factors, suggesting that current VLMs do not effectively utilize demonstration-level information as intended in MM-ICL.
Authors: Christopher Subia-Waud (Rayonlabs Team)
Abstract: Foundation model fine-tuning faces a fundamental challenge: existing AutoML platforms rely on single optimisation strategies that explore only a fraction of viable hyperparameter configurations. In this white paper, We introduce Gradients, a decentralised AutoML platform that transforms hyperparameter optimisation into a competitive marketplace where independent miners compete to discover optimal configurations. Economic incentives align individual exploration with collective optimisation goals, driving systematic investigation of hyperparameter regions that centralised methods miss. We evaluate our approach across 180 controlled experiments spanning diverse model architectures (70M to 70B parameters) and task types. Gradients achieves an 82.8\% win rate against HuggingFace AutoTrain and 100\% against TogetherAI, Databricks, and Google Cloud, with mean improvements of 11.8\% and 42.1\% respectively. Complex reasoning and retrieval tasks show particularly strong gains of 30-40\%, whilst diffusion models achieve 23.4\% improvements for person-specific generation. These results demonstrate that competitive, economically-driven approaches can systematically discover superior configurations that centralised AutoML consistently miss.
Authors: George Orfanides, Tim Hoheisel, Marwa El Halabi
Abstract: Submodular functions, defined on continuous or discrete domains, arise in numerous applications. We study the minimization of the difference of two submodular (DS) functions, over both domains, extending prior work restricted to set functions. We show that all functions on discrete domains and all smooth functions on continuous domains are DS. For discrete domains, we observe that DS minimization is equivalent to minimizing the difference of two convex (DC) functions, as in the set function case. We propose a novel variant of the DC Algorithm (DCA) and apply it to the resulting DC Program, obtaining comparable theoretical guarantees as in the set function case. The algorithm can be applied to continuous domains via discretization. Experiments demonstrate that our method outperforms baselines in integer compressive sensing and integer least squares.
Authors: Tim Vieira, Tianyu Liu, Clemente Pasti, Yahya Emara, Brian DuSell, Benjamin LeBrun, Mario Giulianelli, Juan Luis Gastaldi, Timothy J. O'Donnell, Ryan Cotterell
Abstract: Modern language models represent probability distributions over character strings as distributions over (shorter) token strings derived via a deterministic tokenizer, such as byte-pair encoding. While this approach is highly effective at scaling up language models to large corpora, its current incarnations have a concerning property: the model assigns nonzero probability mass to an exponential number of $\it{noncanonical}$ token encodings of each character string -- these are token strings that decode to valid character strings but are impossible under the deterministic tokenizer (i.e., they will never be seen in any training corpus, no matter how large). This misallocation is both erroneous, as noncanonical strings never appear in training data, and wasteful, diverting probability mass away from plausible outputs. These are avoidable mistakes! In this work, we propose methods to enforce canonicality in token-level language models, ensuring that only canonical token strings are assigned positive probability. We present two approaches: (1) canonicality by conditioning, leveraging test-time inference strategies without additional training, and (2) canonicality by construction, a model parameterization that guarantees canonical outputs but requires training. We demonstrate that fixing canonicality mistakes improves the likelihood of held-out data for several models and corpora.
Authors: Dmitrii Vorobev, Artem Prosvetov, Karim Elhadji Daou
Abstract: We propose a computationally efficient method for real-time three-dimensional football trajectory reconstruction from a single broadcast camera. In contrast to previous work, our approach introduces a multi-mode state model with $W$ discrete modes to significantly accelerate optimization while preserving centimeter-level accuracy -- even in cases of severe occlusion, motion blur, and complex backgrounds. The system operates on standard CPUs and achieves low latency suitable for live broadcast settings. Extensive evaluation on a proprietary dataset of 6K-resolution Russian Premier League matches demonstrates performance comparable to multi-camera systems, without the need for specialized or costly infrastructure. This work provides a practical method for accessible and accurate 3D ball tracking in professional football environments.
Authors: Mingquan Lin, Gregory Holste, Song Wang, Yiliang Zhou, Yishu Wei, Imon Banerjee, Pengyi Chen, Tianjie Dai, Yuexi Du, Nicha C. Dvornek, Yuyan Ge, Zuowei Guo, Shouhei Hanaoka, Dongkyun Kim, Pablo Messina, Yang Lu, Denis Parra, Donghyun Son, \'Alvaro Soto, Aisha Urooj, Ren\'e Vidal, Yosuke Yamagishi, Zefan Yang, Ruichi Zhang, Yang Zhou, Leo Anthony Celi, Ronald M. Summers, Zhiyong Lu, Hao Chen, Adam Flanders, George Shih, Zhangyang Wang, Yifan Peng
Abstract: The CXR-LT series is a community-driven initiative designed to enhance lung disease classification using chest X-rays (CXR). It tackles challenges in open long-tailed lung disease classification and enhances the measurability of state-of-the-art techniques. The first event, CXR-LT 2023, aimed to achieve these goals by providing high-quality benchmark CXR data for model development and conducting comprehensive evaluations to identify ongoing issues impacting lung disease classification performance. Building on the success of CXR-LT 2023, the CXR-LT 2024 expands the dataset to 377,110 chest X-rays (CXRs) and 45 disease labels, including 19 new rare disease findings. It also introduces a new focus on zero-shot learning to address limitations identified in the previous event. Specifically, CXR-LT 2024 features three tasks: (i) long-tailed classification on a large, noisy test set, (ii) long-tailed classification on a manually annotated "gold standard" subset, and (iii) zero-shot generalization to five previously unseen disease findings. This paper provides an overview of CXR-LT 2024, detailing the data curation process and consolidating state-of-the-art solutions, including the use of multimodal models for rare disease detection, advanced generative approaches to handle noisy labels, and zero-shot learning strategies for unseen diseases. Additionally, the expanded dataset enhances disease coverage to better represent real-world clinical settings, offering a valuable resource for future research. By synthesizing the insights and innovations of participating teams, we aim to advance the development of clinically realistic and generalizable diagnostic models for chest radiography.
Authors: Tuomas Oikarinen, Ge Yan, Akshay Kulkarni, Tsui-Wei Weng
Abstract: Interpreting individual neurons or directions in activations space is an important component of mechanistic interpretability. As such, many algorithms have been proposed to automatically produce neuron explanations, but it is often not clear how reliable these explanations are, or which methods produce the best explanations. This can be measured via crowd-sourced evaluations, but they can often be noisy and expensive, leading to unreliable results. In this paper, we carefully analyze the evaluation pipeline and develop a cost-effective and highly accurate crowdsourced evaluation strategy. In contrast to previous human studies that only rate whether the explanation matches the most highly activating inputs, we estimate whether the explanation describes neuron activations across all inputs. To estimate this effectively, we introduce a novel application of importance sampling to determine which inputs are the most valuable to show to raters, leading to around 30x cost reduction compared to uniform sampling. We also analyze the label noise present in crowd-sourced evaluations and propose a Bayesian method to aggregate multiple ratings leading to a further ~5x reduction in number of ratings required for the same accuracy. Finally, we use these methods to conduct a large-scale study comparing the quality of neuron explanations produced by the most popular methods for two different vision models.
Authors: Junhao Chen, Yulia Tsvetkov, Xiaochuang Han
Abstract: Recent progress in multimodal generation has increasingly combined autoregressive (AR) and diffusion-based approaches, leveraging their complementary strengths: AR models capture long-range dependencies and produce fluent, context-aware outputs, while diffusion models operate in continuous latent spaces to refine high-fidelity visual details. However, existing hybrids often lack systematic guidance on how and why to allocate model capacity between these paradigms. In this work, we introduce MADFormer, a Mixed Autoregressive and Diffusion Transformer that serves as a testbed for analyzing AR-diffusion trade-offs. MADFormer partitions image generation into spatial blocks, using AR layers for one-pass global conditioning across blocks and diffusion layers for iterative local refinement within each block. Through controlled experiments on FFHQ-1024 and ImageNet, we identify two key insights: (1) block-wise partitioning significantly improves performance on high-resolution images, and (2) vertically mixing AR and diffusion layers yields better quality-efficiency balances--improving FID by up to 75% under constrained inference compute. Our findings offer practical design principles for future hybrid generative models.
Authors: Stephanie Fu, Tyler Bonnen, Devin Guillory, Trevor Darrell
Abstract: Language provides a natural interface to specify and evaluate performance on visual tasks. To realize this possibility, vision language models (VLMs) must successfully integrate visual and linguistic information. Our work compares VLMs to a direct readout of their visual encoders to understand their ability to integrate across these modalities. Across a series of vision-centric benchmarks (e.g., depth estimation, correspondence), we find that VLMs perform substantially worse than their visual encoders, dropping to near-chance performance. We investigate these results through a series of analyses across the entire VLM: namely 1) the degradation of vision representations, 2) brittleness to task prompt, and 3) the language model's role in solving the task. We find that the bottleneck in performing these vision-centric tasks lies in this third category; VLMs are not effectively using visual information easily accessible throughout the entire model, and they inherit the language priors present in the LLM. Our work helps diagnose the failure modes of open-source VLMs, and presents a series of evaluations useful for future investigations into visual understanding within VLMs.
Authors: Xun Huang, Zhengqi Li, Guande He, Mingyuan Zhou, Eli Shechtman
Abstract: We introduce Self Forcing, a novel training paradigm for autoregressive video diffusion models. It addresses the longstanding issue of exposure bias, where models trained on ground-truth context must generate sequences conditioned on their own imperfect outputs during inference. Unlike prior methods that denoise future frames based on ground-truth context frames, Self Forcing conditions each frame's generation on previously self-generated outputs by performing autoregressive rollout with key-value (KV) caching during training. This strategy enables supervision through a holistic loss at the video level that directly evaluates the quality of the entire generated sequence, rather than relying solely on traditional frame-wise objectives. To ensure training efficiency, we employ a few-step diffusion model along with a stochastic gradient truncation strategy, effectively balancing computational cost and performance. We further introduce a rolling KV cache mechanism that enables efficient autoregressive video extrapolation. Extensive experiments demonstrate that our approach achieves real-time streaming video generation with sub-second latency on a single GPU, while matching or even surpassing the generation quality of significantly slower and non-causal diffusion models. Project website: http://self-forcing.github.io/
Authors: Anh-Quan Cao, Ivan Lopes, Raoul de Charette
Abstract: Multi-task learning for dense prediction is limited by the need for extensive annotation for every task, though recent works have explored training with partial task labels. Leveraging the generalization power of diffusion models, we extend the partial learning setup to a zero-shot setting, training a multi-task model on multiple synthetic datasets, each labeled for only a subset of tasks. Our method, StableMTL, repurposes image generators for latent regression. Adapting a denoising framework with task encoding, per-task conditioning and a tailored training scheme. Instead of per-task losses requiring careful balancing, a unified latent loss is adopted, enabling seamless scaling to more tasks. To encourage inter-task synergy, we introduce a multi-stream model with a task-attention mechanism that converts N-to-N task interactions into efficient 1-to-N attention, promoting effective cross-task sharing. StableMTL outperforms baselines on 7 tasks across 8 benchmarks.
Authors: Yang Zheng, Eli Shlizerman
Abstract: Random Recurrent Neural Networks (RRNN) are the simplest recurrent networks to model and extract features from sequential data. The simplicity however comes with a price; RRNN are known to be susceptible to diminishing/exploding gradient problem when trained with gradient-descent based optimization. To enhance robustness of RRNN, alternative training approaches have been proposed. Specifically, FORCE learning approach proposed a recursive least squares alternative to train RRNN and was shown to be applicable even for the challenging task of target-learning, where the network is tasked with generating dynamic patterns with no guiding input. While FORCE training indicates that solving target-learning is possible, it appears to be effective only in a specific regime of network dynamics (edge-of-chaos). We thereby investigate whether initialization of RRNN connectivity according to a tailored distribution can guarantee robust FORCE learning. We are able to generate such distribution by inference of four generating principles constraining the spectrum of the network Jacobian to remain in stability region. This initialization along with FORCE learning provides a robust training method, i.e., Robust-FORCE (R-FORCE). We validate R-FORCE performance on various target functions for a wide range of network configurations and compare with alternative methods. Our experiments indicate that R-FORCE facilitates significantly more stable and accurate target-learning for a wide class of RRNN. Such stability becomes critical in modeling multi-dimensional sequences as we demonstrate on modeling time-series of human body joints during physical movements.
Authors: Santiago Gonzalez, Xin Qiu, Risto Miikkulainen
Abstract: Evolutionary computation can be used to optimize several different aspects of neural network architectures. For instance, the TaylorGLO method discovers novel, customized loss functions, resulting in improved performance, faster training, and improved data utilization. A likely reason is that such functions discourage overfitting, leading to effective regularization. This paper demonstrates theoretically that this is indeed the case for TaylorGLO. Learning rule decomposition reveals that evolved loss functions balance two factors: the pull toward zero error, and a push away from it to avoid overfitting. This is a general principle that may be used to understand other regularization techniques as well (as demonstrated in this paper for label smoothing). The theoretical analysis leads to a constraint that can be utilized to find more effective loss functions in practice; the mechanism also results in networks that are more robust (as demonstrated in this paper with adversarial inputs). The analysis in this paper thus constitutes a first step towards understanding regularization, and demonstrates the power of evolutionary neural architecture search in general.
Authors: Mohamad Kazem Shirani Faradonbeh, Sadegh Shirani, Mohsen Bayati
Abstract: Linear diffusion processes serve as canonical continuous-time models for dynamic decision-making under uncertainty. These systems evolve according to drift matrices that specify the instantaneous rates of change in the expected system state, while also experiencing continuous random disturbances modeled by Brownian noise. For instance, in medical applications such as artificial pancreas systems, the drift matrices represent the internal dynamics of glucose concentrations. Classical results in stochastic control provide optimal policies under perfect knowledge of the drift matrices. However, practical decision-making scenarios typically feature uncertainty about the drift; in medical contexts, such parameters are patient-specific and unknown, requiring adaptive policies for efficiently learning the drift matrices while ensuring system stability and optimal performance. We study the Thompson sampling (TS) algorithm for decision-making in linear diffusion processes with unknown drift matrices. For this algorithm that designs control policies as if samples from a posterior belief about the parameters fully coincide with the unknown truth, we establish efficiency. That is, Thompson sampling learns optimal control actions fast, incurring only a square-root of time regret, and also learns to stabilize the system in a short time period. To our knowledge, this is the first such result for TS in a diffusion process control problem. Moreover, our empirical simulations in three settings that involve blood-glucose and flight control demonstrate that TS significantly improves regret, compared to the state-of-the-art algorithms, suggesting it explores in a more guarded fashion. Our theoretical analysis includes characterization of a certain optimality manifold that relates the geometry of the drift matrices to the optimal control of the diffusion process, among others.
Authors: Artavazd Maranjyan, Mher Safaryan, Peter Richt\'arik
Abstract: We study a class of distributed optimization algorithms that aim to alleviate high communication costs by allowing clients to perform multiple local gradient-type training steps before communication. In a recent breakthrough, Mishchenko et al. (2022) proved that local training, when properly executed, leads to provable communication acceleration, and this holds in the strongly convex regime without relying on any data similarity assumptions. However, their ProxSkip method requires all clients to take the same number of local training steps in each communication round. We propose a redesign of the ProxSkip method, allowing clients with ``less important'' data to get away with fewer local training steps without impacting the overall communication complexity of the method. In particular, we prove that our modified method, GradSkip, converges linearly under the same assumptions and has the same accelerated communication complexity, while the number of local gradient steps can be reduced relative to a local condition number. We further generalize our method by extending the randomness of probabilistic alternations to arbitrary unbiased compression operators and by considering a generic proximable regularizer. This generalization, which we call GradSkip+, recovers several related methods in the literature as special cases. Finally, we present an empirical study on carefully designed toy problems that confirm our theoretical claims.
Authors: Zhaolin Ren, Tongzheng Ren, Haitong Ma, Na Li, Bo Dai
Abstract: This paper proposes an approach, Spectral Dynamics Embedding Control (SDEC), to optimal control for nonlinear stochastic systems. This method reveals an infinite-dimensional feature representation induced by the system's nonlinear stochastic dynamics, enabling a linear representation of the state-action value function. For practical implementation, this representation is approximated using finite-dimensional trucations, specifically via two prominent kernel approximation methods: random feature truncation and Nystrom approximation. To characterize the effectiveness of these approximations, we provide an in-depth theoretical analysis to characterize the approximation error arising from the finite-dimension truncation and statistical error due to finite-sample approximation in both policy evaluation and policy optimization. Empirically, our algorithm performs favorably against existing stochastic control algorithms on several benchmark problems.
Authors: Yulin Zhu, Xing Ai, Yevgeniy Vorobeychik, Kai Zhou
Abstract: We investigate adversarial robustness of unsupervised Graph Contrastive Learning (GCL) against structural attacks. First, we provide a comprehensive empirical and theoretical analysis of existing attacks, revealing how and why they downgrade the performance of GCL. Inspired by our analytic results, we present a robust GCL framework that integrates a homophily-driven sanitation view, which can be learned jointly with contrastive learning. A key challenge this poses, however, is the non-differentiable nature of the sanitation objective. To address this challenge, we propose a series of techniques to enable gradient-based end-to-end robust GCL. Moreover, we develop a fully unsupervised hyperparameter tuning method which, unlike prior approaches, does not require knowledge of node labels. We conduct extensive experiments to evaluate the performance of our proposed model, GCHS (Graph Contrastive Learning with Homophily-driven Sanitation View), against two state of the art structural attacks on GCL. Our results demonstrate that GCHS consistently outperforms all state of the art baselines in terms of the quality of generated node embeddings as well as performance on two important downstream tasks.
Authors: Xiao-Yin Liu, Xiao-Hu Zhou, Mei-Jiang Gui, Guo-Tao Li, Xiao-Liang Xie, Shi-Qi Liu, Shuang-Yi Wang, Qi-Chao Zhang, Biao Luo, Zeng-Guang Hou
Abstract: Model-based reinforcement learning (RL), which learns an environment model from the offline dataset and generates more out-of-distribution model data, has become an effective approach to the problem of distribution shift in offline RL. Due to the gap between the learned and actual environment, conservatism should be incorporated into the algorithm to balance accurate offline data and imprecise model data. The conservatism of current algorithms mostly relies on model uncertainty estimation. However, uncertainty estimation is unreliable and leads to poor performance in certain scenarios, and the previous methods ignore differences between the model data, which brings great conservatism. To address the above issues, this paper proposes a milDly cOnservative Model-bAsed offlINe RL algorithm (DOMAIN) without estimating model uncertainty, and designs the adaptive sampling distribution of model samples, which can adaptively adjust the model data penalty. In this paper, we theoretically demonstrate that the Q value learned by the DOMAIN outside the region is a lower bound of the true Q value, the DOMAIN is less conservative than previous model-based offline RL algorithms, and has the guarantee of safety policy improvement. The results of extensive experiments show that DOMAIN outperforms prior RL algorithms and the average performance has improved by 1.8% on the D4RL benchmark.
Authors: Tuan Truong, Hoang-Phi Nguyen, Haocheng Luo, Tung Pham, Mehrtash Harandi, Dinh Phung, Trung Le
Abstract: Recent studies highlight the effectiveness of flat minima in enhancing generalization, with sharpness-aware minimization (SAM) achieving state-of-the-art performance. Additionally, insights into the intrinsic geometry of the loss landscape have shown promise for improving model generalization. Building on these advancements, we introduce a novel sharpness-aware, geometry-aware teleportation mechanism to further enhance robustness and generalization. The core innovation of our approach is to decompose each iteration into a teleportation step within a local orbit and a sharpness-aware step that transitions between different orbits, leveraging the Riemannian quotient manifold. Our approach is grounded in a theoretical framework that analyzes the generalization gap between population loss and worst-case empirical loss within the context of Riemannian manifolds. To demonstrate the effectiveness of our method, we evaluate and compare our algorithm on diverse vision benchmarks with various datasets and Riemannian manifolds.
Authors: Ren-Jian Wang, Ke Xue, Yutong Wang, Peng Yang, Haobo Fu, Qiang Fu, Chao Qian
Abstract: Diversity plays a significant role in many problems, such as ensemble learning, reinforcement learning, and combinatorial optimization. How to define the diversity measure is a longstanding problem. Many methods rely on expert experience to define a proper behavior space and then obtain the diversity measure, which is, however, challenging in many scenarios. In this paper, we propose the problem of learning a behavior space from human feedback and present a general method called Diversity from Human Feedback (DivHF) to solve it. DivHF learns a behavior descriptor consistent with human preference by querying human feedback. The learned behavior descriptor can be combined with any distance measure to define a diversity measure. We demonstrate the effectiveness of DivHF by integrating it with the Quality-Diversity optimization algorithm MAP-Elites and conducting experiments on the QDax suite. The results show that the behavior learned by DivHF is much more consistent with human requirements than the one learned by direct data-driven approaches without human feedback, and makes the final solutions more diverse under human preference. Our contributions include formulating the problem, proposing the DivHF method, and demonstrating its effectiveness through experiments.
Authors: Jinglong Luo, Yehong Zhang, Zhuo Zhang, Jiaqi Zhang, Xin Mu, Hui Wang, Yue Yu, Zenglin Xu
Abstract: With the growing use of Transformer models hosted on cloud platforms to offer inference services, privacy concerns are escalating, especially concerning sensitive data like investment plans and bank account details. Secure Multi-Party Computing (SMPC) emerges as a promising solution to protect the privacy of inference data and model parameters. However, the application of SMPC in Privacy-Preserving Inference (PPI) for Transformer models often leads to considerable slowdowns or declines in performance. This is largely due to the multitude of nonlinear operations in the Transformer architecture, which are not well-suited to SMPC and difficult to circumvent or optimize effectively. To address this concern, we introduce a comprehensive PPI framework called SecFormer to achieve fast and accurate PPI for Transformer models. We successfully eliminate the high-cost exponential and maximum operations in PPI without sacrificing model performance and develop a suite of efficient SMPC protocols by employing suitable numerical computation methods to boost other complex nonlinear functions in PPI, including GeLU, LayerNorm, and a redesigned Softmax. Our extensive experiments reveal that SecFormer outperforms MPCFormer in performance, showing improvements of $3.4\%$ and $24.7\%$ for BERT$_{\text{BASE}}$ and BERT$_{\text{LARGE}}$, respectively. In terms of efficiency, SecFormer is 3.57 and 3.58 times faster than PUMA for BERT$_{\text{BASE}}$ and BERT$_{\text{LARGE}}$, demonstrating its effectiveness and speed.
Authors: Quang-Huy Nguyen, Jin Peng Zhou, Zhenzhen Liu, Khanh-Huyen Bui, Kilian Q. Weinberger, Wei-Lun Chao, Dung D. Le
Abstract: Recent object detectors have achieved impressive accuracy in identifying objects seen during training. However, real-world deployment often introduces novel and unexpected objects, referred to as out-of-distribution (OOD) objects, posing significant challenges to model trustworthiness. Modern object detectors are typically overconfident, making it unreliable to use their predictions alone for OOD detection. To address this, we propose leveraging an auxiliary model as a complementary solution. Specifically, we utilize an off-the-shelf text-to-image generative model, such as Stable Diffusion, which is trained with objective functions distinct from those of discriminative object detectors. We hypothesize that this fundamental difference enables the detection of OOD objects by measuring inconsistencies between the models. Concretely, for a given detected object bounding box and its predicted in-distribution class label, we perform class-conditioned inpainting on the image with the object removed. If the object is OOD, the inpainted image is likely to deviate significantly from the original, making the reconstruction error a robust indicator of OOD status. Extensive experiments demonstrate that our approach consistently surpasses existing zero-shot and non-zero-shot OOD detection methods, establishing a robust framework for enhancing object detection systems in dynamic environments.
Authors: Xingyue Huang, Miguel Romero Orth, Pablo Barcel\'o, Michael M. Bronstein, \.Ismail \.Ilkan Ceylan
Abstract: Link prediction with knowledge graphs has been thoroughly studied in graph machine learning, leading to a rich landscape of graph neural network architectures with successful applications. Nonetheless, it remains challenging to transfer the success of these architectures to relational hypergraphs, where the task of link prediction is over $k$-ary relations, which is substantially harder than link prediction with knowledge graphs. In this paper, we propose a framework for link prediction with relational hypergraphs, unlocking applications of graph neural networks to fully relational structures. Theoretically, we conduct a thorough analysis of the expressive power of the resulting model architectures via corresponding relational Weisfeiler-Leman algorithms and also via logical expressiveness. Empirically, we validate the power of the proposed model architectures on various relational hypergraph benchmarks. The resulting model architectures substantially outperform every baseline for inductive link prediction, and lead to state-of-the-art results for transductive link prediction.
Authors: Jianming Lv, Sijun Xia, Depin Liang, Wei Chen
Abstract: Traditional model-free feature selection methods treat each feature independently while disregarding the interrelationships among features, which leads to relatively poor performance compared with the model-aware methods. To address this challenge, we propose an efficient model-free feature selection framework via elastic expansion and compression of the features, namely EasyFS, to achieve better performance than state-of-the-art model-aware methods while sharing the characters of efficiency and flexibility with the existing model-free methods. In particular, EasyFS expands the feature space by using the random non-linear projection network to achieve the non-linear combinations of the original features, so as to model the interrelationships among the features and discover most correlated features. Meanwhile, a novel redundancy measurement based on the change of coding rate is proposed for efficient filtering of redundant features. Comprehensive experiments on 21 different datasets show that EasyFS outperforms state-of-the art methods up to 10.9\% in the regression tasks and 5.7\% in the classification tasks while saving more than 94\% of the time.
Authors: Nirjhar Das, Souradip Chakraborty, Aldo Pacchiano, Sayak Ray Chowdhury
Abstract: Large Language Models (LLMs) aligned using Reinforcement Learning from Human Feedback (RLHF) have shown remarkable generation abilities in numerous tasks. However, collecting high-quality human preferences creates costly bottlenecks in practical deployments, and hence, training data are often budgeted. In these scenarios, it is crucial to collect training data (e.g., contexts, a pair of generations for each context, and a preference indicating which generation is better) carefully, yet most of the existing methods sample contexts uniformly at random from a given collection. Given this, under the Bradley-Terry-Luce preference model and with a small budget of training data, we show that uniform sampling of contexts could lead to a policy (i.e., an aligned model) that suffers a constant sub-optimality gap from the optimal policy. This highlights the need for an adaptive context sampling strategy for effective alignment under a small sample budget. To address this, we reformulate RLHF within the contextual preference bandit framework, treating generations as actions, and give a nearly complete characterization of the sub-optimality gap in terms of both lower and upper bounds. First, when the action set is a $d$-dimensional hypercube and the number of samples is $T$, we show an $\Omega(d/\sqrt{T})$ lower bound. Next, we propose an algorithm, $\textit{Active Preference Optimization}$ ($\texttt{APO}$), that iteratively collects preferences for the most uncertain contexts. We show that the sub-optimality gap of the policy learned via $\texttt{APO}$ matches the lower bound up to a log factor and a non-linearity constant. Finally, we perform experiments on practical datasets to validate $\texttt{APO}$'s efficacy over existing methods, establishing it as a sample-efficient and cost-effective solution for LLM alignment.
Authors: David Buterez, Jon Paul Janet, Dino Oglic, Pietro Lio
Abstract: There has been a recent surge in transformer-based architectures for learning on graphs, mainly motivated by attention as an effective learning mechanism and the desire to supersede handcrafted operators characteristic of message passing schemes. However, concerns over their empirical effectiveness, scalability, and complexity of the pre-processing steps have been raised, especially in relation to much simpler graph neural networks that typically perform on par with them across a wide range of benchmarks. To tackle these shortcomings, we consider graphs as sets of edges and propose a purely attention-based approach consisting of an encoder and an attention pooling mechanism. The encoder vertically interleaves masked and vanilla self-attention modules to learn an effective representations of edges, while allowing for tackling possible misspecifications in input graphs. Despite its simplicity, the approach outperforms fine-tuned message passing baselines and recently proposed transformer-based methods on more than 70 node and graph-level tasks, including challenging long-range benchmarks. Moreover, we demonstrate state-of-the-art performance across different tasks, ranging from molecular to vision graphs, and heterophilous node classification. The approach also outperforms graph neural networks and transformers in transfer learning settings, and scales much better than alternatives with a similar performance level or expressive power.
Authors: Zhaolin Ren, Na Li
Abstract: This paper presents a new approach for batch Bayesian Optimization (BO) called Thompson Sampling-Regret to Sigma Ratio directed sampling (TS-RSR), where we sample a new batch of actions by minimizing a Thompson Sampling approximation of a regret to uncertainty ratio. Our sampling objective is able to coordinate the actions chosen in each batch in a way that minimizes redundancy between points whilst focusing on points with high predictive means or high uncertainty. Theoretically, we provide rigorous convergence guarantees on our algorithm's regret, and numerically, we demonstrate that our method attains state-of-the-art performance on a range of challenging synthetic and realistic test functions, where it outperforms several competitive benchmark batch BO algorithms.
Authors: Seungjae Jung, Gunsoo Han, Daniel Wontae Nam, Kyoung-Woon On
Abstract: In real-world services such as ChatGPT, aligning models based on user feedback is crucial for improving model performance. However, due to the simplicity and convenience of providing feedback, users typically offer only basic binary signals, such as 'thumbs-up' or 'thumbs-down'. Most existing alignment research, on the other hand, relies on preference-based approaches that require both positive and negative responses as a pair. We propose Binary Classifier Optimization (BCO), a technique that effectively aligns LLMs using only binary feedback. BCO trains a binary classifier, where the logit serves as an implicit reward, effectively minimizing the Direct Preference Optimization (DPO) loss. We demonstrate that the binary cross-entropy loss employed in classifier training acts as an upper bound for the DPO loss. Additionally, a novel reward shift technique further minimizes the gap between the losses. We validate our methodology in two settings: first, on a paired preference dataset, where our method performs on par with DPO; and second, on a Likert-5 scale annotation dataset which stems from real users' queries. Our model consistently demonstrates effective and robust alignment across four base LLMs and three different datasets, showcasing the strength of our approach to learning from binary signals.
Authors: Sho Sonoda, Yuka Hashimoto, Isao Ishikawa, Masahiro Ikeda
Abstract: We present a constructive universal approximation theorem for learning machines equipped with joint-group-equivariant feature maps, called the joint-equivariant machines, based on the group representation theory. ``Constructive'' here indicates that the distribution of parameters is given in a closed-form expression known as the ridgelet transform. Joint-group-equivariance encompasses a broad class of feature maps that generalize classical group-equivariance. Particularly, fully-connected networks are not group-equivariant but are joint-group-equivariant. Our main theorem also unifies the universal approximation theorems for both shallow and deep networks. Until this study, the universality of deep networks has been shown in a different manner from the universality of shallow networks, but our results discuss them on common ground. Now we can understand the approximation schemes of various learning machines in a unified manner. As applications, we show the constructive universal approximation properties of four examples: depth-$n$ joint-equivariant machine, depth-$n$ fully-connected network, depth-$n$ group-convolutional network, and a new depth-$2$ network with quadratic forms whose universality has not been known.
Authors: H\"useyin Tun\c{c}, Do\u{g}anay \"Ozese, \c{S}. \.Ilker Birbil, Donato Maragno, Marco Caserta, Mustafa Baydo\u{g}an
Abstract: Incorporating domain-specific constraints into machine learning models is essential for generating predictions that are both accurate and feasible in real-world applications. This paper introduces new methods for training Output-Constrained Regression Trees (OCRT), addressing the limitations of traditional decision trees in constrained multi-target regression tasks. We propose three approaches: M-OCRT, which uses split-based mixed integer programming to enforce constraints; E-OCRT, which employs an exhaustive search for optimal splits and solves constrained prediction problems at each decision node; and EP-OCRT, which applies post-hoc constrained optimization to tree predictions. To illustrate their potential uses in ensemble learning, we also introduce a random forest framework working under convex feasible sets. We validate the proposed methods through a computational study both on synthetic and industry-driven hierarchical time series datasets. Our results demonstrate that imposing constraints on decision tree training results in accurate and feasible predictions.
Authors: Najwa Laabid, Severi Rissanen, Markus Heinonen, Arno Solin, Vikas Garg
Abstract: Graph diffusion models, dominant in graph generative modeling, remain underexplored for graph-to-graph translation tasks like chemical reaction prediction. We demonstrate that standard permutation equivariant denoisers face fundamental limitations in these tasks due to their inability to break symmetries in noisy inputs. To address this, we propose aligning input and target graphs to break input symmetries while preserving permutation equivariance in non-matching graph portions. Using retrosynthesis (i.e., the task of predicting precursors for synthesis of a given target molecule) as our application domain, we show how alignment dramatically improves discrete diffusion model performance from 5% to a SOTA-matching 54.7% top-1 accuracy. Code is available at https://github.com/Aalto-QuML/DiffAlign.
Authors: Pengxiang Li, Lu Yin, Xiaowei Gao, Shiwei Liu
Abstract: The rapid advancements in Large Language Models (LLMs) have revolutionized various natural language processing tasks. However, the substantial size of LLMs presents significant challenges in training or fine-tuning. While parameter-efficient approaches such as low-rank adaptation (LoRA) have gained popularity, they often compromise performance compared to full-rank fine-tuning. In this paper, we propose Outlier-weighed Layerwise Sampling (OWS), a new memory-efficient fine-tuning approach, inspired by the layerwise outlier distribution of LLMs. Unlike LoRA, which adds extra adapters to all layers, OWS strategically assigns higher sampling probabilities to layers with more outliers, selectively sampling only a few layers and fine-tuning their pre-trained weights. To further increase the number of fine-tuned layers without a proportional rise in memory costs, we incorporate gradient low-rank projection, further boosting the approach's performance. Our extensive experiments across various architectures, including LLaMa2 and Mistral, demonstrate that OWS consistently outperforms baseline approaches, including full fine-tuning. Specifically, it achieves up to a 1.1% average accuracy gain on the Commonsense Reasoning benchmark, a 3.0% improvement on MMLU, and a notable 10% boost on MT-Bench, while being more memory efficient. OWS allows us to fine-tune 7B LLMs with only 21GB of memory. Our code is available at https://github.com/pixeli99/OWS.
Authors: Hsiao-Ying Lu, Yiran Li, Ujwal Pratap Krishna Kaluvakolanu Thyagarajan, Kwan-Liu Ma
Abstract: Graph Neural Networks (GNNs) achieve outstanding performance across graph-based tasks but remain difficult to interpret. In this paper, we revisit foundational assumptions underlying model-level explanation methods for GNNs, namely: (1) maximizing classification confidence yields representative explanations, (2) a single explanation suffices for an entire class of graphs, and (3) explanations are inherently trustworthy. We identify pitfalls resulting from these assumptions: methods that optimize for classification confidence may overlook partially learned patterns; topological diversity across graph subsets within the same class is often underrepresented; and explanations alone offer limited support for building user trust when applied to new datasets or models. This paper introduces GNNAnatomy, a distillation-based method designed to generate explanations while avoiding these pitfalls. GNNAnatomy first characterizes graph topology using graphlets, a set of fundamental substructures. We then train a transparent multilayer perceptron surrogate to directly approximate GNN predictions based on the graphlet representations. By analyzing the weights assigned to each graphlet, we identify the most discriminative topologies, which serve as GNN explanations. To account for structural diversity within a class, GNNAnatomy generates explanations at the required granularity through an interface that supports human-AI teaming. This interface helps users identify subsets of graphs where distinct critical substructures drive class differentiation, enabling multi-grained explanations. Additionally, by enabling exploration and linking explanations back to input graphs, the interface fosters greater transparency and trust. We evaluate GNNAnatomy on both synthetic and real-world datasets through quantitative metrics and qualitative comparisons with state-of-the-art model-level explainable GNN methods.
Authors: Jiachen T. Wang, Prateek Mittal, Dawn Song, Ruoxi Jia
Abstract: Data Shapley provides a principled framework for attributing data's contribution within machine learning contexts. However, existing approaches require re-training models on different data subsets, which is computationally intensive, foreclosing their application to large-scale models. Furthermore, they produce the same attribution score for any models produced by running the learning algorithm, meaning they cannot perform targeted attribution towards a specific model obtained from a single run of the algorithm. This paper introduces In-Run Data Shapley, which addresses these limitations by offering scalable data attribution for a target model of interest. In its most efficient implementation, our technique incurs negligible additional runtime compared to standard model training. This dramatic efficiency improvement makes it possible to perform data attribution for the foundation model pretraining stage for the first time. We present several case studies that offer fresh insights into pretraining data's contribution and discuss their implications for copyright in generative AI and pretraining data curation.
Authors: Maayan Ehrenberg, Roy Ganz, Nir Rosenfeld
Abstract: Adversarial training aims to defend against adversaries: malicious opponents whose sole aim is to harm predictive performance in any way possible. This presents a rather harsh perspective, which we assert results in unnecessarily conservative training. As an alternative, we propose to model opponents as simply pursuing their own goals--rather than working directly against the classifier. Employing tools from strategic modeling, our approach enables knowledge or beliefs regarding the opponent's possible incentives to be used as inductive bias for learning. Accordingly, our method of strategic training is designed to defend against all opponents within an 'incentive uncertainty set'. This resorts to adversarial learning when the set is maximal, but offers potential gains when the set can be appropriately reduced. We conduct a series of experiments that show how even mild knowledge regarding the opponent's incentives can be useful, and that the degree of potential gains depends on how these incentives relate to the structure of the learning task.
Authors: Pankayaraj Pathmanathan, Souradip Chakraborty, Xiangyu Liu, Yongyuan Liang, Furong Huang
Abstract: Recent advancements in Reinforcement Learning with Human Feedback (RLHF) have significantly impacted the alignment of Large Language Models (LLMs). The sensitivity of reinforcement learning algorithms such as Proximal Policy Optimization (PPO) has led to new line work on Direct Policy Optimization (DPO), which treats RLHF in a supervised learning framework. The increased practical use of these RLHF methods warrants an analysis of their vulnerabilities. In this work, we investigate the vulnerabilities of DPO to poisoning attacks under different scenarios and compare the effectiveness of preference poisoning, a first of its kind. We comprehensively analyze DPO's vulnerabilities under different types of attacks, i.e., backdoor and non-backdoor attacks, and different poisoning methods across a wide array of language models, i.e., LLama 7B, Mistral 7B, and Gemma 7B. We find that unlike PPO-based methods, which, when it comes to backdoor attacks, require at least 4\% of the data to be poisoned to elicit harmful behavior, we exploit the true vulnerabilities of DPO more simply so we can poison the model with only as much as 0.5\% of the data. We further investigate the potential reasons behind the vulnerability and how well this vulnerability translates into backdoor vs non-backdoor attacks.
Authors: Yuchang Zhu, Jintang Li, Yatao Bian, Zibin Zheng, Liang Chen
Abstract: Recent studies have highlighted fairness issues in Graph Neural Networks (GNNs), where they produce discriminatory predictions against specific protected groups categorized by sensitive attributes such as race and age. While various efforts to enhance GNN fairness have made significant progress, these approaches are often tailored to specific sensitive attributes. Consequently, they necessitate retraining the model from scratch to accommodate changes in the sensitive attribute requirement, resulting in high computational costs. To gain deeper insights into this issue, we approach the graph fairness problem from a causal modeling perspective, where we identify the confounding effect induced by the sensitive attribute as the underlying reason. Motivated by this observation, we formulate the fairness problem in graphs from an invariant learning perspective, which aims to learn invariant representations across environments. Accordingly, we propose a graph fairness framework based on invariant learning, namely FairINV, which enables the training of fair GNNs to accommodate various sensitive attributes within a single training session. Specifically, FairINV incorporates sensitive attribute partition and trains fair GNNs by eliminating spurious correlations between the label and various sensitive attributes. Experimental results on several real-world datasets demonstrate that FairINV significantly outperforms state-of-the-art fairness approaches, underscoring its effectiveness. Our code is available via: https://github.com/ZzoomD/FairINV/.
Authors: Viet-Hoang Tran, Trang Pham, Tho Tran, Minh Khoi Nguyen Nhat, Thanh Chu, Tam Le, Tan M. Nguyen
Abstract: Many variants of Optimal Transport (OT) have been developed to address its heavy computation. Among them, notably, Sliced Wasserstein (SW) is widely used for application domains by projecting the OT problem onto one-dimensional lines, and leveraging the closed-form expression of the univariate OT to reduce the computational burden. However, projecting measures onto low-dimensional spaces can lead to a loss of topological information. To mitigate this issue, in this work, we propose to replace one-dimensional lines with a more intricate structure, called tree systems. This structure is metrizable by a tree metric, which yields a closed-form expression for OT problems on tree systems. We provide an extensive theoretical analysis to formally define tree systems with their topological properties, introduce the concept of splitting maps, which operate as the projection mechanism onto these structures, then finally propose a novel variant of Radon transform for tree systems and verify its injectivity. This framework leads to an efficient metric between measures, termed Tree-Sliced Wasserstein distance on Systems of Lines (TSW-SL). By conducting a variety of experiments on gradient flows, image style transfer, and generative models, we illustrate that our proposed approach performs favorably compared to SW and its variants.
Authors: Jintao Yan, Tan Chen, Yuxuan Sun, Zhaojun Nan, Sheng Zhou, Zhisheng Niu
Abstract: Leveraging the computing and sensing capabilities of vehicles, vehicular federated learning (VFL) has been applied to edge training for connected vehicles. The dynamic and interconnected nature of vehicular networks presents unique opportunities to harness direct vehicle-to-vehicle (V2V) communications, enhancing VFL training efficiency. In this paper, we formulate a stochastic optimization problem to optimize the VFL training performance, considering the energy constraints and mobility of vehicles, and propose a V2V-enhanced dynamic scheduling (VEDS) algorithm to solve it. The model aggregation requirements of VFL and the limited transmission time due to mobility result in a stepwise objective function, which presents challenges in solving the problem. We thus propose a derivative-based drift-plus-penalty method to convert the long-term stochastic optimization problem to an online mixed integer nonlinear programming (MINLP) problem, and provide a theoretical analysis to bound the performance gap between the online solution and the offline optimal solution. Further analysis of the scheduling priority reduces the original problem into a set of convex optimization problems, which are efficiently solved using the interior-point method. Experimental results demonstrate that compared with the state-of-the-art benchmarks, the proposed algorithm enhances the image classification accuracy on the CIFAR-10 dataset by 4.20% and reduces the average displacement errors on the Argoverse trajectory prediction dataset by 9.82%.
Authors: Fanzeng Xia, Hao Liu, Yisong Yue, Tongxin Li
Abstract: In-Context Reinforcement Learning (ICRL) is a frontier paradigm to solve Reinforcement Learning (RL) problems in the foundation model era. While ICRL capabilities have been demonstrated in transformers through task-specific training, the potential of Large Language Models (LLMs) out-of-the-box remains largely unexplored. This paper investigates whether LLMs can generalize cross-domain to perform ICRL under the problem of Dueling Bandits (DB), a stateless preference-based RL setting. We find that the top-performing LLMs exhibit a notable zero-shot capacity for relative decision-making, which translates to low short-term weak regret across all DB environment instances by quickly including the best arm in duels. However, an optimality gap still exists between LLMs and classic DB algorithms in terms of strong regret. LLMs struggle to converge and consistently exploit even when explicitly prompted to do so, and are sensitive to prompt variations. To bridge this gap, we propose an agentic flow framework: LLM with Enhanced Algorithmic Dueling (LEAD), which integrates off-the-shelf DB algorithm support with LLM agents through fine-grained adaptive interplay. We show that LEAD has theoretical guarantees inherited from classic DB algorithms on both weak and strong regret. We validate its efficacy and robustness even with noisy and adversarial prompts. The design of such an agentic framework sheds light on how to enhance the trustworthiness of general-purpose LLMs generalized to in-context decision-making tasks.
Authors: Ajay Jaiswal, Yifan Wang, Lu Yin, Shiwei Liu, Runjin Chen, Jiawei Zhao, Ananth Grama, Yuandong Tian, Zhangyang Wang
Abstract: Large Language Models' (LLMs) weight matrices can often be expressed in low-rank form with potential to relax memory and compute resource requirements. Unlike prior efforts that focus on developing novel matrix decompositions, in this work we study the non-uniform low-rank properties of weight matrices in LLMs through the lens of stabilizing gradient subspace. First, we provide a theoretical framework to understand the stabilization of gradient subspaces through Hessian analysis. Second, we empirically establish an important relationship between gradient dynamics and low-rank expressiveness of weight matrices. Our findings reveal that different LLM components exhibit varying levels of converged low-rank structures, necessitating variable rank reduction across them to minimize drop in performance due to compression. Drawing on this result, we present Weight Low-Rank Projection(WeLore) that unifies weight compression and memory-efficient fine-tuning into one, in a data-agnostic and one-shot manner. When used as a compression technique, WeLore categorizes weight matrices into Low-rank Components (LRCs) and Non-Low-rank Components (N-LRCs) and suitably encodes them for minimum performance loss. Our gradient dynamics perspective illustrates that LRCs tend to have better fine-tuning capabilities and their standalone fine-tuning can closely mimic and sometimes outperform the training loss trajectory and performance of full fine-tuning with notable memory and compute footprint reduction. Codes are available at https://github.com/VITA-Group/WeLore.
Authors: Birte Boes, Jaan-Willem Simon, Hagen Holthusen
Abstract: In this work, we extend the existing framework of inelastic constitutive artificial neural networks (iCANNs) by incorporating plasticity to increase their applicability to model more complex material behavior. The proposed approach ensures objectivity, material symmetry, and thermodynamic consistency, providing a robust basis for automatic model discovery of constitutive equations at finite strains. These are predicted by discovering formulations for the Helmholtz free energy and plastic potentials for the yield function and evolution equations in terms of feed-forward networks. Our framework captures both linear and nonlinear kinematic hardening behavior. Investigation of our model's prediction showed that the extended iCANNs successfully predict both linear and nonlinear kinematic hardening behavior based on experimental and artificially generated datasets, showcasing promising capabilities of this framework. Nonetheless, challenges remain in discovering more complex yield criteria with tension-compression asymmetry and addressing deviations in experimental data at larger strains. Despite these limitations, the proposed framework provides a promising basis for incorporating plasticity into iCANNs, offering a platform for advancing in the field of automated model discovery.
Authors: Zeyu Zhang, Haiying Shen
Abstract: In large-language models, memory constraints in the Key-Value Cache (KVC) pose a challenge during inference. In this work, we propose FDC, a fast KV dimensionality compression system that eliminates the decompression overhead incurred in the existing KV dimensionality compression system, Palu, and reduces attention time. Moreover, FDC employs adaptive compression, tailoring KV compression rates across heads and layers based on their contributions to inference to maximize overall compression while maintaining an accuracy loss constraint. Additionally, FDC enhances the attention kernel to balance the uneven workloads caused by the adaptive compression approach to further reduce attention computation latency. Comprehensive experiments demonstrate that compared to Palu, FDC can reduce Job Completion Time (JCT) by up to 64%, and delivers up to 1.97X throughput under the same latency, while maintaining 99% of the accuracy without compression. When state-of-the-art eviction and quantization methods are combined with FDC, they exhibit similar improvements compared to those combined with Palu. We open-sourced the code.
Authors: Chuwei Wang, Julius Berner, Zongyi Li, Di Zhou, Jiayun Wang, Jane Bae, Anima Anandkumar
Abstract: Accurately predicting the long-term behavior of chaotic systems is crucial for various applications such as climate modeling. However, achieving such predictions typically requires iterative computations over a dense spatiotemporal grid to account for the unstable nature of chaotic systems, which is expensive and impractical in many real-world situations. An alternative approach to such a full-resolved simulation is using a coarse grid and then correcting its errors through a \textit{closure model}, which approximates the overall information from fine scales not captured in the coarse-grid simulation. Recently, ML approaches have been used for closure modeling, but they typically require a large number of training samples from expensive fully-resolved simulations (FRS). In this work, we prove an even more fundamental limitation, i.e., the standard approach to learning closure models suffers from a large approximation error for generic problems, no matter how large the model is, and it stems from the non-uniqueness of the mapping. We propose an alternative end-to-end learning approach using a physics-informed neural operator (PINO) that overcomes this limitation by not using a closure model or a coarse-grid solver. We first train the PINO model on data from a coarse-grid solver and then fine-tune it with (a small amount of) FRS and physics-based losses on a fine grid. The discretization-free nature of neural operators means that they do not suffer from the restriction of a coarse grid that closure models face, and they can provably approximate the long-term statistics of chaotic systems. In our experiments, our PINO model achieves a 330x speedup compared to FRS with a relative error $\sim 10\%$. In contrast, the closure model coupled with a coarse-grid solver is $60$x slower than PINO while having a much higher error $\sim186\%$ when the closure model is trained on the same FRS dataset.
Authors: Yuan Tian, Tianyi Zhang
Abstract: Recent advances in large language models (LLMs) have transformed software development by automatically generating code from natural language. Yet challenges remain in generating fully correct code that aligns with user intent. Our study reveals that LLMs tend to pay less attention to user prompts as more code tokens are generated. We hypothesize that this attention dilution issue is an important reason for code generation errors. To mitigate this issue, we propose Selective Prompt Anchoring (SPA) to guide code LLMs to pay more attention to user intent when generating code. We evaluate SPA using six base LLMs across six benchmarks. Our results demonstrate that SPA enhances Pass@1 by up to 12.9%, consistently outperforming SOTA code generation methods in all settings. Our code is available at https://github.com/magic-YuanTian/Selective-Prompt-Anchoring.
URLs: https://github.com/magic-YuanTian/Selective-Prompt-Anchoring.
Authors: Yao Shu, Wenyang Hu, See-Kiong Ng, Bryan Kian Hsiang Low, Fei Richard Yu
Abstract: Large Language Models (LLMs) have become indispensable in numerous real-world applications. However, fine-tuning these models at scale, especially in federated settings where data privacy and communication efficiency are critical, presents significant challenges. Existing approaches often resort to parameter-efficient fine-tuning (PEFT) to mitigate communication overhead, but this typically comes at the cost of model accuracy. To this end, we propose federated full-parameter tuning at scale for LLMs (Ferret), the first first-order method with shared randomness to enable scalable full-parameter tuning of LLMs across decentralized data sources while maintaining competitive model accuracy. Ferret accomplishes this through three aspects: (i) it employs widely used first-order methods for efficient local updates; (ii) it projects these updates into a low-dimensional space to considerably reduce communication overhead; and (iii) it reconstructs local updates from this low-dimensional space with shared randomness to facilitate effective full-parameter global aggregation, ensuring fast convergence and competitive final performance. Our rigorous theoretical analyses and insights along with extensive experiments, show that Ferret significantly enhances the scalability of existing federated full-parameter tuning approaches by achieving high computational efficiency, reduced communication overhead, and fast convergence, all while maintaining competitive model accuracy. Our implementation is available at https://github.com/allen4747/Ferret.
Authors: Chuheng Zhang, Wei Shen, Li Zhao, Xuyun Zhang, Xiaolong Xu, Wanchun Dou, Jiang Bian
Abstract: While direct policy optimization methods exist, pioneering LLMs are fine-tuned with reinforcement learning from human feedback (RLHF) to generate better responses under the supervision of a reward model learned from preference data. One major challenge of RLHF is the inaccuracy of the intermediate reward model, especially in the tasks that requires complex reasoning for the reward model to score a response. We find that the reliability of the reward model varies across responses assigned with different rewards. This motivates us to filter the samples whose rewards may be unreliable to improve the signal-to-noise ratio during policy learning, resulting in Policy Filtration for Proximal Policy Optimization (PF-PPO). To choose a proper policy filtering strategy, we use the coefficient of determination (R2) between the rewards and actual scores on filtered samples as the metrics to help us find promising strategies since it measures how well the rewards filtered by PF-PPO indicate real performance. We provide extensive experiments to validate the effectiveness of PF-PPO in code generation and math reasoning tasks. In code generation, PF-PPO achieves the state-of-the-art performance of 7-billion-parameter models on HumanEval (+7.9%), MBPP (+0.7%), and LeetCode Contest (+10.0%) which is a more challenging benchmark created by us. In math reasoning, PF-PPO yields performance increase using different reward models and benchmarks (Ape210K and CMATH). Code is available on https://github.com/DtYXs/verl/tree/pf-ppo.
Authors: Wenhao Zhao, Qiushui Xu, Linjie Xu, Lei Song, Jinyu Wang, Chunlai Zhou, Jiang Bian
Abstract: Recently, incorporating knowledge from pretrained language models (PLMs) into decision transformers (DTs) has generated significant attention in offline reinforcement learning (RL). These PLMs perform well in RL tasks, raising an intriguing question: what kind of knowledge from PLMs has been transferred to RL to achieve such good results? This work first dives into this problem by analyzing each head quantitatively and points out Markov head, a crucial component that exists in the attention heads of PLMs. It leads to extreme attention on the last-input token and performs well only in short-term environments. Furthermore, we prove that this extreme attention cannot be changed by re-training embedding layer or fine-tuning. Inspired by our analysis, we propose a general method GPT2-DTMA, which equips a pretrained DT with Mixture of Attention (MoA), to accommodate diverse attention requirements during fine-tuning. Extensive experiments corroborate our theorems and demonstrate the effectiveness of GPT2-DTMA: it achieves comparable performance in short-term environments while significantly narrowing the performance gap in long-term environments.
Authors: Mingyue Cheng, Jintao Zhang, Zhiding Liu, Chunli Liu
Abstract: Intraoperative hypotension (IOH) prediction using past physiological signals is crucial, as IOH may lead to inadequate organ perfusion and significantly elevate the risk of severe complications and mortality. However, current methods often rely on static modeling, overlooking the complex temporal dependencies and the inherently non-stationary nature of physiological signals. We propose a Hybrid Multi-Factor (HMF) network that formulates IOH prediction as a dynamic sequence forecasting task, explicitly capturing both temporal dependencies and physiological non-stationarity. We represent signal dynamics as multivariate time series and decompose them into trend and seasonal components, enabling separate modeling of long-term and periodic variations. Each component is encoded with a patch-based Transformer to balance computational efficiency and feature representation. To address distributional drift from evolving signals, we introduce a symmetric normalization mechanism. Experiments on both public and real-world clinical datasets show that HMF significantly outperforms competitive baselines. We hope HMF offers new insights into IOH prediction and ultimately promotes safer surgical care. Our code is available at https://github.com/Mingyue-Cheng/HMF.
Authors: Anwar Said, Waseem Abbas, Xenofon Koutsoukos
Abstract: Graph machine learning, particularly using graph neural networks, heavily relies on node features. However, many real-world systems, such as social and biological networks, lack node features due to privacy concerns, incomplete data, or collection limitations. Structural and positional encoding are commonly used to address this but are constrained by the maximum values of the encoded properties, such as the highest node degree. This limitation makes them impractical for scale-free networks and applications involving large or non-categorical properties. This paper introduces PropEnc, a novel and versatile encoder to generate expressive node embedding from any graph metric. By combining histogram construction with reversed index encoding, PropEnc offers a flexible solution that supports low-dimensional representations and diverse input types, effectively mitigating sparsity issues while improving computational efficiency. Additionally, it replicates one-hot encoding or approximates indices with high accuracy, making it adaptable to a wide range of graph applications. We validate PropEnc through extensive experiments on graph classification task across several social networks lacking node features. The empirical results demonstrate that PropEnc offers an efficient mechanism for constructing node features from various graph metrics.
Authors: Parham Oveissi, Turibius Rozario, Ankit Goel
Abstract: The application of neural networks in modeling dynamic systems has become prominent due to their ability to estimate complex nonlinear functions. Despite their effectiveness, neural networks face challenges in long-term predictions, where the prediction error diverges over time, thus degrading their accuracy. This paper presents a neural filter to enhance the accuracy of long-term state predictions of neural network-based models of dynamic systems. Motivated by the extended Kalman filter, the neural filter combines the neural network state predictions with the measurements from the physical system to improve the estimated state's accuracy. The neural filter's improvements in prediction accuracy are demonstrated through applications to four nonlinear dynamical systems. Numerical experiments show that the neural filter significantly improves prediction accuracy and bounds the state estimate covariance, outperforming the neural network predictions. Furthermore, it is also shown that the accuracy of a poorly trained neural network model can be improved to the same level as that of an adequately trained neural network model, potentially decreasing the training cost and required data to train a neural network.
Authors: Mucong Ding, Chenghao Deng, Jocelyn Choo, Zichu Wu, Aakriti Agrawal, Avi Schwarzschild, Tianyi Zhou, Tom Goldstein, John Langford, Anima Anandkumar, Furong Huang
Abstract: While generalization over tasks from easy to hard is crucial to profile language models (LLMs), the datasets with fine-grained difficulty annotations for each problem across a broad range of complexity are still blank. Aiming to address this limitation, we present Easy2Hard-Bench, a consistently formatted collection of 6 benchmark datasets spanning various domains, such as mathematics and programming problems, chess puzzles, and reasoning questions. Each problem within these datasets is annotated with numerical difficulty scores. To systematically estimate problem difficulties, we collect abundant performance data on attempts to each problem by humans in the real world or LLMs on the prominent leaderboard. Leveraging the rich performance data, we apply well-established difficulty ranking systems, such as Item Response Theory (IRT) and Glicko-2 models, to uniformly assign numerical difficulty scores to problems. Moreover, datasets in Easy2Hard-Bench distinguish themselves from previous collections by a higher proportion of challenging problems. Through extensive experiments with six state-of-the-art LLMs, we provide a comprehensive analysis of their performance and generalization capabilities across varying levels of difficulty, with the aim of inspiring future research in LLM generalization. The datasets are available at https://huggingface.co/datasets/furonghuang-lab/Easy2Hard-Bench.
URLs: https://huggingface.co/datasets/furonghuang-lab/Easy2Hard-Bench.
Authors: Tuan Truong, Quyen Tran, Quan Pham-Ngoc, Nhat Ho, Dinh Phung, Trung Le
Abstract: We introduce Flat Hilbert Bayesian Inference (FHBI), an algorithm designed to enhance generalization in Bayesian inference. Our approach involves an iterative two-step procedure with an adversarial functional perturbation step and a functional descent step within a reproducing kernel Hilbert space. This methodology is supported by a theoretical analysis that extends previous findings on generalization ability from finite-dimensional Euclidean spaces to infinite-dimensional functional spaces. To evaluate the effectiveness of FHBI, we conduct comprehensive comparisons against nine baseline methods on the \texttt{VTAB-1K} benchmark, which encompasses 19 diverse datasets across various domains with diverse semantics. Empirical results demonstrate that FHBI consistently outperforms the baselines by notable margins, highlighting its practical efficacy.
Authors: Aakriti Agrawal, Mucong Ding, Zora Che, Chenghao Deng, Anirudh Satheesh, John Langford, Furong Huang
Abstract: How can we harness the collective capabilities of multiple Large Language Models (LLMs) to create an even more powerful model? This question forms the foundation of our research, where we propose an innovative approach to weak-to-strong (w2s) generalization-a critical problem in AI alignment. Our work introduces an easy-to-hard (e2h) framework for studying the feasibility of w2s generalization, where weak models trained on simpler tasks collaboratively supervise stronger models on more complex tasks. This setup mirrors real-world challenges, where direct human supervision is limited. To achieve this, we develop a novel AdaBoost-inspired ensemble method, demonstrating that an ensemble of weak supervisors can enhance the performance of stronger LLMs across classification and generative tasks on difficult QA datasets. In several cases, our ensemble approach matches the performance of models trained on ground-truth data, establishing a new benchmark for w2s generalization. We observe an improvement of up to 14% over existing baselines and average improvements of 5% and 4% for binary classification and generative tasks, respectively. This research points to a promising direction for enhancing AI through collective supervision, especially in scenarios where labeled data is sparse or insufficient.
Authors: Emanuele Sansone, Tim Lebailly, Tinne Tuytelaars
Abstract: We present a principled and simplified design of the projector and loss function for non-contrastive self-supervised learning based on hyperdimensional computing. We theoretically demonstrate that this design introduces an inductive bias that encourages representations to be simultaneously decorrelated and clustered, without explicitly enforcing these properties. This bias provably enhances generalization and suffices to avoid known training failure modes, such as representation, dimensional, cluster, and intracluster collapses. We validate our theoretical findings on image datasets, including SVHN, CIFAR-10, CIFAR-100, and ImageNet-100. Our approach effectively combines the strengths of feature decorrelation and cluster-based self-supervised learning methods, overcoming training failure modes while achieving strong generalization in clustering and linear classification tasks.
Authors: Guy Kornowski, Daogao Liu, Kunal Talwar
Abstract: We study differentially private (DP) optimization algorithms for stochastic and empirical objectives which are neither smooth nor convex, and propose methods that return a Goldstein-stationary point with sample complexity bounds that improve on existing works. We start by providing a single-pass $(\epsilon,\delta)$-DP algorithm that returns an $(\alpha,\beta)$-stationary point as long as the dataset is of size $\widetilde{\Omega}(\sqrt{d}/\alpha\beta^{3}+d/\epsilon\alpha\beta^{2})$, which is $\Omega(\sqrt{d})$ times smaller than the algorithm of Zhang et al. [2024] for this task, where $d$ is the dimension. We then provide a multi-pass polynomial time algorithm which further improves the sample complexity to $\widetilde{\Omega}\left(d/\beta^2+d^{3/4}/\epsilon\alpha^{1/2}\beta^{3/2}\right)$, by designing a sample efficient ERM algorithm, and proving that Goldstein-stationary points generalize from the empirical loss to the population loss.
Authors: Xing Lei, Xuetao Zhang, Zifeng Zhuang, Donglin Wang
Abstract: A novel class of advanced algorithms, termed Goal-Conditioned Weighted Supervised Learning (GCWSL), has recently emerged to tackle the challenges posed by sparse rewards in goal-conditioned reinforcement learning (RL). GCWSL consistently delivers strong performance across a diverse set of goal-reaching tasks due to its simplicity, effectiveness, and stability. However, GCWSL methods lack a crucial capability known as trajectory stitching, which is essential for learning optimal policies when faced with unseen skills during testing. This limitation becomes particularly pronounced when the replay buffer is predominantly filled with sub-optimal trajectories. In contrast, traditional TD-based RL methods, such as Q-learning, which utilize Dynamic Programming, do not face this issue but often experience instability due to the inherent difficulties in value function approximation. In this paper, we propose Q-learning Weighted Supervised Learning (Q-WSL), a novel framework designed to overcome the limitations of GCWSL by incorporating the strengths of Dynamic Programming found in Q-learning. Q-WSL leverages Dynamic Programming results to output the optimal action of (state, goal) pairs across different trajectories within the replay buffer. This approach synergizes the strengths of both Q-learning and GCWSL, effectively mitigating their respective weaknesses and enhancing overall performance. Empirical evaluations on challenging goal-reaching tasks demonstrate that Q-WSL surpasses other goal-conditioned approaches in terms of both performance and sample efficiency. Additionally, Q-WSL exhibits notable robustness in environments characterized by binary reward structures and environmental stochasticity.
Authors: Yang Chen, Long Yang, Yitao Liang, Zhouchen Lin
Abstract: Low-Dimension-to-High-Dimension (LDHD) generalization is a special case of Out-of-Distribution (OOD) generalization, where the training data are restricted to a low-dimensional subspace of the high-dimensional testing space. Assuming that each instance is generated from a latent variable and the dimension of the latent variable reflects the problem scale, the inherent scaling challenge in length generalization can be captured by the LDHD generalization in the latent space. We theoretically demonstrate that LDHD generalization is generally unattainable without exploiting prior knowledge to provide appropriate inductive bias. Specifically, we explore LDHD generalization in Boolean functions. We verify that different architectures trained with (S)GD converge to \emph{min-degree interpolators w.r.t. different independent sets}. LDHD generalization is achievable if and only if the target function coincides with this inductive bias. Applying the insights from LDHD generalization to length generalization, we explain the effectiveness of CoT as changing the structure latent space to enable better LDHD generalization. We also propose a principle for position embedding design to handle both the inherent LDHD generalization and the nuisances such as the data format. Following the principle, we propose a novel position embedding called RPE-Square that remedies the RPE for dealing with the data format nuisance.
Authors: Kevin Galim, Wonjun Kang, Yuchen Zeng, Hyung Il Koo, Kangwook Lee
Abstract: Deep State Space Models (SSMs), such as Mamba (Gu & Dao, 2024), have become powerful tools for language modeling, offering high performance and linear scalability with sequence length. However, the application of parameter-efficient fine-tuning (PEFT) methods to SSM-based models remains largely underexplored. We start by investigating two fundamental questions on existing PEFT methods: (i) How do they perform on SSM-based models? (ii) Which parameters should they target for optimal results? Our analysis shows that LoRA and its variants consistently outperform all other PEFT methods. While LoRA is effective for linear projection matrices, it fails on SSM modules-yet still outperforms other methods applicable to SSMs, indicating their limitations. This underscores the need for a specialized SSM tuning approach. To address this, we propose Sparse Dimension Tuning (SDT), a PEFT method tailored for SSM modules. Combining SDT for SSMs with LoRA for linear projection matrices, we achieve state-of-the-art performance across extensive experiments.
Authors: Xiaoyan Hu, Ho-fung Leung, Farzan Farnia
Abstract: Selecting a sample generation scheme from multiple prompt-based generative models, including large language models (LLMs) and prompt-guided image and video generation models, is typically addressed by choosing the model that maximizes an averaged evaluation score. However, this score-based selection overlooks the possibility that different models achieve the best generation performance for different types of text prompts. An online identification of the best generation model for various input prompts can reduce the costs associated with querying sub-optimal models. In this work, we explore the possibility of varying rankings of text-based generative models for different text prompts and propose an online learning framework to predict the best data generation model for a given input prompt. The proposed PAK-UCB algorithm addresses a contextual bandit (CB) setting with shared context variables across the arms, utilizing the generated data to update kernel-based functions that predict the score of each model available for unseen text prompts. Additionally, we leverage random Fourier features (RFF) to accelerate the online learning process of PAK-UCB. Our numerical experiments on real and simulated text-to-image and image-to-text generative models show that RFF-UCB performs successfully in identifying the best generation model across different sample types. The code is available at: github.com/yannxiaoyanhu/dgm-online-select.
Authors: Da Long, Zhitong Xu, Guang Yang, Akil Narayan, Shandian Zhe
Abstract: Modern physics simulation often involves multiple functions of interests, and traditional numerical approaches are known to be complex and computationally costly. While machine learning-based surrogate models can offer significant cost reductions, most focus on a single task, such as forward prediction, and typically lack uncertainty quantification -- an essential component in many applications. To overcome these limitations, we propose Arbitrarily-Conditioned Multi-Functional Diffusion (ACM-FD), a versatile probabilistic surrogate model for multi-physics emulation. ACM-FD can perform a wide range of tasks within a single framework, including forward prediction, various inverse problems, and simulating data for entire systems or subsets of quantities conditioned on others. Specifically, we extend the standard Denoising Diffusion Probabilistic Model (DDPM) for multi-functional generation by modeling noise as Gaussian processes (GP). We propose a random-mask based, zero-regularized denoising loss to achieve flexible and robust conditional generation. We induce a Kronecker product structure in the GP covariance matrix, substantially reducing the computational cost and enabling efficient training and sampling. We demonstrate the effectiveness of ACM-FD across several fundamental multi-physics systems.
Authors: Philippe Chlenski, Quentin Chu, Raiyan R. Khan, Kaizhu Du, Antonio Khalil Moretti, Itsik Pe'er
Abstract: Decision trees (DTs) and their random forest (RF) extensions are workhorses of classification and regression in Euclidean spaces. However, algorithms for learning in non-Euclidean spaces are still limited. We extend DT and RF algorithms to product manifolds: Cartesian products of several hyperbolic, hyperspherical, or Euclidean components. Such manifolds handle heterogeneous curvature while still factorizing neatly into simpler components, making them compelling embedding spaces for complex datasets. Our novel angular reformulation respects manifold geometry while preserving the algorithmic properties that make decision trees effective. In the special cases of single-component manifolds, our method simplifies to its Euclidean or hyperbolic counterparts, or introduces hyperspherical DT algorithms, depending on the curvature. In benchmarks on a diverse suite of 57 classification, regression, and link prediction tasks, our product RFs ranked first on 29 tasks and came in the top 2 for 41. This highlights the value of product RFs as straightforward yet powerful new tools for data analysis in product manifolds. Code for our method is available at https://github.com/pchlenski/manify.
Authors: Chaoxi Niu, Hezhe Qiao, Changlu Chen, Ling Chen, Guansong Pang
Abstract: Graph anomaly detection (GAD), which aims to identify nodes in a graph that significantly deviate from normal patterns, plays a crucial role in broad application domains. However, existing GAD methods are one-model-for-one-dataset approaches, i.e., training a separate model for each graph dataset. This largely limits their applicability in real-world scenarios. To overcome this limitation, we propose a novel zero-shot generalist GAD approach UNPrompt that trains a one-for-all detection model, requiring the training of one GAD model on a single graph dataset and then effectively generalizing to detect anomalies in other graph datasets without any retraining or fine-tuning. The key insight in UNPrompt is that i) the predictability of latent node attributes can serve as a generalized anomaly measure and ii) generalized normal and abnormal graph patterns can be learned via latent node attribute prediction in a properly normalized node attribute space. UNPrompt achieves a generalist mode for GAD through two main modules: one module aligns the dimensionality and semantics of node attributes across different graphs via coordinate-wise normalization, while another module learns generalized neighborhood prompts that support the use of latent node attribute predictability as an anomaly score across different datasets. Extensive experiments on real-world GAD datasets show that UNPrompt significantly outperforms diverse competing methods under the generalist GAD setting, and it also has strong superiority under the one-model-for-one-dataset setting. Code is available at https://github.com/mala-lab/UNPrompt.
Authors: Tian Li, Tianyi Zhou, Jeffrey A. Bilmes
Abstract: Sharpness-Aware Minimization (SAM) has been demonstrated to improve the generalization performance of overparameterized models by seeking flat minima on the loss landscape through optimizing model parameters that incur the largest loss within a neighborhood. Nevertheless, such min-max formulations are computationally challenging especially when the problem is highly non-convex. Additionally, focusing only on the worst-case local solution while ignoring potentially many other local solutions may be suboptimal when searching for flat minima. In this work, we propose Tilted SAM (TSAM), a smoothed generalization of SAM inspired by exponential tilting that effectively assigns higher priority to local solutions that incur larger losses. TSAM is parameterized by a tilt hyperparameter $t$ and reduces to SAM as $t$ approaches infinity. We show that TSAM is smoother than SAM and thus easier to optimize, and it explicitly favors flatter minima. We develop algorithms motivated by the discretization of Hamiltonian dynamics to solve TSAM. Empirically, TSAM arrives at flatter local minima and results in superior test performance than the baselines of SAM and ERM across a range of image and text tasks.
Authors: Shlomo Libo Feigin, Maximilian Fleissner, Debarghya Ghoshdastidar
Abstract: Data augmentations play an important role in the recent success of self-supervised learning (SSL). While augmentations are commonly understood to encode invariances between different views into the learned representations, this interpretation overlooks the impact of the pretraining architecture and suggests that SSL would require diverse augmentations which resemble the data to work well. However, these assumptions do not align with empirical evidence, encouraging further theoretical understanding to guide the principled design of augmentations in new domains. To this end, we use kernel theory to derive analytical expressions for data augmentations that achieve desired target representations after pretraining. We consider non-contrastive and contrastive losses, namely VICReg, Barlow Twins and the Spectral Contrastive Loss, and provide an algorithm to construct such augmentations. Our analysis shows that augmentations need not be similar to the data to learn useful representations, nor be diverse, and that the architecture has a significant impact on the optimal augmentations.
Authors: Yushi Yang, Filip Sondej, Harry Mayne, Andrew Lee, Adam Mahdi
Abstract: Safety fine-tuning algorithms reduce harmful outputs in language models, yet their mechanisms remain under-explored. Direct Preference Optimization (DPO) is a popular choice of algorithm, but prior explanations, attributing its effects solely to dampened toxic neurons in the MLP layers, are incomplete. In this study, we analyse four language models (Llama-3.1-8B, Gemma-2-2B, Mistral-7B, GPT-2-Medium) and show that toxic neurons only account for 2.5% to 24% of DPO's effects across models. Instead, DPO balances distributed activation shifts across all MLP neurons to create a net toxicity reduction. We attribute this reduction to four neuron groups, two aligned with reducing toxicity and two promoting anti-toxicity, whose combined effects replicate DPO across models. To further validate this understanding, we develop an activation editing method mimicking DPO through distributed shifts along a toxicity representation. This method outperforms DPO in reducing toxicity while preserving perplexity, without requiring any weight updates. Our work provides a mechanistic understanding of DPO and introduces an efficient, tuning-free alternative for safety fine-tuning.
Authors: Huashan Sun, Yizhe Yang, Yinghao Li, Jiawei Li, Yang Gao
Abstract: Although substantial efforts have been made to mitigate catastrophic forgetting in continual learning, the intrinsic mechanisms are not well understood. In this work, we demonstrate the existence of "pseudo forgetting": the performance degradation on previous tasks is not attributed to a loss of capabilities, but rather to the failure of the instructions to activate the appropriate model abilities. We show that the model's performance on previous tasks can be restored through two simple interventions: (1) providing partial external correct rationale, and (2) appending semantically meaningless suffixes to the original instructions, to guide the generation of correct rationales. Through empirical analysis of the internal mechanisms governing rationale generation, we reveal that models exhibiting pseudo forgetting show reduced instruction dependence during rationale generation, leading to suboptimal activation of their inherent capabilities. Based on this insight, we propose Rationale-Guidance Difficulty based Replay (RGD-R) framework that dynamically allocates replay data based on the model's ability to correctly leverage the intrinsic capabilities. Experimental results demonstrate that RGD-R effectively mitigates pseudo forgetting while maintaining model plasticity.
Authors: Melda Yeghaian, Zuhir Bodalal, Daan van den Broek, John B A G Haanen, Regina G H Beets-Tan, Stefano Trebeschi, Marcel A J van Gerven
Abstract: Purpose: Immunotherapies have revolutionized the landscape of cancer treatments. However, our understanding of response patterns in advanced cancers treated with immunotherapy remains limited. By leveraging routinely collected noninvasive longitudinal and multimodal data with artificial intelligence, we could unlock the potential to transform immunotherapy for cancer patients, paving the way for personalized treatment approaches. Methods: In this study, we developed a novel artificial neural network architecture, multimodal transformer-based simple temporal attention (MMTSimTA) network, building upon a combination of recent successful developments. We integrated pre- and on-treatment blood measurements, prescribed medications and CT-based volumes of organs from a large pan-cancer cohort of 694 patients treated with immunotherapy to predict mortality at three, six, nine and twelve months. Different variants of our extended MMTSimTA network were implemented and compared to baseline methods incorporating intermediate and late fusion based integration methods. Results: The strongest prognostic performance was demonstrated using a variant of the MMTSimTA model with area under the curves (AUCs) of $0.84 \pm $0.04, $0.83 \pm $0.02, $0.82 \pm $0.02, $0.81 \pm $0.03 for 3-, 6-, 9-, and 12-month survival prediction, respectively. Discussion: Our findings show that integrating noninvasive longitudinal data using our novel architecture yields an improved multimodal prognostic performance, especially in short-term survival prediction. Conclusion: Our study demonstrates that multimodal longitudinal integration of noninvasive data using deep learning may offer a promising approach for personalized prognostication in immunotherapy-treated cancer patients.
Authors: Xinyu Su, Feng Liu, Yanchuan Chang, Egemen Tanin, Majid Sarvi, Jianzhong Qi
Abstract: Traffic forecasting is crucial for transportation systems optimisation. Current models minimise the mean forecasting errors, often favouring periodic events prevalent in the training data, while overlooking critical aperiodic ones like traffic incidents. To address this, we propose DualCast, a dual-branch framework that disentangles traffic signals into intrinsic spatial-temporal patterns and external environmental contexts, including aperiodic events. DualCast also employs a cross-time attention mechanism to capture high-order spatial-temporal relationships from both periodic and aperiodic patterns. DualCast is versatile. We integrate it with recent traffic forecasting models, consistently reducing their forecasting errors by up to 9.6% on multiple real datasets. Our source code is available at https://github.com/suzy0223/DualCast.
Authors: Cong Wang, Weizhe Yang, Haiping Wang, Renjie Yang, Jing Li, Zhijun Wang, Yixiong Wei, Xianli Huang, Chenshu Hu, Zhaoyang Liu, Xinyao Yu, Changqing Zou, Zhifeng Zhao
Abstract: Rapid reconstruction of 2D plasma profiles from line-integral measurements is important in nuclear fusion. This paper introduces a physics-informed model architecture called Onion, that can enhance the performance of models and be adapted to various backbone networks. The model under Onion incorporates physical information by a multiplication process and applies the physics-informed loss function according to the principle of line integration. Prediction results demonstrate that the additional input of physical information improves the deep learning model's ability, leading to a reduction in the average relative error E_1 between the reconstruction profiles and the target profiles by approximately 0.84x10^(-2) on synthetic datasets and about 0.06x10^(-2) on experimental datasets. Furthermore, the implementation of the Softplus activation function in the final two fully connected layers improves model performance. This enhancement results in a reduction in the E_1 by approximately 1.06x10^(-2) on synthetic datasets and about 0.11x10^(-2) on experimental datasets. The incorporation of the physics-informed loss function has been shown to correct the model's predictions, bringing the back-projections closer to the actual inputs and reducing the errors associated with inversion algorithms. Besides, we have developed a synthetic data model to generate customized line-integral diagnostic datasets and have also collected soft x-ray diagnostic datasets from EAST and HL-2A. This study achieves reductions in reconstruction errors, and accelerates the development of surrogate models in fusion research.
Authors: Varshita Kolipaka, Akshit Sinha, Debangan Mishra, Sumit Kumar, Arvindh Arun, Shashwat Goel, Ponnurangam Kumaraguru
Abstract: Graph Neural Networks (GNNs) are increasingly being used for a variety of ML applications on graph data. Because graph data does not follow the independently and identically distributed (i.i.d.) assumption, adversarial manipulations or incorrect data can propagate to other data points through message passing, which deteriorates the model's performance. To allow model developers to remove the adverse effects of manipulated entities from a trained GNN, we study the recently formulated problem of Corrective Unlearning. We find that current graph unlearning methods fail to unlearn the effect of manipulations even when the whole manipulated set is known. We introduce a new graph unlearning method, Cognac, which can unlearn the effect of the manipulation set even when only 5% of it is identified. It recovers most of the performance of a strong oracle with fully corrected training data, even beating retraining from scratch without the deletion set while being 8x more efficient. We hope our work assists GNN developers in mitigating harmful effects caused by issues in real-world data, post-training. Our code is publicly available at https://github.com/cognac-gnn-unlearning/corrective-unlearning-for-gnns
URLs: https://github.com/cognac-gnn-unlearning/corrective-unlearning-for-gnns
Authors: Nicol\'as Garc\'ia Trillos, Aditya Kumar Akash, Sixu Li, Konstantin Riedl, Yuhua Zhu
Abstract: Adversarial attacks pose significant challenges in many machine learning applications, particularly in the setting of distributed training and federated learning, where malicious agents seek to corrupt the training process with the goal of jeopardizing and compromising the performance and reliability of the final models. In this paper, we address the problem of robust federated learning in the presence of such attacks by formulating the training task as a bi-level optimization problem. We conduct a theoretical analysis of the resilience of consensus-based bi-level optimization (CB$^2$O), an interacting multi-particle metaheuristic optimization method, in adversarial settings. Specifically, we provide a global convergence analysis of CB$^2$O in mean-field law in the presence of malicious agents, demonstrating the robustness of CB$^2$O against a diverse range of attacks. Thereby, we offer insights into how specific hyperparameter choices enable to mitigate adversarial effects. On the practical side, we extend CB$^2$O to the clustered federated learning setting by proposing FedCB$^2$O, a novel interacting multi-particle system, and design a practical algorithm that addresses the demands of real-world applications. Extensive experiments demonstrate the robustness of the FedCB$^2$O algorithm against label-flipping attacks in decentralized clustered federated learning scenarios, showcasing its effectiveness in practical contexts.
Authors: Kangjie Zheng, Siyue Liang, Junwei Yang, Bin Feng, Zequn Liu, Wei Ju, Zhiping Xiao, Ming Zhang
Abstract: SMILES, a crucial textual representation of molecular structures, has garnered significant attention as a foundation for pre-trained language models (LMs). However, most existing pre-trained SMILES LMs focus solely on the single-token level supervision during pre-training, failing to fully leverage the substructural information of molecules. This limitation makes the pre-training task overly simplistic, preventing the models from capturing richer molecular semantic information. Moreover, during pre-training, these SMILES LMs only process corrupted SMILES inputs, never encountering any valid SMILES, which leads to a train-inference mismatch. To address these challenges, we propose SMI-Editor, a novel edit-based pre-trained SMILES LM. SMI-Editor disrupts substructures within a molecule at random and feeds the resulting SMILES back into the model, which then attempts to restore the original SMILES through an editing process. This approach not only introduces fragment-level training signals, but also enables the use of valid SMILES as inputs, allowing the model to learn how to reconstruct complete molecules from these incomplete structures. As a result, the model demonstrates improved scalability and an enhanced ability to capture fragment-level molecular information. Experimental results show that SMI-Editor achieves state-of-the-art performance across multiple downstream molecular tasks, and even outperforming several 3D molecular representation models.
Authors: Peter Lalor
Abstract: Machine learning methods in gamma spectroscopy have the potential to provide accurate, real-time classification of unknown radioactive samples. However, obtaining sufficient experimental training data is often prohibitively expensive and time-consuming, and models trained solely on simulated data can struggle to generalize to the unpredictable range of real-world operating scenarios. In this study, we explore how supervised domain adaptation techniques can improve radioisotope identification models by transferring knowledge between different data domains. We begin by pretraining a model for radioisotope identification using data from a synthetic source domain, and then fine-tune it for a new target domain that shares the same label space. Our analysis indicates that fine-tuned models significantly outperform those trained exclusively on source-domain data or solely on target-domain data, particularly in the intermediate data regime ($\approx 10^2$ to $10^5$ target training samples). This conclusion is consistent across four different machine learning architectures (MLP, CNN, Transformer, and LSTM). Furthermore, our findings show that fine-tuned Transformers yield a statistically significant improvement in test performance compared to the other architectures. Overall, this study serves as a proof of concept for applying supervised domain adaptation techniques to gamma spectroscopy in scenarios where experimental data is limited.
Authors: Daokun Zhang, Russell Tsuchida, Dino Sejdinovic
Abstract: Label distribution learning (LDL) provides a framework wherein a distribution over categories rather than a single category is predicted, with the aim of addressing ambiguity in labeled data. Existing research on LDL mainly focuses on the task of point estimation, i.e., finding an optimal distribution in the probability simplex conditioned on the given sample. In this paper, we propose a novel label distribution learning model SNEFY-LDL, which estimates a probability distribution of all possible label distributions over the simplex, by unleashing the expressive power of the recently introduced Squared Neural Family (SNEFY), a new class of tractable probability models. As a way to summarize the fitted model, we derive the closed-form label distribution mean, variance and covariance conditioned on the given sample, which can be used to predict the ground-truth label distributions, construct label distribution confidence intervals, and measure the correlations between different labels. Moreover, more information about the label distribution prediction uncertainties can be acquired from the modeled probability density function. Extensive experiments on conformal prediction, active learning and ensemble learning are conducted, verifying SNEFY-LDL's great effectiveness in LDL uncertainty quantification. The source code of this paper is available at https://github.com/daokunzhang/SNEFY-LDL.
Authors: Kai Yuan, Jiahao Zhang, Yidi Wang, Xiaobing Pei
Abstract: Adversarial attacks on Graph Neural Networks aim to perturb the performance of the learner by carefully modifying the graph topology and node attributes. Existing methods achieve attack stealthiness by constraining the modification budget and differences in graph properties. However, these methods typically disrupt task-relevant primary semantics directly, which results in low defensibility and detectability of the attack. In this paper, we propose an Adversarial Attack on High-level Semantics for Graph Neural Networks (AHSG), which is a graph structure attack model that ensures the retention of primary semantics. By combining latent representations with shared primary semantics, our model retains detectable attributes and relational patterns of the original graph while leveraging more subtle changes to carry out the attack. Then we use the Projected Gradient Descent algorithm to map the latent representations with attack effects to the adversarial graph. Through experiments on robust graph deep learning models equipped with defense strategies, we demonstrate that AHSG outperforms other state-of-the-art methods in attack effectiveness. Additionally, using Contextual Stochastic Block Models to detect the attacked graph further validates that our method preserves the primary semantics of the graph.
Authors: Yixin Ren, Haocheng Zhang, Yewei Xia, Hao Zhang, Jihong Guan, Shuigeng Zhou
Abstract: Score-based causal discovery methods can effectively identify causal relationships by evaluating candidate graphs and selecting the one with the highest score. One popular class of scores is kernel-based generalized score functions, which can adapt to a wide range of scenarios and work well in practice because they circumvent assumptions about causal mechanisms and data distributions. Despite these advantages, kernel-based generalized score functions pose serious computational challenges in time and space, with a time complexity of $\mathcal{O}(n^3)$ and a memory complexity of $\mathcal{O}(n^2)$, where $n$ is the sample size. In this paper, we propose an approximate kernel-based generalized score function with $\mathcal{O}(n)$ time and space complexities by using low-rank technique and designing a set of rules to handle the complex composite matrix operations required to calculate the score, as well as developing sampling algorithms for different data types to benefit the handling of diverse data types efficiently. Our extensive causal discovery experiments on both synthetic and real-world data demonstrate that compared to the state-of-the-art method, our method can not only significantly reduce computational costs, but also achieve comparable accuracy, especially for large datasets.
Authors: Baojun Che, Yifan Chen, Zhenghao Huan, Daniel Zhengyu Huang, Weijie Wang
Abstract: This paper is concerned with the approximation of probability distributions known up to normalization constants, with a focus on Bayesian inference for large-scale inverse problems in scientific computing. In this context, key challenges include costly repeated evaluations of forward models, multimodality, and inaccessible gradients for the forward model. To address them, we develop a variational inference framework that combines Fisher-Rao natural gradient with specialized quadrature rules to enable derivative free updates of Gaussian mixture variational families. The resulting method, termed Derivative Free Gaussian Mixture Variational Inference (DF-GMVI), guarantees covariance positivity and affine invariance, offering a stable and efficient framework for approximating complex posterior distributions. The effectiveness of DF-GMVI is demonstrated through numerical experiments on challenging scenarios, including distributions with multiple modes, infinitely many modes, and curved modes in spaces with up to 100 dimensions. The method's practicality is further demonstrated in a large-scale application, where it successfully recovers the initial conditions of the Navier-Stokes equations from solution data at positive times.
Authors: Xinye Qu, Longxiao Liu, Wenjie Huang
Abstract: In this paper, we propose a novel reinforcement learning algorithm for inventory management of newly launched products with no historical demand information. The algorithm follows the classic Dyna-$Q$ structure, balancing the model-free and model-based approaches, while accelerating the training process of Dyna-$Q$ and mitigating the model discrepancy generated by the model-based feedback. Based on the idea of transfer learning, warm-start information from the demand data of existing similar products can be incorporated into the algorithm to further stabilize the early-stage training and reduce the variance of the estimated optimal policy. Our approach is validated through a case study of bakery inventory management with real data. The adjusted Dyna-$Q$ shows up to a 23.7\% reduction in average daily cost compared with $Q$-learning, and up to a 77.5\% reduction in training time within the same horizon compared with classic Dyna-$Q$. By using transfer learning, it can be found that the adjusted Dyna-$Q$ has the lowest total cost, lowest variance in total cost, and relatively low shortage percentages among all the benchmarking algorithms under a 30-day testing.
Authors: Michael J. Zellinger, Matt Thomson
Abstract: Understanding the reliability of large language models (LLMs) has recently garnered significant attention. Given LLMs' propensity to hallucinate, as well as their high sensitivity to prompt design, it is already challenging to predict the performance of an individual LLM. However, the problem becomes more complex for compound LLM systems such as cascades, where in addition to each model's standalone performance, we must understand how the error rates of different models interact. In this paper, we present a probabilistic model for the joint performance distribution of a sequence of LLMs, which enables a framework for rationally tuning the confidence thresholds of a LLM cascade using continuous optimization. Compared to selecting confidence thresholds using Bayesian optimization, our parametric Markov-copula model yields more favorable error-cost trade-offs, improving the area under the error-cost curve by 4.3% on average for cascades with $k\geq 3$ models. In the low-sample regime with $n \leq 30$ training examples, the performance improvement widens to 10.2%, suggesting that our framework's inductive assumptions about the interactions between the error rates of different LLMs enhance sample efficiency. Overall, our Markov-copula model provides a rational basis for tuning LLM cascade performance and points to the potential of probabilistic methods in analyzing systems of LLMs.
Authors: Vladimir Vovk
Abstract: This paper argues for a wider use of the functional theory of randomness, a modification of the algorithmic theory of randomness getting rid of unspecified additive constants. Both theories are useful for understanding relationships between the assumptions of IID data and data exchangeability. While the assumption of IID data is standard in machine learning, conformal prediction relies on data exchangeability. Nouretdinov, V'yugin, and Gammerman showed, using the language of the algorithmic theory of randomness, that conformal prediction is a universal method under the assumption of IID data. In this paper (written for the Alex Gammerman Festschrift) I will selectively review connections between exchangeability and the property of being IID, early history of conformal prediction, my encounters and collaboration with Alex and other interesting people, and a translation of Nouretdinov et al.'s results into the language of the functional theory of randomness, which moves it closer to practice. Namely, the translation says that every confidence predictor that is valid for IID data can be transformed to a conformal predictor without losing much in predictive efficiency.
Authors: Pengxiang Zhao, Xiaoming Yuan
Abstract: Large Language Models (LLMs) face significant deployment challenges due to their substantial resource requirements. While low-bit quantized weights can reduce memory usage and improve inference efficiency, current hardware lacks native support for mixed-precision General Matrix Multiplication (mpGEMM), resulting in inefficient dequantization-based implementations. Moreover, uniform quantization methods often fail to capture weight distributions adequately, leading to performance degradation. We propose GANQ (GPU-Adaptive Non-Uniform Quantization), a layer-wise post-training non-uniform quantization framework optimized for hardware-efficient lookup table-based mpGEMM. GANQ achieves superior quantization performance by utilizing a training-free, GPU-adaptive optimization algorithm to efficiently reduce layer-wise quantization errors. Extensive experiments demonstrate GANQ's ability to reduce the perplexity gap from the FP16 baseline compared to state-of-the-art methods for both 3-bit and 4-bit quantization. Furthermore, when deployed on a single NVIDIA RTX 4090 GPU, GANQ's quantized models achieve up to 2.57$\times$ speedup over the baseline, advancing memory and inference efficiency in LLM deployment.
Authors: Jiacheng Liang, Yuhui Wang, Changjiang Li, Rongyi Zhu, Tanqiu Jiang, Neil Gong, Ting Wang
Abstract: GraphRAG advances retrieval-augmented generation (RAG) by structuring external knowledge as multi-scale knowledge graphs, enabling language models to integrate both broad context and granular details in their generation. While GraphRAG has demonstrated success across domains, its security implications remain largely unexplored. To bridge this gap, this work examines GraphRAG's vulnerability to poisoning attacks, uncovering an intriguing security paradox: existing RAG poisoning attacks are less effective under GraphRAG than conventional RAG, due to GraphRAG's graph-based indexing and retrieval; yet, the same features also create new attack surfaces. We present GragPoison, a novel attack that exploits shared relations in the underlying knowledge graph to craft poisoning text capable of compromising multiple queries simultaneously. GragPoison employs three key strategies: (i) relation injection to introduce false knowledge, (ii) relation enhancement to amplify poisoning influence, and (iii) narrative generation to embed malicious content within coherent text. Empirical evaluation across diverse datasets and models shows that GragPoison substantially outperforms existing attacks in terms of effectiveness (up to 98% success rate) and scalability (using less than 68% poisoning text) on multiple variations of GraphRAG. We also explore potential defensive measures and their limitations, identifying promising directions for future research.
Authors: Yuewen Mei, Tong Nie, Jian Sun, Ye Tian
Abstract: Ensuring and improving the safety of autonomous driving systems (ADS) is crucial for the deployment of highly automated vehicles, especially in safety-critical events. To address the rarity issue, adversarial scenario generation methods are developed, in which behaviors of traffic participants are manipulated to induce safety-critical events. However, existing methods still face two limitations. First, identification of the adversarial participant directly impacts the effectiveness of the generation. However, the complexity of real-world scenarios, with numerous participants and diverse behaviors, makes identification challenging. Second, the potential of generated safety-critical scenarios to continuously improve ADS performance remains underexplored. To address these issues, we propose LLM-attacker: a closed-loop adversarial scenario generation framework leveraging large language models (LLMs). Specifically, multiple LLM agents are designed and coordinated to identify optimal attackers. Then, the trajectories of the attackers are optimized to generate adversarial scenarios. These scenarios are iteratively refined based on the performance of ADS, forming a feedback loop to improve ADS. Experimental results show that LLM-attacker can create more dangerous scenarios than other methods, and the ADS trained with it achieves a collision rate half that of training with normal scenarios. This indicates the ability of LLM-attacker to test and enhance the safety and robustness of ADS. Video demonstrations are provided at: https://drive.google.com/file/d/1Zv4V3iG7825oyiKbUwS2Y-rR0DQIE1ZA/view.
URLs: https://drive.google.com/file/d/1Zv4V3iG7825oyiKbUwS2Y-rR0DQIE1ZA/view.
Authors: Song Bian, Minghao Yan, Shivaram Venkataraman
Abstract: Scaling laws are powerful tools to predict the performance of large language models. However, current scaling laws fall short of accounting for inference costs. In this work, we first show that model architecture affects inference latency, where models of the same size can have up to 3.5x difference in latency. To tackle this challenge, we modify the Chinchilla scaling laws to co-optimize the model parameter count, the number of training tokens, and the model architecture. Due to the reason that models of similar training loss exhibit gaps in downstream evaluation, we also propose a novel method to train inference-efficient models based on the revised scaling laws. We perform extensive empirical studies to fit and evaluate our inference-aware scaling laws. We vary model parameters from 80M to 1B, training tokens from 1.6B to 30B, and model shapes, training 63 models. Guided by our inference-efficient scaling law and model selection method, we release the Morph-1B model, which improves inference latency by 1.8x while maintaining accuracy on downstream tasks compared to open-source models, pushing the Pareto frontier of accuracy-latency tradeoff. Notably, our experiments reveal that wider and shallower models can yield efficiency gains while preserving accuracy.
Authors: Han Zhong, Yutong Yin, Shenao Zhang, Xiaojun Xu, Yuanxin Liu, Yifei Zuo, Zhihan Liu, Boyi Liu, Sirui Zheng, Hongyi Guo, Liwei Wang, Mingyi Hong, Zhaoran Wang
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks, yet generating reliable reasoning processes remains a significant challenge. We present a unified probabilistic framework that formalizes LLM reasoning through a novel graphical model incorporating latent thinking processes and evaluation signals. Within this framework, we introduce the Bootstrapping Reinforced Thinking Process (BRiTE) algorithm, which works in two steps. First, it generates high-quality rationales by approximating the optimal thinking process through reinforcement learning, using a novel reward shaping mechanism. Second, it enhances the base LLM by maximizing the joint probability of rationale generation with respect to the model's parameters. Theoretically, we demonstrate BRiTE's convergence at a rate of $1/T$ with $T$ representing the number of iterations. Empirical evaluations on math and coding benchmarks demonstrate that our approach consistently improves performance across different base models without requiring human-annotated thinking processes. In addition, BRiTE demonstrates superior performance compared to existing algorithms that bootstrap thinking processes use alternative methods such as rejection sampling, and can even match or exceed the results achieved through supervised fine-tuning with human-annotated data.
Authors: Yao Shu, Qixin Zhang, Kun He, Zhongxiang Dai
Abstract: Recently, zeroth-order (ZO) optimization plays an essential role in scenarios where gradient information is inaccessible or unaffordable, such as black-box systems and resource-constrained environments. While existing adaptive methods such as ZO-AdaMM have shown promise, they are fundamentally limited by their underutilization of moment information during optimization, usually resulting in underperforming convergence. To overcome these limitations, this paper introduces Refined Adaptive Zeroth-Order Optimization (R-AdaZO). Specifically, we first show the untapped variance reduction effect of first moment estimate on ZO gradient estimation, which improves the accuracy and stability of ZO updates. We then refine the second moment estimate based on these variance-reduced gradient estimates to better capture the geometry of the optimization landscape, enabling a more effective scaling of ZO updates. We present rigorous theoretical analysis to show (a) the first analysis to the variance reduction of first moment estimate in ZO optimization, (b) the improved second moment estimates with a more accurate approximation of its variance-free ideal, (c) the first variance-aware convergence framework for adaptive ZO methods, which may be of independent interest, and (d) the faster convergence of R-AdaZO than existing baselines like ZO-AdaMM. Our extensive experiments, including synthetic problems, black-box adversarial attack, and memory-efficient fine-tuning of large language models (LLMs), further verify the superior convergence of R-AdaZO, indicating that R-AdaZO offers an improved solution for real-world ZO optimization challenges.
Authors: Chaouki Ben Issaid, Praneeth Vepakomma, Mehdi Bennis
Abstract: Federated multi-task learning (FMTL) aims to simultaneously learn multiple related tasks across clients without sharing sensitive raw data. However, in the decentralized setting, existing FMTL frameworks are limited in their ability to capture complex task relationships and handle feature and sample heterogeneity across clients. To address these challenges, we introduce a novel sheaf-theoretic-based approach for FMTL. By representing client relationships using cellular sheaves, our framework can flexibly model interactions between heterogeneous client models. We formulate the sheaf-based FMTL optimization problem using sheaf Laplacian regularization and propose the Sheaf-FMTL algorithm to solve it. We show that the proposed framework provides a unified view encompassing many existing federated learning (FL) and FMTL approaches. Furthermore, we prove that our proposed algorithm, Sheaf-FMTL, achieves a sublinear convergence rate in line with state-of-the-art decentralized FMTL algorithms. Extensive experiments show that although Sheaf-FMTL introduces computational and storage overhead due to the management of interaction maps, it achieves substantial communication savings in terms of transmitted bits when compared to decentralized FMTL baselines. This trade-off makes Sheaf-FMTL especially suitable for cross-silo FL scenarios, where managing model heterogeneity and ensuring communication efficiency are essential, and where clients have adequate computational resources.
Authors: Shaofeng Yin, Jialong Wu, Siqiao Huang, Xingjian Su, Xu He, Jianye Hao, Mingsheng Long
Abstract: Heterogeneity in sensors and actuators across environments poses a significant challenge to building large-scale pre-trained world models on top of this low-dimensional sensor information. In this work, we explore pre-training world models for heterogeneous environments by addressing key transfer barriers in both data diversity and model flexibility. We introduce UniTraj, a unified dataset comprising over one million trajectories from 80 environments, designed to scale data while preserving critical diversity. Additionally, we propose TrajWorld, a novel architecture capable of flexibly handling varying sensor and actuator information and capturing environment dynamics in-context. Pre-training TrajWorld on UniTraj yields substantial gains in transition prediction, achieves a new state-of-the-art for off-policy evaluation, and also delivers superior online performance of model predictive control. To the best of our knowledge, this work, for the first time, demonstrates the transfer benefits of world models across heterogeneous and complex control environments. Code and data are available at https://github.com/thuml/TrajWorld.
Authors: Muxing Li, Zesheng Ye, Yixuan Li, Andy Song, Guangquan Zhang, Feng Liu
Abstract: To detect unauthorized data usage in training large-scale generative models (e.g., ChatGPT or Midjourney), membership inference attacks (MIA) have proven effective in distinguishing a single training instance (a member) from a single non-training instance (a non-member). This success is mainly credited to a memorization effect: models tend to perform better on a member than a non-member. However, we find that standard MIAs fail against distilled generative models (i.e., student models) that are increasingly deployed in practice for efficiency (e.g., ChatGPT 4o-mini). Trained exclusively on data generated from a large-scale model (a teacher model), the student model lacks direct exposure to any members (teacher's training data), nullifying the memorization effect that standard MIAs rely on. This finding reveals a serious privacy loophole, where generation-service providers could deploy a student model whose teacher was potentially trained on unauthorized data, yet claim the deployed model is clean because it was not directly trained on such data. Hence, are distilled models inherently unauditable for upstream privacy violations, and should we discard them when we care about privacy? We contend no, as we uncover a memory chain connecting the student and teacher's member data: the distribution of student-generated data aligns more closely with the distribution of the teacher's members than with non-members, thus we can detect unauthorized data usage even when direct instance-level memorization is absent. This leads us to posit that MIAs on distilled generative models should shift from instance-level scores to distribution-level statistics. We further propose three principles of distribution-based MIAs for detecting unauthorized training data through distilled generative models, and validate our position through an exemplar framework. We lastly discuss the implications of our position.
Authors: Tuan Truong, Chau Nguyen, Huy Nguyen, Minh Le, Trung Le, Nhat Ho
Abstract: Low-rank Adaptation (LoRA) has emerged as a powerful method for fine-tuning large-scale foundation models. Despite its popularity, the theoretical understanding of LoRA has remained limited. This paper presents a theoretical analysis of LoRA by examining its connection to the Mixture of Experts models. Under this framework, we show that simple reparameterizations of the LoRA matrices can notably accelerate the low-rank matrix estimation process. In particular, we prove that reparameterization can reduce the data needed to achieve a desired estimation error from an exponential to a polynomial scale. Motivated by this insight, we propose Reparameterized Low-Rank Adaptation (RepLoRA), which incorporates lightweight MLPs to reparameterize the LoRA matrices. Extensive experiments across multiple domains demonstrate that RepLoRA consistently outperforms vanilla LoRA. Notably, with limited data, RepLoRA surpasses LoRA by a margin of up to 40.0% and achieves LoRA's performance with only 30.0% of the training data, highlighting both the theoretical and empirical robustness of our PEFT method.
Authors: Guibin Zhang, Luyang Niu, Junfeng Fang, Kun Wang, Lei Bai, Xiang Wang
Abstract: Large Language Model (LLM)-empowered multi-agent systems extend the cognitive boundaries of individual agents through disciplined collaboration and interaction, while constructing these systems often requires labor-intensive manual designs. Despite the availability of methods to automate the design of agentic workflows, they typically seek to identify a static, complex, one-size-fits-all system, which, however, fails to dynamically allocate inference resources based on the difficulty and domain of each query. To address this challenge, we shift away from the pursuit of a monolithic agentic system, instead optimizing the \textbf{agentic supernet}, a probabilistic and continuous distribution of agentic architectures. We introduce MaAS, an automated framework that samples query-dependent agentic systems from the supernet, delivering high-quality solutions and tailored resource allocation (\textit{e.g.}, LLM calls, tool calls, token cost). Comprehensive evaluation across six benchmarks demonstrates that MaAS \textbf{(I)} requires only $6\sim45\%$ of the inference costs of existing handcrafted or automated multi-agent systems, \textbf{(II)} surpasses them by $0.54\%\sim11.82\%$, and \textbf{(III)} enjoys superior cross-dataset and cross-LLM-backbone transferability.
Authors: Shaopeng Fu, Liang Ding, Jingfeng Zhang, Di Wang
Abstract: Jailbreak attacks against large language models (LLMs) aim to induce harmful behaviors in LLMs through carefully crafted adversarial prompts. To mitigate attacks, one way is to perform adversarial training (AT)-based alignment, i.e., training LLMs on some of the most adversarial prompts to help them learn how to behave safely under attacks. During AT, the length of adversarial prompts plays a critical role in the robustness of aligned LLMs. While long-length adversarial prompts during AT might lead to strong LLM robustness, their synthesis however is very resource-consuming, which may limit the application of LLM AT. This paper focuses on adversarial suffix jailbreak attacks and unveils that to defend against a jailbreak attack with an adversarial suffix of length $\Theta(M)$, it is enough to align LLMs on prompts with adversarial suffixes of length $\Theta(\sqrt{M})$. Theoretically, we analyze the adversarial in-context learning of linear transformers on linear regression tasks and prove a robust generalization bound for trained transformers. The bound depends on the term $\Theta(\sqrt{M_{\text{test}}}/M_{\text{train}})$, where $M_{\text{train}}$ and $M_{\text{test}}$ are the numbers of adversarially perturbed in-context samples during training and testing. Empirically, we conduct AT on popular open-source LLMs and evaluate their robustness against jailbreak attacks of different adversarial suffix lengths. Results confirm a positive correlation between the attack success rate and the ratio of the square root of the adversarial suffix length during jailbreaking to the length during AT. Our findings show that it is practical to defend against ``long-length'' jailbreak attacks via efficient ``short-length'' AT. The code is available at https://github.com/fshp971/adv-icl.
Authors: Chenhao Si, Ming Yan, Xin Li, Zhihong Xia
Abstract: We propose compleX-PINN, a novel physics-informed neural network (PINN) architecture incorporating a learnable activation function inspired by the Cauchy integral theorem. By optimizing the activation parameters, compleX-PINN achieves high accuracy with just a single hidden layer. Empirically, we demonstrate that compleX-PINN solves high-dimensional problems that pose significant challenges for PINNs. Our results show that compleX-PINN consistently achieves substantially greater precision, often improving accuracy by an order of magnitude, on these complex tasks.
Authors: Stefano Zampini, Jacob K. Christopher, Luca Oneto, Davide Anguita, Ferdinando Fioretto
Abstract: Stable diffusion models represent the state-of-the-art in data synthesis across diverse domains and hold transformative potential for applications in science and engineering, e.g., by facilitating the discovery of novel solutions and simulating systems that are computationally intractable to model explicitly. While there is increasing effort to incorporate physics-based constraints into generative models, existing techniques are either limited in their applicability to latent diffusion frameworks or lack the capability to strictly enforce domain-specific constraints. To address this limitation this paper proposes a novel integration of stable diffusion models with constrained optimization frameworks, enabling the generation of outputs satisfying stringent physical and functional requirements. The effectiveness of this approach is demonstrated through material design experiments requiring adherence to precise morphometric properties, challenging inverse design tasks involving the generation of materials inducing specific stress-strain responses, and copyright-constrained content generation tasks.
Authors: Niccol\`o Ajroldi, Antonio Orvieto, Jonas Geiping
Abstract: Averaging checkpoints along the training trajectory is a simple yet powerful approach to improve the generalization performance of Machine Learning models and reduce training time. Motivated by these potential gains, and in an effort to fairly and thoroughly benchmark this technique, we present an extensive evaluation of averaging techniques in modern Deep Learning, which we perform using AlgoPerf \citep{dahl_benchmarking_2023}, a large-scale benchmark for optimization algorithms. We investigate whether weight averaging can reduce training time, improve generalization, and replace learning rate decay, as suggested by recent literature. Our evaluation across seven architectures and datasets reveals that averaging significantly accelerates training and yields considerable efficiency gains, at the price of a minimal implementation and memory cost, while mildly improving generalization across all considered workloads. Finally, we explore the relationship between averaging and learning rate annealing and show how to optimally combine the two to achieve the best performances.
Authors: Tianlang Chen, Charilaos Kanatsoulis, Jure Leskovec
Abstract: Predictive tasks on relational databases are critical in real-world applications spanning e-commerce, healthcare, and social media. To address these tasks effectively, Relational Deep Learning (RDL) encodes relational data as graphs, enabling Graph Neural Networks (GNNs) to exploit relational structures for improved predictions. However, existing RDL methods often overlook the intrinsic structural properties of the graphs built from relational databases, leading to modeling inefficiencies, particularly in handling many-to-many relationships. Here we introduce RelGNN, a novel GNN framework specifically designed to leverage the unique structural characteristics of the graphs built from relational databases. At the core of our approach is the introduction of atomic routes, which are simple paths that enable direct single-hop interactions between the source and destination nodes. Building upon these atomic routes, RelGNN designs new composite message passing and graph attention mechanisms that reduce redundancy, highlight key signals, and enhance predictive accuracy. RelGNN is evaluated on 30 diverse real-world tasks from Relbench (Fey et al., 2024), and achieves state-of-the-art performance on the vast majority of tasks, with improvements of up to 25%. Code is available at https://github.com/snap-stanford/RelGNN.
Authors: Ashab Uddin, Ahmed Hamdi Sakr, Ning Zhang
Abstract: The increasing complexity of Intelligent Transportation Systems (ITS) has led to significant interest in computational offloading to external infrastructures such as edge servers, vehicular nodes, and UAVs. These dynamic and heterogeneous environments pose challenges for traditional offloading strategies, prompting the exploration of Reinforcement Learning (RL) and Deep Reinforcement Learning (DRL) as adaptive decision-making frameworks. This survey presents a comprehensive review of recent advances in DRL-based offloading for vehicular edge computing (VEC). We classify and compare existing works based on learning paradigms (e.g., single-agent, multi-agent), system architectures (e.g., centralized, distributed, hierarchical), and optimization objectives (e.g., latency, energy, fairness). Furthermore, we analyze how Markov Decision Process (MDP) formulations are applied and highlight emerging trends in reward design, coordination mechanisms, and scalability. Finally, we identify open challenges and outline future research directions to guide the development of robust and intelligent offloading strategies for next-generation ITS.
Authors: Cong Lu, Shengran Hu, Jeff Clune
Abstract: Foundation models have become general-purpose assistants, exhibiting diverse capabilities across numerous domains through training on web-scale data. It remains challenging to precisely characterize even a fraction of the full spectrum of these abilities and potential risks in any new model. Existing evaluation approaches often require significant human effort, and it is taking increasing effort to design ever harder challenges for more capable models. We introduce Automated Capability Discovery (ACD), a framework that designates one foundation model as a scientist to systematically propose open-ended tasks probing the abilities of a subject model (potentially itself). By combining frontier models with ideas from the field of open-endedness, ACD automatically and systematically uncovers a diverse spectrum of surprising capabilities and failures in the subject model. We demonstrate ACD across a range of foundation models (including the GPT, Claude, and Llama series), showing that it automatically generates thousands of distinct tasks, which are then clustered to reveal dozens of broader capability areas and failure modes, that would be challenging for any single team to uncover. We further validate our method's automated scoring with extensive human surveys, observing high agreement between model-generated and human evaluations. By leveraging foundation models' ability to both create tasks and self-evaluate, ACD is a significant step toward scalable, automated evaluation of novel AI systems. All code and evaluation logs are open-sourced at https://github.com/conglu1997/ACD.
Authors: Nikita Morozov, Ian Maksimov, Daniil Tiapkin, Sergey Samsonov
Abstract: Generative Flow Networks (GFlowNets) are a family of generative models that learn to sample objects from a given probability distribution, potentially known up to a normalizing constant. Instead of working in the object space, GFlowNets proceed by sampling trajectories in an appropriately constructed directed acyclic graph environment, greatly relying on the acyclicity of the graph. In our paper, we revisit the theory that relaxes the acyclicity assumption and present a simpler theoretical framework for non-acyclic GFlowNets in discrete environments. Moreover, we provide various novel theoretical insights related to training with fixed backward policies, the nature of flow functions, and connections between entropy-regularized RL and non-acyclic GFlowNets, which naturally generalize the respective concepts and theoretical results from the acyclic setting. In addition, we experimentally re-examine the concept of loss stability in non-acyclic GFlowNet training, as well as validate our own theoretical findings.
Authors: Amirhesam Abedsoltan, Huaqing Zhang, Kaiyue Wen, Hongzhou Lin, Jingzhao Zhang, Mikhail Belkin
Abstract: Large language models (LLMs) exhibit remarkable task generalization, solving tasks they were never explicitly trained on with only a few demonstrations. This raises a fundamental question: When can learning from a small set of tasks generalize to a large task family? In this paper, we investigate task generalization through the lens of autoregressive compositional structure, where each task is a composition of $T$ operations, and each operation is among a finite family of $D$ subtasks. This yields a total class of size $D^T$. We first show that generalization to all $D^T$ tasks is theoretically achievable by training on only $\widetilde{O}(D)$ tasks. Empirically, we demonstrate that Transformers achieve such exponential task generalization on sparse parity functions via In-context Learning (ICL) and chain-of-thought (CoT) reasoning. We further show generalization in arithmetic and translation, beyond parity functions.
Authors: Tianren Zhang, Guanyu Chen, Feng Chen
Abstract: Humans develop world models that capture the underlying generation process of data. Whether neural networks can learn similar world models remains an open problem. In this work, we present the first theoretical results for this problem, showing that in a multi-task setting, models with a low-degree bias provably recover latent data-generating variables under mild assumptions--even if proxy tasks involve complex, non-linear functions of the latents. However, such recovery is sensitive to model architecture. Our analysis leverages Boolean models of task solutions via the Fourier-Walsh transform and introduces new techniques for analyzing invertible Boolean transforms, which may be of independent interest. We illustrate the algorithmic implications of our results and connect them to related research areas, including self-supervised learning, out-of-distribution generalization, and the linear representation hypothesis in large language models.
Authors: Guhao Feng, Yihan Geng, Jian Guan, Wei Wu, Liwei Wang, Di He
Abstract: Diffusion language models have emerged as a promising approach for text generation. One would naturally expect this method to be an efficient replacement for autoregressive models since multiple tokens can be sampled in parallel during each diffusion step. However, its efficiency-accuracy trade-off is not yet well understood. In this paper, we present a rigorous theoretical analysis of a widely used type of diffusion language model, the Masked Diffusion Model (MDM), and find that its effectiveness heavily depends on the target evaluation metric. Under mild conditions, we prove that when using perplexity as the metric, MDMs can achieve near-optimal perplexity in sampling steps regardless of sequence length, demonstrating that efficiency can be achieved without sacrificing performance. However, when using the sequence error rate--which is important for understanding the "correctness" of a sequence, such as a reasoning chain--we show that the required sampling steps must scale linearly with sequence length to obtain "correct" sequences, thereby eliminating MDM's efficiency advantage over autoregressive models. Our analysis establishes the first theoretical foundation for understanding the benefits and limitations of MDMs. All theoretical findings are supported by empirical studies.
Authors: Julien Siems, Timur Carstensen, Arber Zela, Frank Hutter, Massimiliano Pontil, Riccardo Grazzi
Abstract: Linear Recurrent Neural Networks (linear RNNs) have emerged as competitive alternatives to Transformers for sequence modeling, offering efficient training and linear-time inference. However, existing architectures face a fundamental trade-off between expressivity and efficiency, dictated by the structure of their state-transition matrices. Diagonal matrices, used in models such as Mamba, GLA, or mLSTM, yield fast runtime but have limited expressivity. To address this, recent architectures such as DeltaNet and RWKV-7 adopted a diagonal plus rank-1 structure, which allows simultaneous token and channel mixing, improving associative recall and, as recently shown, state-tracking when allowing negative eigenvalues in the state-transition matrices. Building on the interpretation of DeltaNet's recurrence as performing one step of online gradient descent per token on an associative recall loss, we introduce DeltaProduct, which instead takes multiple ($n_h$) steps per token. This naturally leads to diagonal plus rank-$n_h$ state-transition matrices, formed as products of $n_h$ generalized Householder transformations, providing a tunable mechanism to balance expressivity and efficiency. We provide a detailed theoretical characterization of the state-tracking capability of DeltaProduct in finite precision, showing how it improves by increasing $n_h$. Our extensive experiments demonstrate that DeltaProduct outperforms DeltaNet in both state-tracking and language modeling, while also showing significantly improved length extrapolation capabilities.
Authors: Madhab Barman, Madhurima Panja, Nachiketa Mishra, Tanujit Chakraborty
Abstract: Tuberculosis (TB) remains a formidable global health challenge, driven by complex spatiotemporal transmission dynamics and influenced by factors such as population mobility and behavioral changes. We propose an Epidemic-Guided Deep Learning (EGDL) approach that fuses mechanistic epidemiological principles with advanced deep learning techniques to enhance early warning systems and intervention strategies for TB outbreaks. Our framework is built upon a modified networked Susceptible-Infectious-Recovered (MN-SIR) model augmented with a saturated incidence rate and graph Laplacian diffusion, capturing both long-term transmission dynamics and region-specific population mobility patterns. Compartmental model parameters are rigorously estimated using Bayesian inference via the Markov Chain Monte Carlo approach. Theoretical analysis leveraging the comparison principle and Green's formula establishes global stability properties of the disease-free and endemic equilibria. Building on these epidemiological insights, we design two forecasting architectures, EGDL-Parallel and EGDL-Series, that integrate the mechanistic outputs of the MN-SIR model within deep neural networks. This integration mitigates the overfitting risks commonly encountered in data-driven methods and filters out noise inherent in surveillance data, resulting in reliable forecasts of real-world epidemic trends. Experiments conducted on TB incidence data from 47 prefectures in Japan and 31 provinces in mainland China demonstrate that our approach delivers robust and accurate predictions across multiple time horizons (short to medium-term forecasts), supporting its generalizability across regions with different population dynamics.
Authors: Haoyu Wang, Shikun Liu, Rongzhe Wei, Pan Li
Abstract: Large language models (LLMs) have recently been introduced to graph learning, aiming to extend their zero-shot generalization success to tasks where labeled graph data is scarce. Among these applications, inference over text-attributed graphs (TAGs) presents unique challenges: existing methods struggle with LLMs' limited context length for processing large node neighborhoods and the misalignment between node embeddings and the LLM token space. To address these issues, we establish two key principles for ensuring generalization and derive the framework LLM-BP accordingly: (1) Unifying the attribute space with task-adaptive embeddings, where we leverage LLM-based encoders and task-aware prompting to enhance generalization of the text attribute embeddings; (2) Developing a generalizable graph information aggregation mechanism, for which we adopt belief propagation with LLM-estimated parameters that adapt across graphs. Evaluations on 11 real-world TAG benchmarks demonstrate that LLM-BP significantly outperforms existing approaches, achieving 8.10% improvement with task-conditional embeddings and an additional 1.71% gain from adaptive aggregation. The code and task-adaptive embeddings are publicly available.
Authors: Ruichen Xu, Kexin Chen
Abstract: Recent theoretical studies (Kou et al., 2023; Cao et al., 2022) have revealed a sharp phase transition from benign to harmful overfitting when the noise-to-feature ratio exceeds a threshold-a situation common in long-tailed data distributions where atypical data is prevalent. However, harmful overfitting rarely happens in overparameterized neural networks. Further experimental results suggested that memorization is necessary for achieving near-optimal generalization error in long-tailed data distributions (Feldman & Zhang, 2020). We argue that this discrepancy between theoretical predictions and empirical observations arises because previous feature-noise data models overlook the heterogeneous nature of noise across different data classes. In this paper, we refine the feature-noise data model by incorporating class-dependent heterogeneous noise and re-examine the overfitting phenomenon in neural networks. Through a comprehensive analysis of the training dynamics, we establish test loss bounds for the refined model. Our findings reveal that neural networks can leverage "data noise" to learn implicit features that improve the classification accuracy for long-tailed data. Our analysis also provides a training-free metric for evaluating data influence on test performance. Experimental validation on both synthetic and real-world datasets supports our theoretical results.
Authors: Nir Rosenfeld, Haifeng Xu
Abstract: Decades of research in machine learning have given us powerful tools for making accurate predictions. This has made such tools appealing for use in social settings and on human inputs. Yet despite a lack of justification for why the generic approach of accuracy maximization can or should improve our collective well-being -- and mounting evidence of likely adverse outcomes -- it remains the widespread default. This position paper asserts that for machine learning to become socially beneficial, it must be embedded within a broader economic framework that explicitly aims to maximize social welfare. The field of welfare economics asks: how should we allocate limited resources among self-interested agents to maximize overall benefits? We contend that this perspective applies to many contemporary applications of machine learning in social contexts, and advocate for its adoption. Rather than disposing of prediction, we propose to leverage this forte of machine learning towards welfare maximization. We demonstrate this idea by portraying a conceptual framework that gradually transitions from accuracy maximization (with awareness to welfare) to welfare maximization (via accurate prediction). We detail applications and use-cases for which this framework can be effective, identify technical challenges and practical opportunities, and highlight future avenues worth pursuing.
Authors: Xingyue Huang, Pablo Barcel\'o, Michael M. Bronstein, \.Ismail \.Ilkan Ceylan, Mikhail Galkin, Juan L Reutter, Miguel Romero Orth
Abstract: Knowledge Graph Foundation Models (KGFMs) are at the frontier for deep learning on knowledge graphs (KGs), as they can generalize to completely novel knowledge graphs with different relational vocabularies. Despite their empirical success, our theoretical understanding of KGFMs remains very limited. In this paper, we conduct a rigorous study of the expressive power of KGFMs. Specifically, we show that the expressive power of KGFMs directly depends on the motifs that are used to learn the relation representations. We then observe that the most typical motifs used in the existing literature are binary, as the representations are learned based on how pairs of relations interact, which limits the model's expressiveness. As part of our study, we design more expressive KGFMs using richer motifs, which necessitate learning relation representations based on, e.g., how triples of relations interact with each other. Finally, we empirically validate our theoretical findings, showing that the use of richer motifs results in better performance on a wide range of datasets drawn from different domains.
Authors: Bishwajit Prasad Gond
Abstract: The increasing complexity of supply chains and the rising costs associated with defective or substandard goods (bad goods) highlight the urgent need for advanced predictive methodologies to mitigate risks and enhance operational efficiency. This research presents a novel framework that integrates Time Series ARIMA (AutoRegressive Integrated Moving Average) models with a proprietary formula specifically designed to calculate bad goods after time series forecasting. By leveraging historical data patterns, including sales, returns, and capacity, the model forecasts potential quality failures, enabling proactive decision-making. ARIMA is employed to capture temporal trends in time series data, while the newly developed formula quantifies the likelihood and impact of defects with greater precision. Experimental results, validated on a dataset spanning 2022-2024 for Organic Beer-G 1 Liter, demonstrate that the proposed method outperforms traditional statistical models, such as Exponential Smoothing and Holt-Winters, in both prediction accuracy and risk evaluation. This study advances the field of predictive analytics by bridging time series forecasting, ARIMA, and risk management in supply chain quality control, offering a scalable and practical solution for minimizing losses due to bad goods.
Authors: Jonathan Light, Wei Cheng, Benjamin Riviere, Wu Yue, Masafumi Oyamada, Mengdi Wang, Yisong Yue, Santiago Paternain, Haifeng Chen
Abstract: Inference scaling methods for large language models often work by breaking problems into steps or groups of tokens, then sampling and selecting the best next steps. However, these steps and their sizes are usually fixed or manually designed based on domain knowledge. We introduce dynamic decomposition, a method that adaptively and automatically breaks down solution and reasoning traces into manageable steps during inference. By allocating compute more effectively - especially by subdividing difficult steps and prioritizing their sampling - dynamic decomposition significantly boosts inference efficiency. Experiments on benchmarks like APPS, MATH, and LiveCodeBench show that dynamic decomposition outperforms fixed strategies such as token-level, sentence-level, and single-step decompositions, reducing the pass@10 error rate by 5.0%, 6.7%, and 10.5% respectively. These results show the promise of dynamic decomposition for improving a broad range of inference scaling techniques.
Authors: Dang Nguyen, Zeman Li, Mohammadhossein Bateni, Vahab Mirrokni, Meisam Razaviyayn, Baharan Mirzasoleiman
Abstract: Synthetic data has the potential to improve the performance, training efficiency, and privacy of real training examples. Nevertheless, existing approaches for synthetic text generation are mostly heuristics and cannot generate human-readable text without compromising the privacy of real data, or provide performance guarantees for training Large Language Models (LLMs). In this work, we propose the first theoretically rigorous approach for generating synthetic human-readable text that provides convergence, performance, and privacy guarantees for fine-tuning LLMs on a target task. To do so, we leverage Alternating Direction Method of Multipliers (ADMM) that iteratively optimizes the embeddings of synthetic examples to match the noisy gradient of the target training or validation data, and maps them to a sequence of text tokens with low perplexity. In doing so, the generated synthetic text guarantees convergence of the model to a close neighborhood of the solution obtained by fine-tuning on real data and preserves their privacy. Experiments on various classification tasks confirm the effectiveness of our proposed approach. Our code is available at https://github.com/BigML-CS-UCLA/GRADMM.
Authors: Taneesh Gupta, Rahul Madhavan, Xuchao Zhang, Chetan Bansal, Saravan Rajmohan
Abstract: Multi-preference optimization enriches language-model alignment beyond pairwise preferences by contrasting entire sets of helpful and undesired responses, thereby enabling richer training signals for large language models. During self-play alignment, these models often produce numerous candidate answers per query, rendering it computationally infeasible to include all responses in the training objective. In this work, we propose $\textit{Active Multi-Preference Optimization}$ (AMPO), a novel approach that combines on-policy generation, a multi-preference group-contrastive loss, and active subset selection. Specifically, we score and embed large candidate pools of responses and then select a small, yet informative, subset that covers reward extremes and distinct semantic clusters for preference optimization. Our contrastive training scheme is capable of identifying not only the best and worst answers but also subtle, underexplored modes that are crucial for robust alignment. Theoretically, we provide guarantees for expected reward maximization using our active selection method, and empirically, AMPO achieves state-of-the-art results on $\textit{AlpacaEval}$ using Llama 8B and Mistral 7B. We release our datasets $\href{https://huggingface.co/Multi-preference-Optimization}{here}$.
Authors: Nicolas Michel, Maorong Wang, Jiangpeng He, Toshihiko Yamasaki
Abstract: Continual Learning (CL) aims to learn from a non-stationary data stream where the underlying distribution changes over time. While recent advances have produced efficient memory-free methods in the offline CL (offCL) setting, where tasks are known in advance and data can be revisited, online CL (onCL) remains dominated by memory-based approaches. The transition from offCL to onCL is challenging, as many offline methods rely on (1) prior knowledge of task boundaries and (2) sophisticated scheduling or optimization schemes, both of which are unavailable when data arrives sequentially and can be seen only once. In this paper, we investigate the adaptation of state-of-the-art memory-free offCL methods to the online setting. We first show that augmenting these methods with lightweight prototypes significantly improves performance, albeit at the cost of increased Gradient Imbalance, resulting in a biased learning towards earlier tasks. To address this issue, we introduce Fine-Grained Hypergradients, an online mechanism for rebalancing gradient updates during training. Our experiments demonstrate that the synergy between prototype memory and hypergradient reweighting substantially enhances the performance of memory-free methods in onCL and surpasses onCL baselines. Code will be released upon acceptance.
Authors: Vladimir Vovk
Abstract: Conformal predictors provide set or functional predictions that are valid under the assumption of randomness, i.e., under the assumption of independent and identically distributed data. The question asked in this paper is whether there are predictors that are valid in the same sense under the assumption of randomness and that are more efficient than conformal predictors. The answer is that the class of conformal predictors is universal in that only limited gains in predictive efficiency are possible. The previous work in this area has relied on the algorithmic theory of randomness and so involved unspecified constants, whereas this paper's results are much more practical. They are also shown to be optimal in some respects.
Authors: Albert Gong, Kamil\.e Stankevi\v{c}i\=ut\.e, Chao Wan, Anmol Kabra, Raphael Thesmar, Johann Lee, Julius Klenke, Carla P. Gomes, Kilian Q. Weinberger
Abstract: High-quality benchmarks are essential for evaluating reasoning and retrieval capabilities of large language models (LLMs). However, curating datasets for this purpose is not a permanent solution as they are prone to data leakage and inflated performance results. To address these challenges, we propose PhantomWiki: a pipeline to generate unique, factually consistent document corpora with diverse question-answer pairs. Unlike prior work, PhantomWiki is neither a fixed dataset, nor is it based on any existing data. Instead, a new PhantomWiki instance is generated on demand for each evaluation. We vary the question difficulty and corpus size to disentangle reasoning and retrieval capabilities respectively, and find that PhantomWiki datasets are surprisingly challenging for frontier LLMs. Thus, we contribute a scalable and data leakage-resistant framework for disentangled evaluation of reasoning, retrieval, and tool-use abilities. Our code is available at https://github.com/kilian-group/phantom-wiki.
Authors: Hao Li, Yu-Hao Huang, Chang Xu, Viktor Schlegel, Renhe Jiang, Riza Batista-Navarro, Goran Nenadic, Jiang Bian
Abstract: Time-series Generation (TSG) is a prominent research area with broad applications in simulations, data augmentation, and counterfactual analysis. While existing methods have shown promise in unconditional single-domain TSG, real-world applications demand for cross-domain approaches capable of controlled generation tailored to domain-specific constraints and instance-level requirements. In this paper, we argue that text can provide semantic insights, domain information and instance-specific temporal patterns, to guide and improve TSG. We introduce ``Text-Controlled TSG'', a task focused on generating realistic time series by incorporating textual descriptions. To address data scarcity in this setting, we propose a novel LLM-based Multi-Agent framework that synthesizes diverse, realistic text-to-TS datasets. Furthermore, we introduce BRIDGE, a hybrid text-controlled TSG framework that integrates semantic prototypes with text description for supporting domain-level guidance. This approach achieves state-of-the-art generation fidelity on 11 of 12 datasets, and improves controllability by up to 12% on MSE and 6% MAE compared to no text input generation, highlighting its potential for generating tailored time-series data.
Authors: Marta Skreta, Tara Akhound-Sadegh, Viktor Ohanesian, Roberto Bondesan, Al\'an Aspuru-Guzik, Arnaud Doucet, Rob Brekelmans, Alexander Tong, Kirill Neklyudov
Abstract: While score-based generative models are the model of choice across diverse domains, there are limited tools available for controlling inference-time behavior in a principled manner, e.g. for composing multiple pretrained models. Existing classifier-free guidance methods use a simple heuristic to mix conditional and unconditional scores to approximately sample from conditional distributions. However, such methods do not approximate the intermediate distributions, necessitating additional `corrector' steps. In this work, we provide an efficient and principled method for sampling from a sequence of annealed, geometric-averaged, or product distributions derived from pretrained score-based models. We derive a weighted simulation scheme which we call Feynman-Kac Correctors (FKCs) based on the celebrated Feynman-Kac formula by carefully accounting for terms in the appropriate partial differential equations (PDEs). To simulate these PDEs, we propose Sequential Monte Carlo (SMC) resampling algorithms that leverage inference-time scaling to improve sampling quality. We empirically demonstrate the utility of our methods by proposing amortized sampling via inference-time temperature annealing, improving multi-objective molecule generation using pretrained models, and improving classifier-free guidance for text-to-image generation. Our code is available at https://github.com/martaskrt/fkc-diffusion.
Authors: Yuyan Ni, Shikun Feng, Haohan Chi, Bowen Zheng, Huan-ang Gao, Wei-Ying Ma, Zhi-Ming Ma, Yanyan Lan
Abstract: Diffusion-based models have shown great promise in molecular generation but often require a large number of sampling steps to generate valid samples. In this paper, we introduce a novel Straight-Line Diffusion Model (SLDM) to tackle this problem, by formulating the diffusion process to follow a linear trajectory. The proposed process aligns well with the noise sensitivity characteristic of molecular structures and uniformly distributes reconstruction effort across the generative process, thus enhancing learning efficiency and efficacy. Consequently, SLDM achieves state-of-the-art performance on 3D molecule generation benchmarks, delivering a 100-fold improvement in sampling efficiency.
Authors: Wonjun Kang, Kevin Galim, Yuchen Zeng, Minjae Lee, Hyung Il Koo, Nam Ik Cho
Abstract: State Space Models (SSMs) have emerged as efficient alternatives to Transformers, mitigating their quadratic computational cost. However, the application of Parameter-Efficient Fine-Tuning (PEFT) methods to SSMs remains largely unexplored. In particular, prompt-based methods like Prompt Tuning and Prefix-Tuning, which are widely used in Transformers, do not perform well on SSMs. To address this, we propose state-based methods as a superior alternative to prompt-based methods. This new family of methods naturally stems from the architectural characteristics of SSMs. State-based methods adjust state-related features directly instead of depending on external prompts. Furthermore, we introduce a novel state-based PEFT method: State-offset Tuning. At every timestep, our method directly affects the state at the current step, leading to more effective adaptation. Through extensive experiments across diverse datasets, we demonstrate the effectiveness of our method. Code is available at https://github.com/furiosa-ai/ssm-state-tuning.
Authors: Pengle Zhang, Jia Wei, Jintao Zhang, Jun Zhu, Jianfei Chen
Abstract: Transformer models have achieved remarkable success across various AI applications but face significant training costs. Low-bit training, such as INT8 training, can leverage computational units with higher throughput, and has already demonstrated its effectiveness on GPT2 models with block-level quantization. However, it struggles with modern Transformer variants incorporating GLU units. This is because those variants demonstrate complex distributions of activation outliers. To address the challenge, we propose Fallback Quantization, implementing mixed-precision GEMM that dynamically falls back 8-bit to 16-bit for activation blocks containing outliers. Experiments show that our approach is robustly competent in both fine-tuning and pretraining settings. Moreover, our method achieves a 1.57x end-to-end training speedup on RTX4090 GPUs.
Authors: Runlong Zhou, Maryam Fazel, Simon S. Du
Abstract: Reinforcement learning from human feedback (RLHF) has become essential for improving language model capabilities, but traditional approaches rely on the assumption that human preferences follow a transitive Bradley-Terry model. This assumption fails to capture the non-transitive nature of populational human preferences. Nash learning from human feedback (NLHF), targeting non-transitive preferences, is a problem of computing the Nash equilibrium (NE) of the two-player constant-sum game defined by the human preference. We introduce Extragradient preference optimization (EGPO), a novel algorithm for NLHF achieving last-iterate linear convergence to the NE of KL-regularized games and polynomial convergence to the NE of original games, while being robust to noise. Unlike previous approaches that rely on nested optimization, we derive an equivalent implementation using gradients of an online variant of the identity preference optimization (IPO) loss, enabling more faithful implementation for neural networks. Our empirical evaluations demonstrate EGPO's superior performance over baseline methods when training for the same number of epochs, as measured by pairwise win-rates using the ground truth preference. These results validate both the theoretical strengths and practical advantages of EGPO for language model alignment with non-transitive human preferences.
Authors: Ping Zhang, Zheda Mai, Quang-Huy Nguyen, Wei-Lun Chao
Abstract: Semi-supervised learning (SSL) leverages abundant unlabeled data alongside limited labeled data to enhance learning. As vision foundation models (VFMs) increasingly serve as the backbone of vision applications, it remains unclear how SSL interacts with these pre-trained models. To address this gap, we develop new SSL benchmark datasets where frozen VFMs underperform and systematically evaluate representative SSL methods. We make a surprising observation: parameter-efficient fine-tuning (PEFT) using only labeled data often matches SSL performance, even without leveraging unlabeled data. This motivates us to revisit self-training, a conceptually simple SSL baseline, where we use the supervised PEFT model to pseudo-label unlabeled data for further training. To overcome the notorious issue of noisy pseudo-labels, we propose ensembling multiple PEFT approaches and VFM backbones to produce more robust pseudo-labels. Empirical results validate the effectiveness of this simple yet powerful approach, providing actionable insights into SSL with VFMs and paving the way for more scalable and practical semi-supervised learning in the era of foundation models.
Authors: Jieming Bian, Lei Wang, Letian Zhang, Jie Xu
Abstract: Fine-tuning large language models (LLMs) in federated settings enables privacy-preserving adaptation but suffers from cross-client interference due to model aggregation. Existing federated LoRA fine-tuning methods, primarily based on FedAvg, struggle with data heterogeneity, leading to harmful cross-client interference and suboptimal personalization. In this work, we propose \textbf{FedALT}, a novel personalized federated LoRA fine-tuning algorithm that fundamentally departs from FedAvg. Instead of using an aggregated model to initialize local training, each client continues training its individual LoRA while incorporating shared knowledge through a separate Rest-of-World (RoW) LoRA component. To effectively balance local adaptation and global information, FedALT introduces an adaptive mixer that dynamically learns input-specific weightings between the individual and RoW LoRA components, drawing conceptual foundations from the Mixture-of-Experts (MoE) paradigm. Through extensive experiments on NLP benchmarks, we demonstrate that FedALT significantly outperforms state-of-the-art personalized federated LoRA fine-tuning methods, achieving superior local adaptation without sacrificing computational efficiency.
Authors: Abhi Kamboj, Minh N. Do
Abstract: Multimodal alignment constructs a joint latent vector space where modalities representing the same concept map to neighboring latent vectors. We formulate this as an inverse problem and show that, under certain conditions, paired data from each modality can map to equivalent latent vectors, which we refer to as perfect alignment. When perfect alignment cannot be achieved, it can be approximated using the Singular Value Decomposition (SVD) of a multimodal data matrix. Experiments on synthetic multimodal Gaussian data verify the effectiveness of our perfect alignment method compared to a learned contrastive alignment method. We further demonstrate the practical application of cross-modal transfer for human action recognition, showing that perfect alignment significantly enhances the model's accuracy. We conclude by discussing how these findings can be applied to various modalities and tasks and the limitations of our method. We hope these findings inspire further exploration of perfect alignment and its applications in representation learning.
Authors: Shih-Min Yang, Martin Magnusson, Johannes A. Stork, Todor Stoyanov
Abstract: Soft Actor-Critic (SAC) has achieved notable success in continuous control tasks but struggles in sparse reward settings, where infrequent rewards make efficient exploration challenging. While novelty-based exploration methods address this issue by encouraging the agent to explore novel states, they are not trivial to apply to SAC. In particular, managing the interaction between novelty-based exploration and SAC's stochastic policy can lead to inefficient exploration and redundant sample collection. In this paper, we propose KEA (Keeping Exploration Alive) which tackles the inefficiencies in balancing exploration strategies when combining SAC with novelty-based exploration. KEA integrates a novelty-augmented SAC with a standard SAC agent, proactively coordinated via a switching mechanism. This coordination allows the agent to maintain stochasticity in high-novelty regions, enhancing exploration efficiency and reducing repeated sample collection. We first analyze this potential issue in a 2D navigation task, and then evaluate KEA on the DeepSea hard-exploration benchmark as well as sparse reward control tasks from the DeepMind Control Suite. Compared to state-of-the-art novelty-based exploration baselines, our experiments show that KEA significantly improves learning efficiency and robustness in sparse reward setups.
Authors: Ran Wang, Xinlei Zhou, Meng Hu, Rihao Li, Wenhui Wu, Yuheng Jia
Abstract: Despite the remarkable success of deep neural networks (DNNs), the security threat of adversarial attacks poses a significant challenge to the reliability of DNNs. In this paper, both theoretically and empirically, we discover a universal phenomenon that has been neglected in previous works, i.e., adversarial attacks tend to shift the distributions of feature statistics. Motivated by this finding, and by leveraging the advantages of uncertainty-aware stochastic methods in building robust models efficiently, we propose an uncertainty-driven feature statistics adjustment module for robustness enhancement, named Feature Statistics with Uncertainty (FSU). It randomly resamples channel-wise feature means and standard deviations of examples from multivariate Gaussian distributions, which helps to reconstruct the perturbed examples and calibrate the shifted distributions. The calibration recovers some domain characteristics of the data for classification, thereby mitigating the influence of perturbations and weakening the ability of attacks to deceive models. The proposed FSU module has universal applicability in training, attacking, predicting, and fine-tuning, demonstrating impressive robustness enhancement ability at a trivial additional time cost. For example, by fine-tuning the well-established models with FSU, the state-of-the-art methods achieve up to 17.13% and 34.82% robustness improvement against powerful AA and CW attacks on benchmark datasets.
Authors: Ashkan Yousefpour, Taeheon Kim, Ryan S. Kwon, Seungbeen Lee, Wonje Jeung, Seungju Han, Alvin Wan, Harrison Ngan, Youngjae Yu, Jonghyun Choi
Abstract: Large Language Models (LLMs) have emerged as powerful tools, but their inherent safety risks - ranging from harmful content generation to broader societal harms - pose significant challenges. These risks can be amplified by the recent adversarial attacks, fine-tuning vulnerabilities, and the increasing deployment of LLMs in high-stakes environments. Existing safety-enhancing techniques, such as fine-tuning with human feedback or adversarial training, are still vulnerable as they address specific threats and often fail to generalize across unseen attacks, or require manual system-level defenses. This paper introduces RepBend, a novel approach that fundamentally disrupts the representations underlying harmful behaviors in LLMs, offering a scalable solution to enhance (potentially inherent) safety. RepBend brings the idea of activation steering - simple vector arithmetic for steering model's behavior during inference - to loss-based fine-tuning. Through extensive evaluation, RepBend achieves state-of-the-art performance, outperforming prior methods such as Circuit Breaker, RMU, and NPO, with up to 95% reduction in attack success rates across diverse jailbreak benchmarks, all with negligible reduction in model usability and general capabilities.
Authors: Tomohiro Hayase, Beno\^it Collins, Nakamasa Inoue
Abstract: Hierarchical inductive biases are hypothesized to promote generalizable policies in reinforcement learning, as demonstrated by explicit hyperbolic latent representations and architectures. Therefore, a more flexible approach is to have these biases emerge naturally from the algorithm. We introduce Free Random Projection, an input mapping grounded in free probability theory that constructs random orthogonal matrices where hierarchical structure arises inherently. The free random projection integrates seamlessly into existing in-context reinforcement learning frameworks by encoding hierarchical organization within the input space without requiring explicit architectural modifications. Empirical results on multi-environment benchmarks show that free random projection consistently outperforms the standard random projection, leading to improvements in generalization. Furthermore, analyses within linearly solvable Markov decision processes and investigations of the spectrum of kernel random matrices reveal the theoretical underpinnings of free random projection's enhanced performance, highlighting its capacity for effective adaptation in hierarchically structured state spaces.
Authors: MaryLena Bleile
Abstract: From clinical dosing algorithms to autonomous robots, sequential decision-making systems routinely operate with missing or incomplete data. Classical reinforcement learning theory, which is commonly used to solve sequential decision problems, assumes Markovian observability, which may not hold under partial observability. Causal inference paradigms formalise ignorability of missingness. We show these views can be unified and generalized in order to guarantee Q-learning convergence even when the Markov property fails. To do so, we introduce the concept of \emph{relative ignorability}. Relative ignorability is a graphical-causal criterion which refines the requirements for accurate decision-making based on incomplete data. Theoretical results and simulations both reveal that non-markovian stochastic processes whose missingness is relatively ignorable with respect to causal estimands can still be optimized using standard Reinforcement Learning algorithms. These results expand the theoretical foundations of safe, data-efficient AI to real-world environments where complete information is unattainable.
Authors: Sujay Nagaraj, Yang Liu, Flavio P. Calmon, Berk Ustun
Abstract: Machine learning models are routinely used to support decisions that affect individuals -- be it to screen a patient for a serious illness or to gauge their response to treatment. In these tasks, we are limited to learning models from datasets with noisy labels. In this paper, we study the instance-level impact of learning under label noise. We introduce a notion of regret for this regime, which measures the number of unforeseen mistakes due to noisy labels. We show that standard approaches to learning under label noise can return models that perform well at a population-level while subjecting individuals to a lottery of mistakes. We present a versatile approach to estimate the likelihood of mistakes at the individual-level from a noisy dataset by training models over plausible realizations of datasets without label noise. This is supported by a comprehensive empirical study of label noise in clinical prediction tasks. Our results reveal how failure to anticipate mistakes can compromise model reliability and adoption -- we demonstrate how we can address these challenges by anticipating and avoiding regretful decisions.
Authors: Michal Lukasik, Lin Chen, Harikrishna Narasimhan, Aditya Krishna Menon, Wittawat Jitkrittum, Felix X. Yu, Sashank J. Reddi, Gang Fu, Mohammadhossein Bateni, Sanjiv Kumar
Abstract: Bipartite ranking is a fundamental supervised learning problem, with the goal of learning a ranking over instances with maximal Area Under the ROC Curve (AUC) against a single binary target label. However, one may often observe multiple binary target labels, e.g., from distinct human annotators. How can one synthesize such labels into a single coherent ranking? In this work, we formally analyze two approaches to this problem -- loss aggregation and label aggregation -- by characterizing their Bayes-optimal solutions. We show that while both approaches can yield Pareto-optimal solutions, loss aggregation can exhibit label dictatorship: one can inadvertently (and undesirably) favor one label over others. This suggests that label aggregation can be preferable to loss aggregation, which we empirically verify.
Authors: Phung Lai, Xiaopeng Jiang, Hai Phan, Cristian Borcea, Khang Tran, An Chen, Vijaya Datta Mayyuri, Ruoming Jin
Abstract: Federated Learning (FL) allows collaborative training among multiple devices without data sharing, thus enabling privacy-sensitive applications on mobile or Internet of Things (IoT) devices, such as mobile health and asset tracking. However, designing an FL system with good model utility that works with low computation/communication overhead on heterogeneous, resource-constrained mobile/IoT devices is challenging. To address this problem, this paper proposes FedX, a novel adaptive model decomposition and quantization FL system for IoT. To balance utility with resource constraints on IoT devices, FedX decomposes a global FL model into different sub-networks with adaptive numbers of quantized bits for different devices. The key idea is that a device with fewer resources receives a smaller sub-network for lower overhead but utilizes a larger number of quantized bits for higher model utility, and vice versa. The quantization operations in FedX are done at the server to reduce the computational load on devices. FedX iteratively minimizes the losses in the devices' local data and in the server's public data using quantized sub-networks under a regularization term, and thus it maximizes the benefits of combining FL with model quantization through knowledge sharing among the server and devices in a cost-effective training process. Extensive experiments show that FedX significantly improves quantization times by up to 8.43X, on-device computation time by 1.5X, and total end-to-end training time by 1.36X, compared with baseline FL systems. We guarantee the global model convergence theoretically and validate local model convergence empirically, highlighting FedX's optimization efficiency.
Authors: Aaron Mueller, Atticus Geiger, Sarah Wiegreffe, Dana Arad, Iv\'an Arcuschin, Adam Belfki, Yik Siu Chan, Jaden Fiotto-Kaufman, Tal Haklay, Michael Hanna, Jing Huang, Rohan Gupta, Yaniv Nikankin, Hadas Orgad, Nikhil Prakash, Anja Reusch, Aruna Sankaranarayanan, Shun Shao, Alessandro Stolfo, Martin Tutek, Amir Zur, David Bau, Yonatan Belinkov
Abstract: How can we know whether new mechanistic interpretability methods achieve real improvements? In pursuit of lasting evaluation standards, we propose MIB, a Mechanistic Interpretability Benchmark, with two tracks spanning four tasks and five models. MIB favors methods that precisely and concisely recover relevant causal pathways or causal variables in neural language models. The circuit localization track compares methods that locate the model components - and connections between them - most important for performing a task (e.g., attribution patching or information flow routes). The causal variable localization track compares methods that featurize a hidden vector, e.g., sparse autoencoders (SAEs) or distributed alignment search (DAS), and align those features to a task-relevant causal variable. Using MIB, we find that attribution and mask optimization methods perform best on circuit localization. For causal variable localization, we find that the supervised DAS method performs best, while SAE features are not better than neurons, i.e., non-featurized hidden vectors. These findings illustrate that MIB enables meaningful comparisons, and increases our confidence that there has been real progress in the field.
Authors: Saksham Rastogi, Pratyush Maini, Danish Pruthi
Abstract: Given how large parts of publicly available text are crawled to pretrain large language models (LLMs), data creators increasingly worry about the inclusion of their proprietary data for model training without attribution or licensing. Their concerns are also shared by benchmark curators whose test-sets might be compromised. In this paper, we present STAMP, a framework for detecting dataset membership-i.e., determining the inclusion of a dataset in the pretraining corpora of LLMs. Given an original piece of content, our proposal involves first generating multiple rephrases, each embedding a watermark with a unique secret key. One version is to be released publicly, while others are to be kept private. Subsequently, creators can compare model likelihoods between public and private versions using paired statistical tests to prove membership. We show that our framework can successfully detect contamination across four benchmarks which appear only once in the training data and constitute less than 0.001% of the total tokens, outperforming several contamination detection and dataset inference baselines. We verify that STAMP preserves both the semantic meaning and utility of the original data. We apply STAMP to two real-world scenarios to confirm the inclusion of paper abstracts and blog articles in the pretraining corpora.
Authors: Cong Xu, Wenbin Liang, Mo Yu, Anan Liu, Ke-Yue Zhang, Shunli Wang, Lizhuang Ma, Jianyong Wang, Jun Wang, Wei Zhang
Abstract: The rapid scaling of models has led to prohibitively high training and fine-tuning costs. A major factor accounting for memory consumption is the widespread use of stateful optimizers (e.g., Adam), which maintain auxiliary information of even 2x the model size in order to achieve optimal convergence. We therefore present SOLO in this work to spawn a novel type of optimizer that requires an extremely light memory footprint. While previous efforts have achieved certain success in 8-bit or 4-bit cases, SOLO enables Adam-style optimizers to maintain quantized states with precision as low as 3 bits, or even 2 bits. This immense progress is due to the identification and resolution of two key challenges: the signal swamping problem in unsigned quantization that results in unchanged state dynamics, and the increased gradient variance in signed quantization that leads to incorrect descent directions. The theoretical analysis suggests a tailored logarithmic quantization for the former and a precision-specific momentum hyperparameter for the latter. SOLO can thus be seamlessly applied to Adam-style optimizers, leading to substantial memory savings with minimal accuracy loss.
Authors: Linjing You, Jiabao Lu, Xiayuan Huang
Abstract: Deep neural networks often degrade under distribution shifts. Although domain adaptation offers a solution, privacy constraints often prevent access to source data, making Test-Time Adaptation (TTA, which adapts using only unlabeled test data) increasingly attractive. However, current TTA methods still face practical challenges: (1) a primary focus on instance-wise alignment, overlooking CORrelation ALignment (CORAL) due to missing source correlations; (2) complex backpropagation operations for model updating, resulting in overhead computation and (3) domain forgetting. To address these challenges, we provide a theoretical analysis to investigate the feasibility of Test-time Correlation Alignment (TCA), demonstrating that correlation alignment between high-certainty instances and test instances can enhance test performances with a theoretical guarantee. Based on this, we propose two simple yet effective algorithms: LinearTCA and LinearTCA+. LinearTCA applies a simple linear transformation to achieve both instance and correlation alignment without additional model updates, while LinearTCA+ serves as a plug-and-play module that can easily boost existing TTA methods. Extensive experiments validate our theoretical insights and show that TCA methods significantly outperforms baselines across various tasks, benchmarks and backbones. Notably, LinearTCA achieves higher accuracy with only 4% GPU memory and 0.6% computation time compared to the best TTA baseline. It also outperforms existing methods on CLIP over 1.86%.
Authors: Qingyuan Wu, Yuhui Wang, Simon Sinong Zhan, Yixuan Wang, Chung-Wei Lin, Chen Lv, Qi Zhu, J\"urgen Schmidhuber, Chao Huang
Abstract: Reinforcement learning (RL) with delays is challenging as sensory perceptions lag behind the actual events: the RL agent needs to estimate the real state of its environment based on past observations. State-of-the-art (SOTA) methods typically employ recursive, step-by-step forecasting of states. This can cause the accumulation of compounding errors. To tackle this problem, our novel belief estimation method, named Directly Forecasting Belief Transformer (DFBT), directly forecasts states from observations without incrementally estimating intermediate states step-by-step. We theoretically demonstrate that DFBT greatly reduces compounding errors of existing recursively forecasting methods, yielding stronger performance guarantees. In experiments with D4RL offline datasets, DFBT reduces compounding errors with remarkable prediction accuracy. DFBT's capability to forecast state sequences also facilitates multi-step bootstrapping, thus greatly improving learning efficiency. On the MuJoCo benchmark, our DFBT-based method substantially outperforms SOTA baselines. Code is available at https://github.com/QingyuanWuNothing/DFBT.
Authors: Thanh Tran, Viet-Hoang Tran, Thanh Chu, Trang Pham, Laurent El Ghaoui, Tam Le, Tan M. Nguyen
Abstract: Tree-Sliced methods have recently emerged as an alternative to the traditional Sliced Wasserstein (SW) distance, replacing one-dimensional lines with tree-based metric spaces and incorporating a splitting mechanism for projecting measures. This approach enhances the ability to capture the topological structures of integration domains in Sliced Optimal Transport while maintaining low computational costs. Building on this foundation, we propose a novel nonlinear projectional framework for the Tree-Sliced Wasserstein (TSW) distance, substituting the linear projections in earlier versions with general projections, while ensuring the injectivity of the associated Radon Transform and preserving the well-definedness of the resulting metric. By designing appropriate projections, we construct efficient metrics for measures on both Euclidean spaces and spheres. Finally, we validate our proposed metric through extensive numerical experiments for Euclidean and spherical datasets. Applications include gradient flows, self-supervised learning, and generative models, where our methods demonstrate significant improvements over recent SW and TSW variants.
Authors: Nisarg Parikh, Nigel Fernandez, Alexander Scarlatos, Simon Woodhead, Andrew Lan
Abstract: Large language models (LLMs) are increasingly used to generate distractors for multiple-choice questions (MCQs), especially in domains like math education. However, existing approaches are limited in ensuring that the generated distractors are consistent with common student errors. We propose LookAlike, a method that improves error-distractor consistency via preference optimization. Our two main innovations are: (a) mining synthetic preference pairs from model inconsistencies, and (b) alternating supervised fine-tuning (SFT) with Direct Preference Optimization (DPO) to stabilize training. Unlike prior work that relies on heuristics or manually annotated preference data, LookAlike uses its own generation inconsistencies as dispreferred samples, thus enabling scalable and stable training. Evaluated on a real-world dataset of 1,400+ math MCQs, LookAlike achieves 51.6% accuracy in distractor generation and 57.2% in error generation under LLM-as-a-judge evaluation, outperforming an existing state-of-the-art method (45.6% / 47.7%). These improvements highlight the effectiveness of preference-based regularization and inconsistency mining for generating consistent math MCQ distractors at scale.
Authors: Jiancong Xiao, Bojian Hou, Zhanliang Wang, Ruochen Jin, Qi Long, Weijie J. Su, Li Shen
Abstract: One of the key technologies for the success of Large Language Models (LLMs) is preference alignment. However, a notable side effect of preference alignment is poor calibration: while the pre-trained models are typically well-calibrated, LLMs tend to become poorly calibrated after alignment with human preferences. In this paper, we investigate why preference alignment affects calibration and how to address this issue. For the first question, we observe that the preference collapse issue in alignment undesirably generalizes to the calibration scenario, causing LLMs to exhibit overconfidence and poor calibration. To address this, we demonstrate the importance of fine-tuning with domain-specific knowledge to alleviate the overconfidence issue. To further analyze whether this affects the model's performance, we categorize models into two regimes: calibratable and non-calibratable, defined by bounds of Expected Calibration Error (ECE). In the calibratable regime, we propose a calibration-aware fine-tuning approach to achieve proper calibration without compromising LLMs' performance. However, as models are further fine-tuned for better performance, they enter the non-calibratable regime. For this case, we develop an EM-algorithm-based ECE regularization for the fine-tuning loss to maintain low calibration error. Extensive experiments validate the effectiveness of the proposed methods.
Authors: Sarah Hartman, Cheng Soon Ong, Julia Powles, Petra Kuhnert
Abstract: This position paper argues that achieving meaningful scientific and societal advances with artificial intelligence (AI) requires a responsible, application-driven approach (RAD) to AI research. As AI is increasingly integrated into society, AI researchers must engage with the specific contexts where AI is being applied. This includes being responsive to ethical and legal considerations, technical and societal constraints, and public discourse. We present the case for RAD-AI to drive research through a three-staged approach: (1) building transdisciplinary teams and people-centred studies; (2) addressing context-specific methods, ethical commitments, assumptions, and metrics; and (3) testing and sustaining efficacy through staged testbeds and a community of practice. We present a vision for the future of application-driven AI research to unlock new value through technically feasible methods that are adaptive to the contextual needs and values of the communities they ultimately serve.
Authors: Gleb Molodtsov, Daniil Medyakov, Sergey Skorik, Nikolas Khachaturov, Shahane Tigranyan, Vladimir Aletov, Aram Avetisyan, Martin Tak\'a\v{c}, Aleksandr Beznosikov
Abstract: Recent advancements in machine learning have improved performance while also increasing computational demands. While federated and distributed setups address these issues, their structure is vulnerable to malicious influences. In this paper, we address a specific threat, Byzantine attacks, where compromised clients inject adversarial updates to derail global convergence. We combine the trust scores concept with trial function methodology to dynamically filter outliers. Our methods address the critical limitations of previous approaches, allowing functionality even when Byzantine nodes are in the majority. Moreover, our algorithms adapt to widely used scaled methods like Adam and RMSProp, as well as practical scenarios, including local training and partial participation. We validate the robustness of our methods by conducting extensive experiments on both synthetic and real ECG data collected from medical institutions. Furthermore, we provide a broad theoretical analysis of our algorithms and their extensions to aforementioned practical setups. The convergence guarantees of our methods are comparable to those of classical algorithms developed without Byzantine interference.
Authors: Jingwei Li, Jing Xu, Zifan Wang, Huishuai Zhang, Jingzhao Zhang
Abstract: One explanation for the strong generalization ability of neural networks is implicit bias. Yet, the definition and mechanism of implicit bias in non-linear contexts remains little understood. In this work, we propose to characterize implicit bias by the count of connected regions in the input space with the same predicted label. Compared with parameter-dependent metrics (e.g., norm or normalized margin), region count can be better adapted to nonlinear, overparameterized models, because it is determined by the function mapping and is invariant to reparametrization. Empirically, we found that small region counts align with geometrically simple decision boundaries and correlate well with good generalization performance. We also observe that good hyper-parameter choices such as larger learning rates and smaller batch sizes can induce small region counts. We further establish the theoretical connections and explain how larger learning rate can induce small region counts in neural networks.
Authors: Kalyan Cherukuri, Aarav Lala, Yash Yardi
Abstract: We propose Q-Policy, a hybrid quantum-classical reinforcement learning (RL) framework that mathematically accelerates policy evaluation and optimization by exploiting quantum computing primitives. Q-Policy encodes value functions in quantum superposition, enabling simultaneous evaluation of multiple state-action pairs via amplitude encoding and quantum parallelism. We introduce a quantum-enhanced policy iteration algorithm with provable polynomial reductions in sample complexity for the evaluation step, under standard assumptions. To demonstrate the technical feasibility and theoretical soundness of our approach, we validate Q-Policy on classical emulations of small discrete control tasks. Due to current hardware and simulation limitations, our experiments focus on showcasing proof-of-concept behavior rather than large-scale empirical evaluation. Our results support the potential of Q-Policy as a theoretical foundation for scalable RL on future quantum devices, addressing RL scalability challenges beyond classical approaches.
Authors: Kalyan Cherukuri, Aarav Lala
Abstract: As generative agents become increasingly capable, alignment of their behavior with complex human values remains a fundamental challenge. Existing approaches often simplify human intent through reduction to a scalar reward, overlooking the multi-faceted nature of human feedback. In this work, we introduce a theoretical framework for preference-based Multi-Objective Inverse Reinforcement Learning (MO-IRL), where human preferences are modeled as latent vector-valued reward functions. We formalize the problem of recovering a Pareto-optimal reward representation from noisy preference queries and establish conditions for identifying the underlying multi-objective structure. We derive tight sample complexity bounds for recovering $\epsilon$-approximations of the Pareto front and introduce a regret formulation to quantify suboptimality in this multi-objective setting. Furthermore, we propose a provably convergent algorithm for policy optimization using preference-inferred reward cones. Our results bridge the gap between practical alignment techniques and theoretical guarantees, providing a principled foundation for learning aligned behaviors in a high-dimension and value-pluralistic environment.
Authors: Qingyu Song, Peiyu Liao, Wenqian Zhao, Yiwen Wang, Shoubo Hu, Hui-Ling Zhen, Ning Jiang, Mingxuan Yuan
Abstract: The increasing deployment of Large Language Models (LLMs) on edge devices, driven by model advancements and hardware improvements, offers significant privacy benefits. However, these on-device LLMs inherently face performance limitations due to reduced model capacity and necessary compression techniques. To address this, we introduce a systematic methodology -- encompassing model capability, development efficiency, and system resources -- for evaluating on-device LLMs. Our comprehensive evaluation, encompassing models from 0.5B to 14B parameters and seven post-training quantization (PTQ) methods on commodity laptops, yields several critical insights: 1) System-level metrics exhibit near-linear scaling with effective bits-per-weight (BPW). 2) A practical threshold exists around $\sim$3.5 effective BPW, larger models subjected to low-bit quantization consistently outperform smaller models utilizing higher bit-precision. 3) Quantization with low BPW incurs marginal accuracy loss but significant memory savings. 4) Determined by low-level implementation specifics power consumption on CPU, where computation-intensive operations spend more power than memory-intensive ones. These findings offer crucial insights and practical guidelines for the efficient deployment and optimized configuration of LLMs on resource-constrained edge devices. Our codebase is available at https://github.com/simmonssong/LLMOnDevice.
Authors: Xinxing Shi, Xiaoyu Jiang, Mauricio A. \'Alvarez
Abstract: Gaussian Process (GP) Variational Autoencoders (VAEs) extend standard VAEs by replacing the fully factorised Gaussian prior with a GP prior, thereby capturing richer correlations among latent variables. However, performing exact GP inference in large-scale GPVAEs is computationally prohibitive, often forcing existing approaches to rely on restrictive kernel assumptions or large sets of inducing points. In this work, we propose a neighbour-driven approximation strategy that exploits local adjacencies in the latent space to achieve scalable GPVAE inference. By confining computations to the nearest neighbours of each data point, our method preserves essential latent dependencies, allowing more flexible kernel choices and mitigating the need for numerous inducing points. Through extensive experiments on tasks including representation learning, data imputation, and conditional generation, we demonstrate that our approach outperforms other GPVAE variants in both predictive performance and computational efficiency.
Authors: Diyuan Wu, Aleksandr Shevchenko, Samet Oymak, Marco Mondelli
Abstract: Token embeddings play a crucial role in language modeling but, despite this practical relevance, their theoretical understanding remains limited. Our paper addresses the gap by characterizing the structure of embeddings obtained via gradient descent. Specifically, we consider a one-layer softmax attention model with a linear head for binary classification, i.e., $\texttt{Softmax}( p^\top E_X^\top ) E_X v = \frac{ \sum_{i=1}^T \exp(p^\top E_{x_i}) E_{x_i}^\top v}{\sum_{j=1}^T \exp(p^\top E_{x_{j}}) }$, where $E_X = [ E_{x_1} , \dots, E_{x_T} ]^\top$ contains the embeddings of the input sequence, $p$ is the embedding of the $\mathrm{\langle cls \rangle}$ token and $v$ the output vector. First, we show that, already after a single step of gradient training with the logistic loss, the embeddings $E_X$ capture the importance of tokens in the dataset by aligning with the output vector $v$ proportionally to the frequency with which the corresponding tokens appear in the dataset. Then, after training $p$ via gradient flow until convergence, the softmax selects the important tokens in the sentence (i.e., those that are predictive of the label), and the resulting $\mathrm{\langle cls \rangle}$ embedding maximizes the margin for such a selection. Experiments on real-world datasets (IMDB, Yelp) exhibit a phenomenology close to that unveiled by our theory.
Authors: Jie Hu, Yi-Ting Ma, Do Young Eun
Abstract: We propose a history-driven target (HDT) framework in Markov Chain Monte Carlo (MCMC) to improve any random walk algorithm on discrete state spaces, such as general undirected graphs, for efficient sampling from target distribution $\boldsymbol{\mu}$. With broad applications in network science and distributed optimization, recent innovations like the self-repellent random walk (SRRW) achieve near-zero variance by prioritizing under-sampled states through transition kernel modifications based on past visit frequencies. However, SRRW's reliance on explicit computation of transition probabilities for all neighbors at each step introduces substantial computational overhead, while its strict dependence on time-reversible Markov chains excludes advanced non-reversible MCMC methods. To overcome these limitations, instead of direct modification of transition kernel, HDT introduces a history-dependent target distribution $\boldsymbol{\pi}[\mathbf{x}]$ to replace the original target $\boldsymbol{\mu}$ in any graph sampler, where $\mathbf{x}$ represents the empirical measure of past visits. This design preserves lightweight implementation by requiring only local information between the current and proposed states and achieves compatibility with both reversible and non-reversible MCMC samplers, while retaining unbiased samples with target distribution $\boldsymbol{\mu}$ and near-zero variance performance. Extensive experiments in graph sampling demonstrate consistent performance gains, and a memory-efficient Least Recently Used (LRU) cache ensures scalability to large general graphs.
Authors: Mudit Gaur, Prashant Trivedi, Sasidhar Kunapuli, Amrit Singh Bedi, Vaneet Aggarwal
Abstract: Diffusion models have demonstrated state-of-the-art performance across vision, language, and scientific domains. Despite their empirical success, prior theoretical analyses of the sample complexity suffer from poor scaling with input data dimension or rely on unrealistic assumptions such as access to exact empirical risk minimizers. In this work, we provide a principled analysis of score estimation, establishing a sample complexity bound of $\widetilde{\mathcal{O}}(\epsilon^{-6})$. Our approach leverages a structured decomposition of the score estimation error into statistical, approximation, and optimization errors, enabling us to eliminate the exponential dependence on neural network parameters that arises in prior analyses. It is the first such result which achieves sample complexity bounds without assuming access to the empirical risk minimizer of score function estimation loss.
Authors: Afrah Farea, Saiful Khan, Reza Daryani, Emre Cenk Ersan, Mustafa Serdar Celebi
Abstract: We introduce neural network architectures that combine physics-informed neural networks (PINNs) with the immersed boundary method (IBM) to solve fluid-structure interaction (FSI) problems. Our approach features two distinct architectures: a Single-FSI network with a unified parameter space, and an innovative Eulerian-Lagrangian network that maintains separate parameter spaces for fluid and structure domains. We study each architecture using standard Tanh and adaptive B-spline activation functions. Empirical studies on a 2D cavity flow problem involving a moving solid structure show that the Eulerian-Lagrangian architecture performs significantly better. The adaptive B-spline activation further enhances accuracy by providing locality-aware representation near boundaries. While our methodology shows promising results in predicting the velocity field, pressure recovery remains challenging due to the absence of explicit force-coupling constraints in the current formulation. Our findings underscore the importance of domain-specific architectural design and adaptive activation functions for modeling FSI problems within the PINN framework.
Authors: Hao Wu, Yuan Gao, Ruiqi Shu, Zean Han, Fan Xu, Zhihong Zhu, Qingsong Wen, Xian Wu, Kun Wang, Xiaomeng Huang
Abstract: Accurately predicting the long-term evolution of turbulence is crucial for advancing scientific understanding and optimizing engineering applications. However, existing deep learning methods face significant bottlenecks in long-term autoregressive prediction, which exhibit excessive smoothing and fail to accurately track complex fluid dynamics. Our extensive experimental and spectral analysis of prevailing methods provides an interpretable explanation for this shortcoming, identifying Spectral Bias as the core obstacle. Concretely, spectral bias is the inherent tendency of models to favor low-frequency, smooth features while overlooking critical high-frequency details during training, thus reducing fidelity and causing physical distortions in long-term predictions. Building on this insight, we propose Turb-L1, an innovative turbulence prediction method, which utilizes a Hierarchical Dynamics Synthesis mechanism within a multi-grid architecture to explicitly overcome spectral bias. It accurately captures cross-scale interactions and preserves the fidelity of high-frequency dynamics, enabling reliable long-term tracking of turbulence evolution. Extensive experiments on the 2D turbulence benchmark show that Turb-L1 demonstrates excellent performance: (I) In long-term predictions, it reduces Mean Squared Error (MSE) by $80.3\%$ and increases Structural Similarity (SSIM) by over $9\times$ compared to the SOTA baseline, significantly improving prediction fidelity. (II) It effectively overcomes spectral bias, accurately reproducing the full enstrophy spectrum and maintaining physical realism in high-wavenumber regions, thus avoiding the spectral distortions or spurious energy accumulation seen in other methods.
Authors: Claas Voelcker, Anastasiia Pedan, Arash Ahmadian, Romina Abachi, Igor Gilitschenski, Amir-massoud Farahmand
Abstract: The idea of value-aware model learning, that models should produce accurate value estimates, has gained prominence in model-based reinforcement learning. The MuZero loss, which penalizes a model's value function prediction compared to the ground-truth value function, has been utilized in several prominent empirical works in the literature. However, theoretical investigation into its strengths and weaknesses is limited. In this paper, we analyze the family of value-aware model learning losses, which includes the popular MuZero loss. We show that these losses, as normally used, are uncalibrated surrogate losses, which means that they do not always recover the correct model and value function. Building on this insight, we propose corrections to solve this issue. Furthermore, we investigate the interplay between the loss calibration, latent model architectures, and auxiliary losses that are commonly employed when training MuZero-style agents. We show that while deterministic models can be sufficient to predict accurate values, learning calibrated stochastic models is still advantageous.
Authors: Zhaokun Wang, Jinyu Guo, Jingwen Pu, Lingfeng Chen, Hongli Pu, Jie Ou, Libo Qin, Wenhong Tian
Abstract: Current parameter-efficient fine-tuning methods for adapting pre-trained language models to downstream tasks are susceptible to interference from noisy data. Conventional noise-handling approaches either rely on laborious data pre-processing or employ model architecture modifications prone to error accumulation. In contrast to existing noise-process paradigms, we propose a noise-robust adaptation method via asymmetric LoRA poisoning experts (LoPE), a novel framework that enhances model robustness to noise only with generated noisy data. Drawing inspiration from the mixture-of-experts architecture, LoPE strategically integrates a dedicated poisoning expert in an asymmetric LoRA configuration. Through a two-stage paradigm, LoPE performs noise injection on the poisoning expert during fine-tuning to enhance its noise discrimination and processing ability. During inference, we selectively mask the dedicated poisoning expert to leverage purified knowledge acquired by normal experts for noise-robust output. Extensive experiments demonstrate that LoPE achieves strong performance and robustness purely through the low-cost noise injection, which completely eliminates the requirement of data cleaning.
Authors: Masahiro Kato, Yuki Ikeda, Kentaro Baba, Takashi Imai, Ryo Inokuchi
Abstract: In this study, we propose a method for identifying potential customers in targeted marketing by applying learning from positive and unlabeled data (PU learning). We consider a scenario in which a company sells a product and can observe only the customers who purchased it. Decision-makers seek to market products effectively based on whether people have loyalty to the company. Individuals with loyalty are those who are likely to remain interested in the company even without additional advertising. Consequently, those loyal customers would likely purchase from the company if they are interested in the product. In contrast, people with lower loyalty may overlook the product or buy similar products from other companies unless they receive marketing attention. Therefore, by focusing marketing efforts on individuals who are interested in the product but do not have strong loyalty, we can achieve more efficient marketing. To achieve this goal, we consider how to learn, from limited data, a classifier that identifies potential customers who (i) have interest in the product and (ii) do not have loyalty to the company. Although our algorithm comprises a single-stage optimization, its objective function implicitly contains two losses derived from standard PU learning settings. For this reason, we refer to our approach as double PU learning. We verify the validity of the proposed algorithm through numerical experiments, confirming that it functions appropriately for the problem at hand.
Authors: Yeping Jin, Lauren Wise, Ioannis Ch. Paschalidis
Abstract: We introduce an innovative approach that incorporates a Distributionally Robust Learning (DRL) approach into Cox regression to enhance the robustness and accuracy of survival predictions. By formulating a DRL framework with a Wasserstein distance-based ambiguity set, we develop a variant Cox model that is less sensitive to assumptions about the underlying data distribution and more resilient to model misspecification and data perturbations. By leveraging Wasserstein duality, we reformulate the original min-max DRL problem into a tractable regularized empirical risk minimization problem, which can be computed by exponential conic programming. We provide guarantees on the finite sample behavior of our DRL-Cox model. Moreover, through extensive simulations and real world case studies, we demonstrate that our regression model achieves superior performance in terms of prediction accuracy and robustness compared with traditional methods.
Authors: Zachary Coalson, Juhan Bae, Nicholas Carlini, Sanghyun Hong
Abstract: We study how training data contributes to the emergence of toxic behaviors in large-language models. Most prior work on reducing model toxicity adopts $reactive$ approaches, such as fine-tuning pre-trained (and potentially toxic) models to align them with human values. In contrast, we propose a $proactive$ approach$-$IF-Guide$-$which leverages influence functions to identify harmful tokens within any training data and suppress their impact during training. To this end, we first show that standard influence functions are ineffective at discovering harmful training records. We then present a novel adaptation that measures token-level attributions from training data to model toxicity, along with techniques for selecting toxic training documents and a learning objective that can be integrated into both pre-training and fine-tuning. Moreover, IF-Guide does not rely on human-preference data, which is typically required by existing alignment methods. In evaluation, we demonstrate that IF-Guide substantially reduces both explicit and implicit toxicity$-$by up to 10$\times$ compared to uncensored models, and up to 3$\times$ compared to baseline alignment methods, e.g., DPO and RAD$-$across both pre-training and fine-tuning scenarios. IF-Guide is computationally efficient: a billion-parameter model is $not$ $necessary$ for computing influence scores; a million-parameter model$-$with 7.5$\times$ fewer parameters$-$can effectively serve as a proxy for identifying harmful data. Our code is publicly available at: https://github.com/ztcoalson/IF-Guide
Authors: Xiaojun Shan, Qi Cao, Xing Han, Haofei Yu, Paul Pu Liang
Abstract: Recent advances in multimodal foundation models have achieved state-of-the-art performance across a range of tasks. These breakthroughs are largely driven by new pre-training paradigms that leverage large-scale, unlabeled multimodal data, followed by instruction fine-tuning on curated labeled datasets and high-quality prompts. While there is growing interest in scaling instruction fine-tuning to ever-larger datasets in both quantity and scale, our findings reveal that simply increasing the number of instruction-tuning tasks does not consistently yield better performance. Instead, we observe that grouping tasks by the common interactions across modalities, such as discovering redundant shared information, prioritizing modality selection with unique information, or requiring synergistic fusion to discover new information from both modalities, encourages the models to learn transferrable skills within a group while suppressing interference from mismatched tasks. To this end, we introduce MINT, a simple yet surprisingly effective task-grouping strategy based on the type of multimodal interaction. We demonstrate that the proposed method greatly outperforms existing task grouping baselines for multimodal instruction tuning, striking an effective balance between generalization and specialization.
Authors: Jiaxi Hu, Yongqi Pan, Jusen Du, Disen Lan, Xiaqiang Tang, Qingsong Wen, Yuxuan Liang, Weigao Sun
Abstract: Recent efficient sequence modeling methods such as Gated DeltaNet, TTT, and RWKV-7 have achieved performance improvements by supervising the recurrent memory management through Delta learning rule. Unlike previous state-space models (e.g., Mamba) and gated linear attentions (e.g., GLA), these models introduce interactions between the recurrent state and the key vector, structurally resembling bilinear systems. In this paper, we first introduce the concept of Bilinear RNNs with a comprehensive analysis on the advantages and limitations of these models. Then, based on closed-loop control theory, we propose a novel Bilinear RNN variant named Comba, which adopts a scalar-plus-low-rank state transition, with both state feedback and output feedback corrections. We also implement a hardware-efficient chunk-wise parallel kernel in Triton and train models with 340M/1.3B parameters on large-scale corpus. Comba demonstrates superior performance and computation efficiency in both language and vision modeling.
Authors: Yang Guo, Yutian Tao, Yifei Ming, Robert D. Nowak, Yingyu Liang
Abstract: Retrieval-augmented generation (RAG) has seen many empirical successes in recent years by aiding the LLM with external knowledge. However, its theoretical aspect has remained mostly unexplored. In this paper, we propose the first finite-sample generalization bound for RAG in in-context linear regression and derive an exact bias-variance tradeoff. Our framework views the retrieved texts as query-dependent noisy in-context examples and recovers the classical in-context learning (ICL) and standard RAG as the limit cases. Our analysis suggests that an intrinsic ceiling on generalization error exists on RAG as opposed to the ICL. Furthermore, our framework is able to model retrieval both from the training data and from external corpora by introducing uniform and non-uniform RAG noise. In line with our theory, we show the sample efficiency of ICL and RAG empirically with experiments on common QA benchmarks, such as Natural Questions and TriviaQA.
Authors: Xiansheng Cai, Sihan Hu, Tao Wang, Yuan Huang, Pan Zhang, Youjin Deng, Kun Chen
Abstract: Fundamental physics often confronts complex symbolic problems with few guiding exemplars or established principles. While artificial intelligence (AI) offers promise, its typical need for vast datasets to learn from hinders its use in these information-scarce frontiers. We introduce learning at criticality (LaC), a reinforcement learning (RL) scheme that tunes Large Language Models (LLMs) to a sharp learning transition, addressing this information scarcity. At this transition, LLMs achieve peak generalization from minimal data, exemplified by 7-digit base-7 addition -- a test of nontrivial arithmetic reasoning. To elucidate this peak, we analyze a minimal concept-network model (CoNet) designed to capture the essence of how LLMs might link tokens. Trained on a single exemplar, this model also undergoes a sharp learning transition. This transition exhibits hallmarks of a second-order phase transition, notably power-law distributed solution path lengths. At this critical point, the system maximizes a ``critical thinking pattern" crucial for generalization, enabled by the underlying scale-free exploration. This suggests LLMs reach peak performance by operating at criticality, where such explorative dynamics enable the extraction of underlying operational rules. We demonstrate LaC in quantum field theory: an 8B-parameter LLM, tuned to its critical point by LaC using a few exemplars of symbolic Matsubara sums, solves unseen, higher-order problems, significantly outperforming far larger models. LaC thus leverages critical phenomena, a physical principle, to empower AI for complex, data-sparse challenges in fundamental physics.
Authors: Seohong Park, Kevin Frans, Deepinder Mann, Benjamin Eysenbach, Aviral Kumar, Sergey Levine
Abstract: In this work, we study the scalability of offline reinforcement learning (RL) algorithms. In principle, a truly scalable offline RL algorithm should be able to solve any given problem, regardless of its complexity, given sufficient data, compute, and model capacity. We investigate if and how current offline RL algorithms match up to this promise on diverse, challenging, previously unsolved tasks, using datasets up to 1000x larger than typical offline RL datasets. We observe that despite scaling up data, many existing offline RL algorithms exhibit poor scaling behavior, saturating well below the maximum performance. We hypothesize that the horizon is the main cause behind the poor scaling of offline RL. We empirically verify this hypothesis through several analysis experiments, showing that long horizons indeed present a fundamental barrier to scaling up offline RL. We then show that various horizon reduction techniques substantially enhance scalability on challenging tasks. Based on our insights, we also introduce a minimal yet scalable method named SHARSA that effectively reduces the horizon. SHARSA achieves the best asymptotic performance and scaling behavior among our evaluation methods, showing that explicitly reducing the horizon unlocks the scalability of offline RL. Code: https://github.com/seohongpark/horizon-reduction
Authors: Viet Nguyen, Changjian Shui, Vijay Giri, Siddarth Arya, Amol Verma, Fahad Razak, Rahul G. Krishnan
Abstract: The distribution of data changes over time; models operating operating in dynamic environments need retraining. But knowing when to retrain, without access to labels, is an open challenge since some, but not all shifts degrade model performance. This paper formalizes and addresses the problem of post-deployment deterioration (PDD) monitoring. We propose D3M, a practical and efficient monitoring algorithm based on the disagreement of predictive models, achieving low false positive rates under non-deteriorating shifts and provides sample complexity bounds for high true positive rates under deteriorating shifts. Empirical results on both standard benchmark and a real-world large-scale internal medicine dataset demonstrate the effectiveness of the framework and highlight its viability as an alert mechanism for high-stakes machine learning pipelines.
Authors: Michal Klein, Alireza Mousavi-Hosseini, Stephen Zhang, Marco Cuturi
Abstract: Flow models transform data gradually from one modality (e.g. noise) onto another (e.g. images). Such models are parameterized by a time-dependent velocity field, trained to fit segments connecting pairs of source and target points. When the pairing between source and target points is given, training flow models boils down to a supervised regression problem. When no such pairing exists, as is the case when generating data from noise, training flows is much harder. A popular approach lies in picking source and target points independently. This can, however, lead to velocity fields that are slow to train, but also costly to integrate at inference time. In theory, one would greatly benefit from training flow models by sampling pairs from an optimal transport (OT) measure coupling source and target, since this would lead to a highly efficient flow solving the Benamou and Brenier dynamical OT problem. In practice, recent works have proposed to sample mini-batches of $n$ source and $n$ target points and reorder them using an OT solver to form better pairs. These works have advocated using batches of size $n\approx 256$, and considered OT solvers that return couplings that are either sharp (using e.g. the Hungarian algorithm) or blurred (using e.g. entropic regularization, a.k.a. Sinkhorn). We follow in the footsteps of these works by exploring the benefits of increasing $n$ by three to four orders of magnitude, and look more carefully on the effect of the entropic regularization $\varepsilon$ used in the Sinkhorn algorithm. Our analysis is facilitated by new scale invariant quantities to report the sharpness of a coupling, while our sharded computations across multiple GPU or GPU nodes allow scaling up $n$. We show that in both synthetic and image generation tasks, flow models greatly benefit when fitted with large Sinkhorn couplings, with a low entropic regularization $\varepsilon$.
Authors: Sajjad Abdoli, Freeman Lewin, Gediminas Vasiliauskas, Fabian Schonholz
Abstract: The development of modern Artificial Intelligence (AI) models, particularly diffusion-based models employed in computer vision and image generation tasks, is undergoing a paradigmatic shift in development methodologies. Traditionally dominated by a "Model Centric" approach, in which performance gains were primarily pursued through increasingly complex model architectures and hyperparameter optimization, the field is now recognizing a more nuanced "Data-Centric" approach. This emergent framework foregrounds the quality, structure, and relevance of training data as the principal driver of model performance. To operationalize this paradigm shift, we introduce the DataSeeds.AI sample dataset (the "DSD"), initially comprised of approximately 10,610 high-quality human peer-ranked photography images accompanied by extensive multi-tier annotations. The DSD is a foundational computer vision dataset designed to usher in a new standard for commercial image datasets. Representing a small fraction of DataSeeds.AI's 100 million-plus image catalog, the DSD provides a scalable foundation necessary for robust commercial and multimodal AI development. Through this in-depth exploratory analysis, we document the quantitative improvements generated by the DSD on specific models against known benchmarks and make the code and the trained models used in our evaluation publicly available.
Authors: Haotian Jiang, Zeyu Bao, Shida Wang, Qianxiao Li
Abstract: The evolution of sequence modeling architectures, from recurrent neural networks and convolutional models to Transformers and structured state-space models, reflects ongoing efforts to address the diverse temporal dependencies inherent in sequential data. Despite this progress, systematically characterizing the strengths and limitations of these architectures remains a fundamental challenge. In this work, we propose a synthetic benchmarking framework to evaluate how effectively different sequence models capture distinct temporal structures. The core of this approach is to generate synthetic targets, each characterized by a memory function and a parameter that determines the strength of temporal dependence. This setup allows us to produce a continuum of tasks that vary in temporal complexity, enabling fine-grained analysis of model behavior concerning specific memory properties. We focus on four representative memory functions, each corresponding to a distinct class of temporal structures. Experiments on several sequence modeling architectures confirm existing theoretical insights and reveal new findings. These results demonstrate the effectiveness of the proposed method in advancing theoretical understanding and highlight the importance of using controllable targets with clearly defined structures for evaluating sequence modeling architectures.
Authors: Yihan Xie, Sijing Li, Tianwei Lin, Zhuonan Wang, Chenglin Yang, Yu Zhong, Wenqiao Zhang, Haoyuan Li, Hao Jiang, Fengda Zhang, Qishan Chen, Jun Xiao, Yueting Zhuang, Beng Chin Ooi
Abstract: We present Heartcare Suite, a multimodal comprehensive framework for finegrained electrocardiogram (ECG) understanding. It comprises three key components: (i) Heartcare-220K, a high-quality, structured, and comprehensive multimodal ECG dataset covering essential tasks such as disease diagnosis, waveform morphology analysis, and rhythm interpretation. (ii) Heartcare-Bench, a systematic and multi-dimensional benchmark designed to evaluate diagnostic intelligence and guide the optimization of Medical Multimodal Large Language Models (Med-MLLMs) in ECG scenarios. and (iii) HeartcareGPT with a tailored tokenizer Bidirectional ECG Abstract Tokenization (Beat), which compresses raw multi-lead signals into semantically rich discrete tokens via duallevel vector quantization and query-guided bidirectional diffusion mechanism. Built upon Heartcare-220K, HeartcareGPT achieves strong generalization and SoTA performance across multiple clinically meaningful tasks. Extensive experiments demonstrate that Heartcare Suite is highly effective in advancing ECGspecific multimodal understanding and evaluation. Our project is available at https://github.com/DCDmllm/Heartcare-Suite .
Authors: Adil Hasan, Thomas Peyrin
Abstract: Significant investments have been made towards the commodification of diffusion models for generation of diverse media. Their mass-market adoption is however still hobbled by the intense hardware resource requirements of diffusion model inference. Model quantization strategies tailored specifically towards diffusion models have been useful in easing this burden, yet have generally explored the Uniform Scalar Quantization (USQ) family of quantization methods. In contrast, Vector Quantization (VQ) methods, which operate on groups of multiple related weights as the basic unit of compression, have seen substantial success in Large Language Model (LLM) quantization. In this work, we apply codebook-based additive vector quantization to the problem of diffusion model compression. Our resulting approach achieves a new Pareto frontier for the extremely low-bit weight quantization on the standard class-conditional benchmark of LDM-4 on ImageNet at 20 inference time steps. Notably, we report sFID 1.92 points lower than the full-precision model at W4A8 and the best-reported results for FID, sFID and ISC at W2A8. We are also able to demonstrate FLOPs savings on arbitrary hardware via an efficient inference kernel, as opposed to savings resulting from small integer operations which may lack broad hardware support.
Authors: Rujikorn Charakorn, Edoardo Cetin, Yujin Tang, Robert Tjarko Lange
Abstract: While Foundation Models provide a general tool for rapid content creation, they regularly require task-specific adaptation. Traditionally, this exercise involves careful curation of datasets and repeated fine-tuning of the underlying model. Fine-tuning techniques enable practitioners to adapt foundation models for many new applications but require expensive and lengthy training while being notably sensitive to hyperparameter choices. To overcome these limitations, we introduce Text-to-LoRA (T2L), a model capable of adapting large language models (LLMs) on the fly solely based on a natural language description of the target task. T2L is a hypernetwork trained to construct LoRAs in a single inexpensive forward pass. After training T2L on a suite of 9 pre-trained LoRA adapters (GSM8K, Arc, etc.), we show that the ad-hoc reconstructed LoRA instances match the performance of task-specific adapters across the corresponding test sets. Furthermore, T2L can compress hundreds of LoRA instances and zero-shot generalize to entirely unseen tasks. This approach provides a significant step towards democratizing the specialization of foundation models and enables language-based adaptation with minimal compute requirements. Our code is available at https://github.com/SakanaAI/text-to-lora
Authors: Bruce W. Lee, Addie Foote, Alex Infanger, Leni Shor, Harish Kamath, Jacob Goldman-Wetzler, Bryce Woodworth, Alex Cloud, Alexander Matt Turner
Abstract: Current LLM unlearning methods are not robust: they can be reverted easily with a few steps of finetuning. This is true even for the idealized unlearning method of training to imitate an oracle model that was never exposed to unwanted information, suggesting that output-based finetuning is insufficient to achieve robust unlearning. In a similar vein, we find that training a randomly initialized student to imitate an unlearned model transfers desired behaviors while leaving undesired capabilities behind. In other words, distillation robustifies unlearning. Building on this insight, we propose Unlearn-Noise-Distill-on-Outputs (UNDO), a scalable method that distills an unlearned model into a partially noised copy of itself. UNDO introduces a tunable tradeoff between compute cost and robustness, establishing a new Pareto frontier on synthetic language and arithmetic tasks. At its strongest setting, UNDO matches the robustness of a model retrained from scratch with perfect data filtering while using only 60-80% of the compute and requiring only 0.01% of the pretraining data to be labeled. We also show that UNDO robustifies unlearning on the more realistic Weapons of Mass Destruction Proxy (WMDP) benchmark. Since distillation is widely used in practice, incorporating an unlearning step beforehand offers a convenient path to robust capability removal.
Authors: Shadi Hanandeh, Omar Elbagalati, Mustafa Hajij
Abstract: To date, pavement management software products and studies on optimizing the prioritization of pavement maintenance and rehabilitation (M&R) have been mainly focused on three parameters; the pre-treatment pavement condition, the rehabilitation cost, and the available budget. Yet, the role of the candidate projects' spatial characteristics in the decision-making process has not been deeply considered. Such a limitation, predominately, allows the recommended M&R projects' schedule to involve simultaneously running but spatially scattered construction sites, which are very challenging to monitor and manage. This study introduces a novel approach to integrate pavement segments' spatial coordinates into the M&R prioritization analysis. The introduced approach aims at combining the pavement segments with converged spatial coordinates to be repaired in the same timeframe without compromising the allocated budget levels or the overall target Pavement Condition Index (PCI). Such a combination would result in minimizing the routing of crews, materials and other equipment among the construction sites and would provide better collaborations and communications between the pavement maintenance teams. Proposed herein is a novel spatial clustering algorithm that automatically finds the projects within a certain budget and spatial constrains. The developed algorithm was successfully validated using 1,800 pavement maintenance projects from two real-life examples of the City of Milton, GA and the City of Tyler, TX.
Authors: Elie Azeraf, Emmanuel Monfrini, Emmanuel Vignon, Wojciech Pieczynski
Abstract: Natural Language Processing (NLP) models' current trend consists of using increasingly more extra-data to build the best models as possible. It implies more expensive computational costs and training time, difficulties for deployment, and worries about these models' carbon footprint reveal a critical problem in the future. Against this trend, our goal is to develop NLP models requiring no extra-data and minimizing training time. To do so, in this paper, we explore Markov chain models, Hidden Markov Chain (HMC) and Pairwise Markov Chain (PMC), for NLP segmentation tasks. We apply these models for three classic applications: POS Tagging, Named-Entity-Recognition, and Chunking. We develop an original method to adapt these models for text segmentation's specific challenges to obtain relevant performances with very short training and execution times. PMC achieves equivalent results to those obtained by Conditional Random Fields (CRF), one of the most applied models for these tasks when no extra-data are used. Moreover, PMC has training times 30 times shorter than the CRF ones, which validates this model given our objectives.
Authors: Raaz Dwivedi, Katherine Tian, Sabina Tomkins, Predrag Klasnja, Susan Murphy, Devavrat Shah
Abstract: We consider after-study statistical inference for sequentially designed experiments wherein multiple units are assigned treatments for multiple time points using treatment policies that adapt over time. Our goal is to provide inference guarantees for the counterfactual mean at the smallest possible scale -- mean outcome under different treatments for each unit and each time -- with minimal assumptions on the adaptive treatment policy. Without any structural assumptions on the counterfactual means, this challenging task is infeasible due to more unknowns than observed data points. To make progress, we introduce a latent factor model over the counterfactual means that serves as a non-parametric generalization of the non-linear mixed effects model and the bilinear latent factor model considered in prior works. For estimation, we use a non-parametric method, namely a variant of nearest neighbors, and establish a non-asymptotic high probability error bound for the counterfactual mean for each unit and each time. Under regularity conditions, this bound leads to asymptotically valid confidence intervals for the counterfactual mean as the number of units and time points grows to $\infty$ together at suitable rates. We illustrate our theory via several simulations and a case study involving data from a mobile health clinical trial HeartSteps.
Authors: Haotian Lin, Matthew Reimherr
Abstract: We study the transfer learning (TL) for the functional linear regression (FLR) under the Reproducing Kernel Hilbert Space (RKHS) framework, observing that the TL techniques in existing high-dimensional linear regression are not compatible with the truncation-based FLR methods, as functional data are intrinsically infinite-dimensional and generated by smooth underlying processes. We measure the similarity across tasks using RKHS distance, allowing the type of information being transferred to be tied to the properties of the imposed RKHS. Building on the hypothesis offset transfer learning paradigm, two algorithms are proposed: one conducts the transfer when positive sources are known, while the other leverages aggregation techniques to achieve robust transfer without prior information about the sources. We establish asymptotic lower bounds for this learning problem and show that the proposed algorithms enjoy a matching upper bound. These analyses provide statistical insights into factors that contribute to the dynamics of the transfer. We also extend the results to functional generalized linear models. The effectiveness of the proposed algorithms is demonstrated via extensive synthetic data as well as real-world data applications.
Authors: Vasilis Syrgkanis, Ruohan Zhan
Abstract: We consider estimation and inference using data collected from reinforcement learning algorithms. These algorithms, characterized by their adaptive experimentation, interact with individual units over multiple stages, dynamically adjusting their strategies based on previous interactions. Our goal is to evaluate a counterfactual policy post-data collection and estimate structural parameters, like dynamic treatment effects, which can be used for credit assignment and determining the effect of earlier actions on final outcomes. Such parameters of interest can be framed as solutions to moment equations, but not minimizers of a population loss function, leading to Z-estimation approaches for static data. However, in the adaptive data collection environment of reinforcement learning, where algorithms deploy nonstationary behavior policies, standard estimators do not achieve asymptotic normality due to the fluctuating variance. We propose a weighted Z-estimation approach with carefully designed adaptive weights to stabilize the time-varying estimation variance. We identify proper weighting schemes to restore the consistency and asymptotic normality of the weighted Z-estimators for target parameters, which allows for hypothesis testing and constructing uniform confidence regions. Primary applications include dynamic treatment effect estimation and dynamic off-policy evaluation.
Authors: Md. Alamin Talukder, Md. Manowarul Islam, Md Ashraf Uddin
Abstract: Brain tumors present a grave risk to human life, demanding precise and timely diagnosis for effective treatment. Inaccurate identification of brain tumors can significantly diminish life expectancy, underscoring the critical need for precise diagnostic methods. Manual identification of brain tumors within vast Magnetic Resonance Imaging (MRI) image datasets is arduous and time-consuming. Thus, the development of a reliable deep learning (DL) model is essential to enhance diagnostic accuracy and ultimately save lives. This study introduces an innovative optimization-based deep ensemble approach employing transfer learning (TL) to efficiently classify brain tumors. Our methodology includes meticulous preprocessing, reconstruction of TL architectures, fine-tuning, and ensemble DL models utilizing weighted optimization techniques such as Genetic Algorithm-based Weight Optimization (GAWO) and Grid Search-based Weight Optimization (GSWO). Experimentation is conducted on the Figshare Contrast-Enhanced MRI (CE-MRI) brain tumor dataset, comprising 3064 images. Our approach achieves notable accuracy scores, with Xception, ResNet50V2, ResNet152V2, InceptionResNetV2, GAWO, and GSWO attaining 99.42%, 98.37%, 98.22%, 98.26%, 99.71%, and 99.76% accuracy, respectively. Notably, GSWO demonstrates superior accuracy, averaging 99.76\% accuracy across five folds on the Figshare CE-MRI brain tumor dataset. The comparative analysis highlights the significant performance enhancement of our proposed model over existing counterparts. In conclusion, our optimized deep ensemble model exhibits exceptional accuracy in swiftly classifying brain tumors. Furthermore, it has the potential to assist neurologists and clinicians in making accurate and immediate diagnostic decisions.
Authors: Tianyuan Wang, Felix Lucka, Tristan van Leeuwen
Abstract: In X-ray Computed Tomography (CT), projections from many angles are acquired and used for 3D reconstruction. To make CT suitable for in-line quality control, reducing the number of angles while maintaining reconstruction quality is necessary. Sparse-angle tomography is a popular approach for obtaining 3D reconstructions from limited data. To optimize its performance, one can adapt scan angles sequentially to select the most informative angles for each scanned object. Mathematically, this corresponds to solving an optimal experimental design (OED) problem. OED problems are high-dimensional, non-convex, bi-level optimization problems that cannot be solved online, i.e., during the scan. To address these challenges, we pose the OED problem as a partially observable Markov decision process in a Bayesian framework, and solve it through deep reinforcement learning. The approach learns efficient non-greedy policies to solve a given class of OED problems through extensive offline training rather than solving a given OED problem directly via numerical optimization. As such, the trained policy can successfully find the most informative scan angles online. We use a policy training method based on the Actor-Critic approach and evaluate its performance on 2D tomography with synthetic data.
Authors: Vik Shirvaikar, Andrea Stor{\aa}s, Xi Lin, Chris Holmes
Abstract: The identification of heterogeneous treatment effects (HTE) across subgroups is of significant interest in clinical trial analysis. Several state-of-the-art HTE estimation methods, including causal forests, apply recursive partitioning for non-parametric identification of relevant covariates and interactions. However, the partitioning criterion is typically based on differences in absolute risk. This can dilute statistical power by masking variation in the relative risk, which is often a more appropriate quantity of clinical interest. In this work, we propose and implement a methodology for modifying causal forests to target relative risk, using a novel node-splitting procedure based on exhaustive generalized linear model comparison. We present results from simulated data that suggest relative risk causal forests can capture otherwise undetected sources of heterogeneity. We implement our method on real-world trial data to explore HTEs for liraglutide in patients with type 2 diabetes.
Authors: Jessica Ojo, Odunayo Ogundepo, Akintunde Oladipo, Kelechi Ogueji, Jimmy Lin, Pontus Stenetorp, David Ifeoluwa Adelani
Abstract: Large-scale multilingual evaluations, such as MEGA, often include only a handful of African languages due to the scarcity of high-quality evaluation data and the limited discoverability of existing African datasets. This lack of representation hinders comprehensive LLM evaluation across a diverse range of languages and tasks. To address these challenges, we introduce AfroBench -- a multi-task benchmark for evaluating the performance of LLMs across 64 African languages, 15 tasks and 22 datasets. AfroBench consists of nine natural language understanding datasets, six text generation datasets, six knowledge and question answering tasks, and one mathematical reasoning task. We present results comparing the performance of prompting LLMs to fine-tuned baselines based on BERT and T5-style models. Our results suggest large gaps in performance between high-resource languages, such as English, and African languages across most tasks; but performance also varies based on the availability of monolingual data resources. Our findings confirm that performance on African languages continues to remain a hurdle for current LLMs, underscoring the need for additional efforts to close this gap. https://mcgill-nlp.github.io/AfroBench/
Authors: Md. Alamin Talukder, Md. Abu Layek, Mohsin Kazi, Md Ashraf Uddin, Sunil Aryal
Abstract: The worldwide COVID-19 pandemic has profoundly influenced the health and everyday experiences of individuals across the planet. It is a highly contagious respiratory disease requiring early and accurate detection to curb its rapid transmission. Initial testing methods primarily revolved around identifying the genetic composition of the coronavirus, exhibiting a relatively low detection rate and requiring a time-intensive procedure. To address this challenge, experts have suggested using radiological imagery, particularly chest X-rays, as a valuable approach within the diagnostic protocol. This study investigates the potential of leveraging radiographic imaging (X-rays) with deep learning algorithms to swiftly and precisely identify COVID-19 patients. The proposed approach elevates the detection accuracy by fine-tuning with appropriate layers on various established transfer learning models. The experimentation was conducted on a COVID-19 X-ray dataset containing 2000 images. The accuracy rates achieved were impressive of 100% for EfficientNetB4 model. The fine-tuned EfficientNetB4 achieved an excellent accuracy score, showcasing its potential as a robust COVID-19 detection model. Furthermore, EfficientNetB4 excelled in identifying Lung disease using Chest X-ray dataset containing 4,350 Images, achieving remarkable performance with an accuracy of 99.17%, precision of 99.13%, recall of 99.16%, and f1-score of 99.14%. These results highlight the promise of fine-tuned transfer learning for efficient lung detection through medical imaging, especially with X-ray images. This research offers radiologists an effective means of aiding rapid and precise COVID-19 diagnosis and contributes valuable assistance for healthcare professionals in accurately identifying affected patients.
Authors: Elad Aigner-Horev, Daniel Rosenberg, Roi Weiss
Abstract: The resilience of a Rademacher chaos is the maximum number of adversarial sign-flips that the chaos can sustain without having its largest atom probability significantly altered. Inspired by probabilistic lower-bound guarantees for the resilience of linear Rademacher chaos, obtained by Bandeira, Ferber, and Kwan (Advances in Mathematics, Vol. $319$, $2017$), we provide probabilistic lower-bound guarantees for the resilience of Rademacher chaos of arbitrary yet sufficiently low degree. Our main results distinguish between Rademacher chaos of order two and those of higher order. In that, our first main result pertains to the resilience of decoupled bilinear Rademacher forms where different asymptotic behaviour is observed for sparse and dense matrices. For our second main result, we bootstrap our first result in order to provide resilience guarantees for quadratic Rademacher chaos. Our third main result, generalises the first and handles the resilience of decoupled Rademacher chaos of arbitrary yet sufficiently low order. Our results for decoupled Rademacher chaos of order two and that of higher order whilst are established through the same conceptual framework, differ substantially. A difference incurred due to the implementation of the same conceptual argument. The order two result is established using Dudley's maximal inequality for sub-Gaussian processes, the Hanson-Wright inequality, as well as the Kolmogorov-Rogozin inequality. To handle higher order chaos, appeals to Dudley's inequality as well as the Hanson-Wright inequality are replaced with tools suited for random tensors. Appeals to the Hanson-Wright inequality are replaced with appeals to a concentration result for random tensors put forth by Adamczak and Wolff. Our results are instance-dependent and thus allow for the efficient computation of resilience guarantees provided the order of the chaos is constant.
Authors: Meiyi Zhu, Caili Guo, Chunyan Feng, Osvaldo Simeone
Abstract: In a membership inference attack (MIA), an attacker exploits the overconfidence exhibited by typical machine learning models to determine whether a specific data point was used to train a target model. In this paper, we analyze the performance of the likelihood ratio attack (LiRA) within an information-theoretical framework that allows the investigation of the impact of the aleatoric uncertainty in the true data generation process, of the epistemic uncertainty caused by a limited training data set, and of the calibration level of the target model. We compare three different settings, in which the attacker receives decreasingly informative feedback from the target model: confidence vector (CV) disclosure, in which the output probability vector is released; true label confidence (TLC) disclosure, in which only the probability assigned to the true label is made available by the model; and decision set (DS) disclosure, in which an adaptive prediction set is produced as in conformal prediction. We derive bounds on the advantage of an MIA adversary with the aim of offering insights into the impact of uncertainty and calibration on the effectiveness of MIAs. Simulation results demonstrate that the derived analytical bounds predict well the effectiveness of MIAs.
Authors: Jikai Jin, Vasilis Syrgkanis
Abstract: Average treatment effect estimation is the most central problem in causal inference with application to numerous disciplines. While many estimation strategies have been proposed in the literature, the statistical optimality of these methods has still remained an open area of investigation, especially in regimes where these methods do not achieve parametric rates. In this paper, we adopt the recently introduced structure-agnostic framework of statistical lower bounds, which poses no structural properties on the nuisance functions other than access to black-box estimators that achieve some statistical estimation rate. This framework is particularly appealing when one is only willing to consider estimation strategies that use non-parametric regression and classification oracles as black-box sub-processes. Within this framework, we prove the statistical optimality of the celebrated and widely used doubly robust estimators for both the Average Treatment Effect (ATE) and the Average Treatment Effect on the Treated (ATT), as well as weighted variants of the former, which arise in policy evaluation.
Authors: Jingtao Ding, Chang Liu, Yu Zheng, Yunke Zhang, Zihan Yu, Ruikun Li, Hongyi Chen, Jinghua Piao, Huandong Wang, Jiazhen Liu, Yong Li
Abstract: Complex networks pervade various real-world systems, from the natural environment to human societies. The essence of these networks is in their ability to transition and evolve from microscopic disorder-where network topology and node dynamics intertwine-to a macroscopic order characterized by certain collective behaviors. Over the past two decades, complex network science has significantly enhanced our understanding of the statistical mechanics, structures, and dynamics underlying real-world networks. Despite these advancements, there remain considerable challenges in exploring more realistic systems and enhancing practical applications. The emergence of artificial intelligence (AI) technologies, coupled with the abundance of diverse real-world network data, has heralded a new era in complex network science research. This survey aims to systematically address the potential advantages of AI in overcoming the lingering challenges of complex network research. It endeavors to summarize the pivotal research problems and provide an exhaustive review of the corresponding methodologies and applications. Through this comprehensive survey-the first of its kind on AI for complex networks-we expect to provide valuable insights that will drive further research and advancement in this interdisciplinary field.
Authors: Junhyeong Cho, Kim Youwang, Hunmin Yang, Tae-Hyun Oh
Abstract: Recent monocular 3D shape reconstruction methods have shown promising zero-shot results on object-segmented images without any occlusions. However, their effectiveness is significantly compromised in real-world conditions, due to imperfect object segmentation by off-the-shelf models and the prevalence of occlusions. To effectively address these issues, we propose a unified regression model that integrates segmentation and reconstruction, specifically designed for occlusion-aware 3D shape reconstruction. To facilitate its reconstruction in the wild, we also introduce a scalable data synthesis pipeline that simulates a wide range of variations in objects, occluders, and backgrounds. Training on our synthetic data enables the proposed model to achieve state-of-the-art zero-shot results on real-world images, using significantly fewer parameters than competing approaches.
Authors: Atsuyuki Miyai, Jingkang Yang, Jingyang Zhang, Yifei Ming, Qing Yu, Go Irie, Yixuan Li, Hai Li, Ziwei Liu, Kiyoharu Aizawa
Abstract: This paper introduces a novel task to evaluate the robust understanding capability of Large Multimodal Models (LMMs), termed $\textbf{Unsolvable Problem Detection (UPD)}$. Multiple-choice question answering (MCQA) is widely used to assess the understanding capability of LMMs, but it does not guarantee that LMMs truly comprehend the answer. UPD assesses the LMM's ability to withhold answers when encountering unsolvable problems of MCQA, verifying whether the model truly understands the answer. UPD encompasses three problems: Absent Answer Detection (AAD), Incompatible Answer Set Detection (IASD), and Incompatible Visual Question Detection (IVQD), covering unsolvable cases like answer-lacking or incompatible choices and image-question mismatches. For the evaluation, we introduce the MM-UPD Bench, a benchmark for assessing performance across various ability dimensions. Our experiments reveal that even most LMMs, which demonstrate adequate performance on existing benchmarks, struggle significantly with MM-UPD, underscoring a novel aspect of trustworthiness that current benchmarks have overlooked. A detailed analysis shows that LMMs have different bottlenecks and chain-of-thought and self-reflection improved performance for LMMs with the bottleneck in their LLM capability. We hope our insights will enhance the broader understanding and development of more reliable LMMs. The code is available at https://github.com/AtsuMiyai/UPD.
Authors: Jerry Yao-Chieh Hu, Bo-Yu Chen, Dennis Wu, Feng Ruan, Han Liu
Abstract: We present a nonparametric interpretation for deep learning compatible modern Hopfield models and utilize this new perspective to debut efficient variants. Our key contribution stems from interpreting the memory storage and retrieval processes in modern Hopfield models as a nonparametric regression problem subject to a set of query-memory pairs. Interestingly, our framework not only recovers the known results from the original dense modern Hopfield model but also fills the void in the literature regarding efficient modern Hopfield models, by introducing \textit{sparse-structured} modern Hopfield models with sub-quadratic complexity. We establish that this sparse model inherits the appealing theoretical properties of its dense analogue -- connection with transformer attention, fixed point convergence and exponential memory capacity. Additionally, we showcase the versatility of our framework by constructing a family of modern Hopfield models as extensions, including linear, random masked, top-$K$ and positive random feature modern Hopfield models. Empirically, we validate our framework in both synthetic and realistic settings for memory retrieval and learning tasks.
Authors: Yuheng Lai, Leying Guan
Abstract: $\textit{Equalized odds}$, an important notion of algorithmic fairness, aims to ensure that sensitive variables, such as race and gender, do not unfairly influence the algorithm's prediction when conditioning on the true outcome. Despite rapid advancements, current research primarily focuses on equalized odds violations caused by a single sensitive attribute, leaving the challenge of simultaneously accounting for multiple attributes under-addressed. We bridge this gap by introducing an in-processing fairness-aware learning approach, FairICP, which integrates adversarial learning with a novel inverse conditional permutation scheme. FairICP offers a flexible and efficient scheme to promote equalized odds under fairness conditions described by complex and multi-dimensional sensitive attributes. The efficacy and adaptability of our method are demonstrated through both simulation studies and empirical analyses of real-world datasets.
Authors: Sunny Sanyal, Ravid Shwartz-Ziv, Alexandros G. Dimakis, Sujay Sanghavi
Abstract: Large Language Models (LLMs) rely on the transformer architecture and its self-attention mechanism to deliver strong performance across tasks. However, we uncover a structural inefficiency in standard pre-trained decoder-style LLMs: in many of the deeper layers, attention matrices frequently collapse to near rank-one, single-column patterns. We refer to these underutilized components as lazy layers, which are redundant and computationally inefficient. To address this, we propose Inheritune, a simple and effective training recipe for building smaller, more efficient, and high performing language models. Inheritune initializes a compact model by inheriting the useful early layers from a larger pre-trained model, then progressively retrains and expands it. Our experiments across multiple models and datasets show that Inheritune trained models, despite having significantly fewer layers, can match or even outperform their larger counterparts. This approach yields compact, performant models and offers a practical path for efficient language model compression. Code is available at https://github.com/sanyalsunny111/LLM-Inheritune
Authors: Grigory Bartosh, Dmitry Vetrov, Christian A. Naesseth
Abstract: Conventional diffusion models typically relies on a fixed forward process, which implicitly defines complex marginal distributions over latent variables. This can often complicate the reverse process' task in learning generative trajectories, and results in costly inference for diffusion models. To address these limitations, we introduce Neural Flow Diffusion Models (NFDM), a novel framework that enhances diffusion models by supporting a broader range of forward processes beyond the standard Gaussian. We also propose a novel parameterization technique for learning the forward process. Our framework provides an end-to-end, simulation-free optimization objective, effectively minimizing a variational upper bound on the negative log-likelihood. Experimental results demonstrate NFDM's strong performance, evidenced by state-of-the-art likelihood estimation. Furthermore, we investigate NFDM's capacity for learning generative dynamics with specific characteristics, such as deterministic straight lines trajectories, and demonstrate how the framework may be adopted for learning bridges between two distributions. The results underscores NFDM's versatility and its potential for a wide range of applications.
Authors: Nishant Luitel, Nirajan Bekoju, Anand Kumar Sah, Subarna Shakya
Abstract: The impact of subword tokenization on language model performance is well-documented for perplexity, with finer granularity consistently reducing this intrinsic metric. However, research on how different tokenization schemes affect a model's understanding capabilities remains limited, particularly for non-Latin script languages. Addressing this gap, we conducted a comprehensive evaluation of six distinct tokenization strategies by pretraining transformer-based language models for Nepali and evaluating their performance across multiple downstream tasks. While recent prominent models like GPT, RoBERTa, Claude, LLaMA, Mistral, Falcon, and MPT have adopted byte-level BPE tokenization, our findings demonstrate that for Nepali, SentencePiece tokenization consistently yields superior results on understanding-based tasks. Unlike previous studies that primarily focused on BERT-based architectures, our research specifically examines sequential transformer models, providing valuable insights for language model development in low-resource languages and highlighting the importance of tokenization strategy beyond perplexity reduction.
Authors: Jingwei Li, Jing Dong, Tianxing He, Jingzhao Zhang
Abstract: Given the rising popularity of AI-generated art and the associated copyright concerns, identifying whether an artwork was used to train a diffusion model is an important research topic. The work approaches this problem from the membership inference attack (MIA) perspective. We first identify the limitation of applying existing MIA methods for proprietary diffusion models: the required access of internal U-nets. To address the above problem, we introduce a novel membership inference attack method that uses only the image-to-image variation API and operates without access to the model's internal U-net. Our method is based on the intuition that the model can more easily obtain an unbiased noise prediction estimate for images from the training set. By applying the API multiple times to the target image, averaging the outputs, and comparing the result to the original image, our approach can classify whether a sample was part of the training set. We validate our method using DDIM and Stable Diffusion setups and further extend both our approach and existing algorithms to the Diffusion Transformer architecture. Our experimental results consistently outperform previous methods.
Authors: Patrick Chao, Yan Sun, Edgar Dobriban, Hamed Hassani
Abstract: Recent progress in large language models enables the creation of realistic machine-generated content. Watermarking is a promising approach to distinguish machine-generated text from human text, embedding statistical signals in the output that are ideally undetectable to humans. We propose a watermarking framework that encodes such signals through an error correcting code. Our method, termed robust binary code (RBC) watermark, introduces no noticeable degradation in quality. We evaluate our watermark on base and instruction fine-tuned models and find that our watermark is robust to edits, deletions, and translations. We provide an information-theoretic perspective on watermarking, a powerful statistical test for detection and for generating $p$-values, and theoretical guarantees. Our empirical findings suggest our watermark is fast, powerful, and robust, comparing favorably to the state-of-the-art.
Authors: Saeed Rashidi, William Won, Sudarshan Srinivasan, Puneet Gupta, Tushar Krishna
Abstract: Distributed Deep Neural Network (DNN) training is a technique to reduce the training overhead by distributing the training tasks into multiple accelerators, according to a parallelization strategy. However, high-performance compute and interconnects are needed for maximum speed-up and linear scaling of the system. Wafer-scale systems are a promising technology that allows for tightly integrating high-end accelerators with high-speed wafer-scale interconnects, making it an attractive platform for distributed training. However, the wafer-scale interconnect should offer high performance and flexibility for various parallelization strategies to enable maximum optimizations for compute and memory usage. In this paper, we propose FRED, a wafer-scale interconnect that is tailored for the high-BW requirements of wafer-scale networks and can efficiently execute communication patterns of different parallelization strategies. Furthermore, FRED supports in-switch collective communication execution that reduces the network traffic by approximately 2X. Our results show that FRED can improve the average end-to-end training time of ResNet-152, Transformer-17B, GPT-3, and Transformer-1T by 1.76X, 1.87X, 1.34X, and 1.4X, respectively when compared to a baseline waferscale 2D-Mesh fabric.
Authors: Xiaokun Luan, Xiyue Zhang, Jingyi Wang, Meng Sun
Abstract: Model reuse techniques can reduce the resource requirements for training high-performance deep neural networks (DNNs) by leveraging existing models. However, unauthorized reuse and replication of DNNs can lead to copyright infringement and economic loss to the model owner. This underscores the need to analyze the reuse relation between DNNs and develop copyright protection techniques to safeguard intellectual property rights. Existing white-box testing-based approaches cannot address the common heterogeneous reuse case where the model architecture is changed, and DNN fingerprinting approaches heavily rely on generating adversarial examples with good transferability, which is known to be challenging in the black-box setting. To bridge the gap, we propose NFARD, a Neuron Functionality Analysis-based Reuse Detector, which only requires normal test samples to detect reuse relations by measuring the models' differences on a newly proposed model characterization, i.e., neuron functionality (NF). A set of NF-based distance metrics is designed to make NFARD applicable to both white-box and black-box settings. Moreover, we devise a linear transformation method to handle heterogeneous reuse cases by constructing the optimal projection matrix for dimension consistency, significantly extending the application scope of NFARD. To the best of our knowledge, this is the first adversarial example-free method that exploits neuron functionality for DNN copyright protection. As a side contribution, we constructed a reuse detection benchmark named Reuse Zoo that covers various practical reuse techniques and popular datasets. Extensive evaluations on this comprehensive benchmark show that NFARD achieves F1 scores of 0.984 and 1.0 for detecting reuse relationships in black-box and white-box settings, respectively, while generating test suites 2 ~ 99 times faster than previous methods.
Authors: Zeliang Zhang, Xiaodong Liu, Hao Cheng, Chenliang Xu, Jianfeng Gao
Abstract: By increasing model parameters but activating them sparsely when performing a task, the use of Mixture-of-Experts (MoE) architecture significantly improves the performance of Large Language Models (LLMs) without increasing the inference cost. However, the memory consumption due to the growing number of experts presents a challenge to the deployment of these models in many real world settings. Our empirical study reveals that some experts encode redundant knowledge during pre-training. We thus propose a method of grouping and pruning similar experts to improve the model's parameter efficiency. We validate the effectiveness of our method by pruning three state-of-the-art MoE architectures, including Mixtral, Deepseek-MoE, and Qwen. The evaluation shows that our method outperforms other model pruning methods on a range of natural language tasks. We will release our code to facilitate future research.
Authors: Abhi Kamboj, Anh Duy Nguyen, Minh N. Do
Abstract: In order to unlock the potential of diverse sensors, we investigate a method to transfer knowledge between time-series modalities using a multimodal \textit{temporal} representation space for Human Activity Recognition (HAR). Specifically, we explore the setting where the modality used in testing has no labeled data during training, which we refer to as Unsupervised Modality Adaptation (UMA). We categorize existing UMA approaches as Student-Teacher or Contrastive Alignment methods. These methods typically compress continuous-time data samples into single latent vectors during alignment, inhibiting their ability to transfer temporal information through real-world temporal distortions. To address this, we introduce Cross-modal Transfer Through Time (C3T), which preserves temporal information during alignment to handle dynamic sensor data better. C3T achieves this by aligning a set of temporal latent vectors across sensing modalities. Our extensive experiments on various camera+IMU datasets demonstrate that C3T outperforms existing methods in UMA by at least 8% in accuracy and shows superior robustness to temporal distortions such as time-shift, misalignment, and dilation. Our findings suggest that C3T has significant potential for developing generalizable models for time-series sensor data, opening new avenues for various multimodal applications.
Authors: Krishnakant Singh, Simone Schaub-Meyer, Stefan Roth
Abstract: Object-centric learning aims to decompose an input image into a set of meaningful object files (slots). These latent object representations enable a variety of downstream tasks. Yet, object-centric learning struggles on real-world datasets, which contain multiple objects of complex textures and shapes in natural everyday scenes. To address this, we introduce Guided Latent Slot Diffusion (GLASS), a novel slot-attention model that learns in the space of generated images and uses semantic and instance guidance modules to learn better slot embeddings for various downstream tasks. Our experiments show that GLASS surpasses state-of-the-art slot-attention methods by a wide margin on tasks such as (zero-shot) object discovery and conditional image generation for real-world scenes. Moreover, GLASS enables the first application of slot attention to the compositional generation of complex, realistic scenes.
Authors: Cheng Shi, Liming Pan, Ivan Dokmani\'c
Abstract: Feature-learning deep nets progressively collapse data to a regular low-dimensional geometry. How this emerges from the collective action of nonlinearity, noise, learning rate, and other factors, has eluded first-principles theories built from microscopic neuronal dynamics. We exhibit a noise-nonlinearity phase diagram that identifies regimes where shallow or deep layers learn more effectively and propose a macroscopic mechanical theory that reproduces the diagram and links feature learning across layers to generalization.
Authors: Cyril Shih-Huan Hsu, Danny De Vleeschauwer, Chrysa Papagianni, Paola Grosso
Abstract: When a network slice spans multiple technology domains, it is crucial for each domain to uphold the End-to-End (E2E) Service Level Agreement (SLA) associated with the slice. Consequently, the E2E SLA must be properly decomposed into partial SLAs that are assigned to each domain involved. In a network slice management system with a two-level architecture, comprising an E2E service orchestrator and local domain controllers, we consider that the orchestrator has access only to historical data regarding the responses of local controllers to previous requests, and this information is used to construct a risk model for each domain. In this study, we extend our previous work by investigating the dynamic nature of real-world systems and introducing an online learning-decomposition framework to tackle the dynamicity. We propose a framework that continuously updates the risk models based on the most recent feedback. This approach leverages key components such as online gradient descent and FIFO memory buffers, which enhance the stability and robustness of the overall process. Our empirical study on an analytic model-based simulator demonstrates that the proposed framework outperforms the state-of-the-art static approach, delivering more accurate and resilient SLA decomposition under varying conditions and data limitations. Furthermore, we provide a comprehensive complexity analysis of the proposed solution.
Authors: Jingying Ma, Qika Lin, Ziyu Jia, Mengling Feng
Abstract: Sleep staging is critical to assess sleep quality and diagnose disorders. Despite advancements in artificial intelligence enabling automated sleep staging, significant challenges remain: (1) Simultaneously extracting prominent temporal and spatial sleep features from multi-channel raw signals, including characteristic sleep waveforms and salient spatial brain networks. (2) Capturing the spatial-temporal coupling patterns essential for accurate sleep staging. To address these challenges, we propose a novel framework named ST-USleepNet, comprising a spatial-temporal graph construction module (ST) and a U-shaped sleep network (USleepNet). The ST module converts raw signals into a spatial-temporal graph based on signal similarity, temporal, and spatial relationships to model spatial-temporal coupling patterns. The USleepNet employs a U-shaped structure for both the temporal and spatial streams, mirroring its original use in image segmentation to isolate significant targets. Applied to raw sleep signals and graph data from the ST module, USleepNet effectively segments these inputs, simultaneously extracting prominent temporal and spatial sleep features. Testing on three datasets demonstrates that ST-USleepNet outperforms existing baselines, and model visualizations confirm its efficacy in extracting prominent sleep features and temporal-spatial coupling patterns across various sleep stages. The code is available at https://github.com/Majy-Yuji/ST-USleepNet.
Authors: Cheng Qian, Hainan Zhang, Lei Sha, Zhiming Zheng
Abstract: With the growing deployment of LLMs in daily applications like chatbots and content generation, efforts to ensure outputs align with human values and avoid harmful content have intensified. However, increasingly sophisticated jailbreak attacks threaten this alignment, aiming to induce unsafe outputs. Current defense efforts either focus on prompt rewriting or detection, which are limited in effectiveness due to the various design of jailbreak prompts, or on output control and detection, which are computationally expensive as they require LLM inference. Therefore, designing a pre-inference defense method that resists diverse jailbreak prompts is crucial for preventing LLM jailbreak attacks. We observe that jailbreak attacks, safe queries, and harmful queries exhibit different clustering patterns within the LLM's hidden state representation space. This suggests that by leveraging the LLM's hidden state representational capabilities, we can analyze the LLM's forthcoming behavior and proactively intervene for defense. In this paper, we propose a jailbreak attack defense strategy based on a Hidden State Filter (HSF), a lossless architectural defense mechanism that enables the model to preemptively identify and reject adversarial inputs before the inference process begins. We activate its defensive potential through an additional plugin module, effectively framing the defense task as a classification problem. Experimental results on two benchmark datasets, utilizing three different LLMs, show that HSF significantly enhances resilience against six cutting-edge jailbreak attacks. It significantly reduces the success rate of jailbreak attacks while minimally impacting responses to benign user queries, with negligible inference overhead, and outperforming defense baselines.Our code and data are available at https://anonymous.4open.science/r/Hidden-State-Filtering-8652/
URLs: https://anonymous.4open.science/r/Hidden-State-Filtering-8652/
Authors: Zhiyang Wang, Juan Cervino, Alejandro Ribeiro
Abstract: In this paper, we study the generalization capabilities of geometric graph neural networks (GNNs). We consider GNNs over a geometric graph constructed from a finite set of randomly sampled points over an embedded manifold with topological information captured. We prove a generalization gap between the optimal empirical risk and the optimal statistical risk of this GNN, which decreases with the number of sampled points from the manifold and increases with the dimension of the underlying manifold. This generalization gap ensures that the GNN trained on a graph on a set of sampled points can be utilized to process other unseen graphs constructed from the same underlying manifold. The most important observation is that the generalization capability can be realized with one large graph instead of being limited to the size of the graph as in previous results. The generalization gap is derived based on the non-asymptotic convergence result of a GNN on the sampled graph to the underlying manifold neural networks (MNNs). We verify this theoretical result with experiments on multiple real-world datasets.
Authors: Sayan Banerjee, Krishnakumar Balasubramanian, Promit Ghosal
Abstract: We provide finite-particle convergence rates for the Stein Variational Gradient Descent (SVGD) algorithm in the Kernelized Stein Discrepancy ($\mathsf{KSD}$) and Wasserstein-2 metrics. Our key insight is that the time derivative of the relative entropy between the joint density of $N$ particle locations and the $N$-fold product target measure, starting from a regular initial distribution, splits into a dominant `negative part' proportional to $N$ times the expected $\mathsf{KSD}^2$ and a smaller `positive part'. This observation leads to $\mathsf{KSD}$ rates of order $1/\sqrt{N}$, in both continuous and discrete time, providing a near optimal (in the sense of matching the corresponding i.i.d. rates) double exponential improvement over the recent result by Shi and Mackey (2024). Under mild assumptions on the kernel and potential, these bounds also grow polynomially in the dimension $d$. By adding a bilinear component to the kernel, the above approach is used to further obtain Wasserstein-2 convergence in continuous time. For the case of `bilinear + Mat\'ern' kernels, we derive Wasserstein-2 rates that exhibit a curse-of-dimensionality similar to the i.i.d. setting. We also obtain marginal convergence and long-time propagation of chaos results for the time-averaged particle laws.
Authors: Tetiana Parshakova, Trevor Hastie, Stephen Boyd
Abstract: We examine a special case of the multilevel factor model, with covariance given by multilevel low rank (MLR) matrix~\cite{parshakova2023factor}. We develop a novel, fast implementation of the expectation-maximization algorithm, tailored for multilevel factor models, to maximize the likelihood of the observed data. This method accommodates any hierarchical structure and maintains linear time and storage complexities per iteration. This is achieved through a new efficient technique for computing the inverse of the positive definite MLR matrix. We show that the inverse of positive definite MLR matrix is also an MLR matrix with the same sparsity in factors, and we use the recursive Sherman-Morrison-Woodbury matrix identity to obtain the factors of the inverse. Additionally, we present an algorithm that computes the Cholesky factorization of an expanded matrix with linear time and space complexities, yielding the covariance matrix as its Schur complement. This paper is accompanied by an open-source package that implements the proposed methods.
Authors: Zeyu Zhang, Haiying Shen
Abstract: The scaling of transformer-based Large Language Models (LLMs) has significantly expanded their context lengths, enabling applications where inputs exceed 100K tokens. Our analysis of a recent Azure LLM inference trace reveals a highly skewed long-tail distribution of input lengths, with approximately 80% of inputs shorter than 2K tokens. Long inputs constitute only a small fraction. Existing cluster-level LLM scheduling strategies, including First-In-First-Out (FIFO), reservation-based, and priority-based approaches, primarily target short-input requests with lengths below 2K and fail to address this heterogeneity, leading to inefficiencies such as head-of-line blocking, resource underutilization, and starvation of long-input requests. We propose PecSched, a Preemptive and Efficient Cluster SCHEDuling system for LLM inference. PecSched introduces the following key techniques: 1) preemptive scheduling that prioritizes short-input requests for their performance; 2) coordinated prefill-decode colocation and disaggregation, which reduces both the duration and frequency of preemptions; 3) fast Sequence Parallelism (SP) that minimizes the prefill time of long-input requests to further reduce the likelihood and frequency of preemptions. Evaluations based on Azure LLM inference trace show that, compared to state-of-the-art cluster-level LLM inference schedulers, PecSched reduces the 99th percentile queueing delay of short-input requests by up to 92% and improves their throughput by up to 595%, without significantly affecting the Job Completion Time (JCT) of long-input requests. We open-sourced our code.
Authors: Yifan Jiang, Kriti Aggarwal, Tanmay Laud, Kashif Munir, Jay Pujara, Subhabrata Mukherjee
Abstract: The rapid progress of Large Language Models (LLMs) has opened up new opportunities across various domains and applications; yet it also presents challenges related to potential misuse. To mitigate such risks, red teaming has been employed as a proactive security measure to probe language models for harmful outputs via jailbreak attacks. However, current jailbreak attack approaches are single-turn with explicit malicious queries that do not fully capture the complexity of real-world interactions. In reality, users can engage in multi-turn interactions with LLM-based chat assistants, allowing them to conceal their true intentions in a more covert manner. To bridge this gap, we, first, propose a new jailbreak approach, RED QUEEN ATTACK. This method constructs a multi-turn scenario, concealing the malicious intent under the guise of preventing harm. We craft 40 scenarios that vary in turns and select 14 harmful categories to generate 56k multi-turn attack data points. We conduct comprehensive experiments on the RED QUEEN ATTACK with four representative LLM families of different sizes. Our experiments reveal that all LLMs are vulnerable to RED QUEEN ATTACK, reaching 87.62% attack success rate on GPT-4o and 75.4% on Llama3-70B. Further analysis reveals that larger models are more susceptible to the RED QUEEN ATTACK, with multi-turn structures and concealment strategies contributing to its success. To prioritize safety, we introduce a straightforward mitigation strategy called RED QUEEN GUARD, which aligns LLMs to effectively counter adversarial attacks. This approach reduces the attack success rate to below 1% while maintaining the model's performance across standard benchmarks. Full implementation and dataset are publicly accessible at https://github.com/kriti-hippo/red_queen.
Authors: Gholamali Aminian, Amir R. Asadi, Tian Li, Ahmad Beirami, Gesine Reinert, Samuel N. Cohen
Abstract: The generalization error (risk) of a supervised statistical learning algorithm quantifies its prediction ability on previously unseen data. Inspired by exponential tilting, \citet{li2020tilted} proposed the {\it tilted empirical risk} (TER) as a non-linear risk metric for machine learning applications such as classification and regression problems. In this work, we examine the generalization error of the tilted empirical risk in the robustness regime under \textit{negative tilt}. Our first contribution is to provide uniform and information-theoretic bounds on the {\it tilted generalization error}, defined as the difference between the population risk and the tilted empirical risk, under negative tilt for unbounded loss function under bounded $(1+\epsilon)$-th moment of loss function for some $\epsilon\in(0,1]$ with a convergence rate of $O(n^{-\epsilon/(1+\epsilon)})$ where $n$ is the number of training samples, revealing a novel application for TER under no distribution shift. Secondly, we study the robustness of the tilted empirical risk with respect to noisy outliers at training time and provide theoretical guarantees under distribution shift for the tilted empirical risk. We empirically corroborate our findings in simple experimental setups where we evaluate our bounds to select the value of tilt in a data-driven manner.
Authors: Alan Papalia, Yulun Tian, David M. Rosen, Jonathan P. How, John J. Leonard
Abstract: This paper presents an overview of the Burer-Monteiro method (BM), a technique that has been applied to solve robot perception problems to certifiable optimality in real-time. BM is often used to solve semidefinite programming relaxations, which can be used to perform global optimization for non-convex perception problems. Specifically, BM leverages the low-rank structure of typical semidefinite programs to dramatically reduce the computational cost of performing optimization. This paper discusses BM in certifiable perception, with three main objectives: (i) to consolidate information from the literature into a unified presentation, (ii) to elucidate the role of the linear independence constraint qualification (LICQ), a concept not yet well-covered in certifiable perception literature, and (iii) to share practical considerations that are discussed among practitioners but not thoroughly covered in the literature. Our general aim is to offer a practical primer for applying BM towards certifiable perception.
Authors: Wanchao Liang, Tianyu Liu, Less Wright, Will Constable, Andrew Gu, Chien-Chin Huang, Iris Zhang, Wei Feng, Howard Huang, Junjie Wang, Sanket Purandare, Gokul Nadathur, Stratos Idreos
Abstract: The development of large language models (LLMs) has been instrumental in advancing state-of-the-art natural language processing applications. Training LLMs with billions of parameters and trillions of tokens require sophisticated distributed systems that enable composing and comparing several state-of-the-art techniques in order to efficiently scale across thousands of accelerators. However, existing solutions are complex, scattered across multiple libraries/repositories, lack interoperability, and are cumbersome to maintain. Thus, curating and empirically comparing training recipes require non-trivial engineering effort. This paper introduces TorchTitan, an open-source, PyTorch-native distributed training system that unifies state-of-the-art techniques, streamlining integration and reducing overhead. TorchTitan enables 3D parallelism in a modular manner with elastic scaling, providing comprehensive logging, checkpointing, and debugging tools for production-ready training. It also incorporates hardware-software co-designed solutions, leveraging features like Float8 training and SymmetricMemory. As a flexible test bed, TorchTitan facilitates custom recipe curation and comparison, allowing us to develop optimized training recipes for Llama 3.1 and provide guidance on selecting techniques for maximum efficiency based on our experiences. We thoroughly assess TorchTitan on the Llama 3.1 family of LLMs, spanning 8 billion to 405 billion parameters, and showcase its exceptional performance, modular composability, and elastic scalability. By stacking training optimizations, we demonstrate accelerations of 65.08% with 1D parallelism at the 128-GPU scale (Llama 3.1 8B), an additional 12.59% with 2D parallelism at the 256-GPU scale (Llama 3.1 70B), and an additional 30% with 3D parallelism at the 512-GPU scale (Llama 3.1 405B) on NVIDIA H100 GPUs over optimized baselines.
Authors: Lorenzo Magnino, Yuchen Zhu, Mathieu Lauri\`ere
Abstract: Optimal stopping is a fundamental problem in optimization with applications in risk management, finance, robotics, and machine learning. We extend the standard framework to a multi-agent setting, named multi-agent optimal stopping (MAOS), where agents cooperate to make optimal stopping decisions in a finite-space, discrete-time environment. Since solving MAOS becomes computationally prohibitive as the number of agents is very large, we study the mean-field optimal stopping (MFOS) problem, obtained as the number of agents tends to infinity. We establish that MFOS provides a good approximation to MAOS and prove a dynamic programming principle (DPP) based on mean-field control theory. We then propose two deep learning approaches: one that learns optimal stopping decisions by simulating full trajectories and another that leverages the DPP to compute the value function and to learn the optimal stopping rule using backward induction. Both methods train neural networks to approximate optimal stopping policies. We demonstrate the effectiveness and the scalability of our work through numerical experiments on 6 different problems in spatial dimension up to 300. To the best of our knowledge, this is the first work to formalize and computationally solve MFOS in discrete time and finite space, opening new directions for scalable MAOS methods.
Authors: Fangru Lin, Shaoguang Mao, Emanuele La Malfa, Valentin Hofmann, Adrian de Wynter, Xun Wang, Si-Qing Chen, Michael Wooldridge, Janet B. Pierrehumbert, Furu Wei
Abstract: Language is not monolithic. While benchmarks, including those designed for multiple languages, are often used as proxies to evaluate the performance of Large Language Models (LLMs), they tend to overlook the nuances of within-language variation and thus fail to model the experience of speakers of non-standard dialects. Focusing on African American Vernacular English (AAVE), we present the first study aimed at objectively assessing the fairness and robustness of LLMs in handling dialects across canonical reasoning tasks, including algorithm, math, logic, and integrated reasoning. We introduce ReDial (Reasoning with Dialect Queries), a benchmark containing 1.2K+ parallel query pairs in Standardized English and AAVE. We hire AAVE speakers, including experts with computer science backgrounds, to rewrite seven popular benchmarks, such as HumanEval and GSM8K. With ReDial, we evaluate widely used LLMs, including GPT, Claude, Llama, Mistral, and the Phi model families. Our findings reveal that almost all of these widely used models show significant brittleness and unfairness to queries in AAVE. Our work establishes a systematic and objective framework for analyzing LLM bias in dialectal queries. Moreover, it highlights how mainstream LLMs provide unfair service to dialect speakers in reasoning tasks, laying a critical foundation for future research.
Authors: Weishan Cai, Wenjun Ma, Yuncheng Jiang
Abstract: The success of current Entity Alignment (EA) task depends largely on the supervision information provided by labeled data. Considering the cost of labeled data, most supervised methods are difficult to apply in practical scenarios. Therefore, more and more works based on contrastive learning, active learning or other deep learning techniques have been developed, to solve the performance bottleneck caused by the lack of labeled data. However, the existing unsupervised EA methods still have some limitations, either their modeling complexity is high or they cannot balance the effectiveness and practicality of alignment. To overcome these issues, we propose a Simplifying and Learnable graph convolutional attention network for Unsupervised Knowledge Graphs alignment method (SLU). Specifically, we first introduce LCAT, a new and simple framework as the backbone network to model the graph structure of two KGs. Then we design a reconstruction method of relation structure based on potential matching relations for efficiently filtering invalid neighborhood information of aligned entities, to improve the usability and scalability of SLU. Impressively, a similarity function based on consistency is proposed to better measure the similarity of candidate entity pairs. Finally, we conduct extensive experiments on three datasets of different sizes (15K and 100K) and different types (cross-lingual and monolingual) to verify the superiority of SLU. Experimental results show that SLU significantly improves alignment accuracy, outperforming 25 supervised or unsupervised methods, and improving 6.4% in Hits@1 over the best baseline in the best case.
Authors: Yujun Zhou, Jingdong Yang, Yue Huang, Kehan Guo, Zoe Emory, Bikram Ghosh, Amita Bedar, Sujay Shekar, Zhenwen Liang, Pin-Yu Chen, Tian Gao, Werner Geyer, Nuno Moniz, Nitesh V Chawla, Xiangliang Zhang
Abstract: Artificial Intelligence (AI) is revolutionizing scientific research, yet its growing integration into laboratory environments presents critical safety challenges. While large language models (LLMs) increasingly assist in tasks ranging from procedural guidance to autonomous experiment orchestration, an "illusion of understanding" may lead researchers to overestimate their reliability. Such overreliance is particularly dangerous in high-stakes laboratory settings, where failures in hazard identification or risk assessment can result in severe accidents. To address these concerns, we propose the Laboratory Safety Benchmark (LabSafety Bench), a comprehensive framework that evaluates large language models and vision language models (VLMs) on their ability to identify potential hazards, assess risks, and predict the consequences of unsafe actions in lab environments. LabSafety Bench comprises 765 multiple-choice questions aligned with US Occupational Safety and Health Administration (OSHA) protocols, along with 404 realistic laboratory scenarios featuring dual evaluation tasks: the Hazards Identification Test and the Consequence Identification Test, with 3128 open-ended questions in total. Evaluations across eight proprietary models, seven open-weight LLMs, and four VLMs reveal that, despite advanced performance on structured assessments, no model achieves the safety threshold required for reliable operation -- none scoring above 70% on the Hazards Identification Test. Moreover, while proprietary models tend to excel in multiple-choice evaluations, their performance in open-ended, real-world scenario responses is comparable to that of open-source models. These findings underscore the urgent need for specialized evaluation frameworks to ensure the safe and responsible deployment of AI in laboratory settings.
Authors: Hrishav Bakul Barua, Kalin Stefanov, Lemuel Lai En Che, Abhinav Dhall, KokSheik Wong, Ganesh Krishnasamy
Abstract: The translation of Low Dynamic Range (LDR) to High Dynamic Range (HDR) images is an important computer vision task. There is a significant amount of research utilizing both conventional non-learning methods and modern data-driven approaches, focusing on using both single-exposed and multi-exposed LDR for HDR image reconstruction. However, most current state-of-the-art methods require high-quality paired {LDR,HDR} datasets for model training. In addition, there is limited literature on using unpaired datasets for this task, that is, the model learns a mapping between domains, i.e., {LDR,HDR}. This paper proposes LLM-HDR, a method that integrates the perception of Large Language Models (LLM) into a modified semantic- and cycle-consistent adversarial architecture that utilizes unpaired {LDR,HDR} datasets for training. The method introduces novel artifact- and exposure-aware generators to address visual artifact removal and an encoder and loss to address semantic consistency, another under-explored topic. LLM-HDR is the first to use an LLM for the {LDR,HDR} translation task in a self-supervised setup. The method achieves state-of-the-art performance across several benchmark datasets and reconstructs high-quality HDR images. The official website of this work is available at: https://github.com/HrishavBakulBarua/LLM-HDR
Authors: Niclas Dern, John P. Cunningham, Geoff Pleiss
Abstract: Classic ensembles generalize better than any single component model. In contrast, recent empirical studies find that modern ensembles of (overparameterized) neural networks may not provide any inherent generalization advantage over single but larger neural networks. This paper clarifies how modern overparameterized ensembles differ from their classic underparameterized counterparts, using ensembles of random feature (RF) regressors as a basis for developing theory. In contrast to the underparameterized regime, where ensembling typically induces regularization and increases generalization, we prove with minimal assumptions that infinite ensembles of overparameterized RF regressors become pointwise equivalent to (single) infinite-width RF regressors, and finite width ensembles rapidly converge to single models with the same parameter budget. These results, which are exact for ridgeless models and approximate for small ridge penalties, imply that overparameterized ensembles and single large models exhibit nearly identical generalization. We further characterize the predictive variance amongst ensemble members, demonstrating that it quantifies the expected effects of increasing capacity rather than capturing any conventional notion of uncertainty. Our results challenge common assumptions about the advantages of ensembles in overparameterized settings, prompting a reconsideration of how well intuitions from underparameterized ensembles transfer to deep ensembles and the overparameterized regime.
Authors: Jan Sellner, Alexander Studier-Fischer, Ahmad Bin Qasim, Silvia Seidlitz, Nicholas Schreck, Minu Tizabi, Manuel Wiesenfarth, Annette Kopp-Schneider, Janne Heinecke, Jule Brandt, Samuel Kn\"odler, Caelan Max Haney, Gabriel Salg, Berkin \"Ozdemir, Maximilian Dietrich, Maurice Stephan Michel, Felix Nickel, Karl-Friedrich Kowalewski, Lena Maier-Hein
Abstract: Novel optical imaging techniques, such as hyperspectral imaging (HSI) combined with machine learning-based (ML) analysis, have the potential to revolutionize clinical surgical imaging. However, these novel modalities face a shortage of large-scale, representative clinical data for training ML algorithms, while preclinical animal data is abundantly available through standardized experiments and allows for controlled induction of pathological tissue states, which is not ethically possible in patients. To leverage this situation, we propose a novel concept called "xeno-learning", a cross-species knowledge transfer paradigm inspired by xeno-transplantation, where organs from a donor species are transplanted into a recipient species. Using a total of 13,874 HSI images from humans as well as porcine and rat models, we show that although spectral signatures of organs differ substantially across species, relative changes resulting from pathologies or surgical manipulation (e.g., malperfusion; injection of contrast agent) are comparable. Such changes learnt in one species can thus be transferred to a new species via a novel "physiology-based data augmentation" method, enabling the large-scale secondary use of preclinical animal data for humans. The resulting ethical, monetary, and performance benefits promise a high impact of the proposed knowledge transfer paradigm on future developments in the field.
Authors: Amir Najafi, Samin Mahdizadeh Sani, Farzan Farnia
Abstract: We address the challenge of certifying the performance of a federated learning model on an unseen target network using only measurements from the source network that trained the model. Specifically, consider a source network "A" with $K$ clients, each holding private, non-IID datasets drawn from heterogeneous distributions, modeled as samples from a broader meta-distribution $\mu$. Our goal is to provide certified guarantees for the model's performance on a different, unseen network "B", governed by an unknown meta-distribution $\mu'$, assuming the deviation between $\mu$ and $\mu'$ is bounded either in Wasserstein distance or an $f$-divergence. We derive worst-case uniform guarantees for both the model's average loss and its risk CDF, the latter corresponding to a novel, adversarially robust version of the Dvoretzky-Kiefer-Wolfowitz (DKW) inequality. In addition, we show how the vanilla DKW bound enables principled certification of the model's true performance on unseen clients within the same (source) network. Our bounds are efficiently computable, asymptotically minimax optimal, and preserve clients' privacy. We also establish non-asymptotic generalization bounds that converge to zero as $K$ grows and the minimum per-client sample size exceeds $\mathcal{O}(\log K)$. Empirical evaluations confirm the practical utility of our bounds across real-world tasks. The project code is available at: github.com/samin-mehdizadeh/Robust-Evaluation-DKW
Authors: Hongpeng Jin, Yanzhao Wu
Abstract: Large Language Models (LLMs) exhibit remarkable human-like predictive capabilities. However, it is challenging to deploy LLMs to provide efficient and adaptive inference services at the edge. This paper proposes a novel Cloud-Edge Collaboration framework for LLMs (CE-CoLLM) to tackle these challenges. First, we identify the transmission of LLM contextual data between the cloud and edge as a key performance bottleneck, which introduces substantial communication overhead that dominates overall inference latency and makes na\"ive cloud-edge collaboration for LLMs inefficient. Second, we introduce a suite of novel techniques, including a latency-aware early exit mechanism and efficient cloud context management, into CE-CoLLM, which collectively reduce communication overhead and preserve LLM inference accuracy. Third, we design two adaptive inference modes to accommodate diverse edge environments: (1) a low-latency standalone edge inference mode that enables reliable edge-side independent LLM inference even under unstable network conditions, and (2) a high-accuracy cloud-edge collaborative inference mode that adaptively leverages cloud resources to enhance prediction accuracy. Extensive experiments on multiple benchmark datasets demonstrate that CE-CoLLM reduces overall inference time by up to 13.81% and offloads over 84.53% of the computational workload from the cloud to the edge, compared to conventional cloud-based LLM deployment, without sacrificing prediction accuracy. The code is provided on GitHub at https://github.com/mlsysx/CE-CoLLM.
Authors: Jiacheng Liang, Zian Wang, Lauren Hong, Shouling Ji, Ting Wang
Abstract: Various watermarking methods (``watermarkers'') have been proposed to identify LLM-generated texts; yet, due to the lack of unified evaluation platforms, many critical questions remain under-explored: i) What are the strengths/limitations of various watermarkers, especially their attack robustness? ii) How do various design choices impact their robustness? iii) How to optimally operate watermarkers in adversarial environments? To fill this gap, we systematize existing LLM watermarkers and watermark removal attacks, mapping out their design spaces. We then develop WaterPark, a unified platform that integrates 10 state-of-the-art watermarkers and 12 representative attacks. More importantly, by leveraging WaterPark, we conduct a comprehensive assessment of existing watermarkers, unveiling the impact of various design choices on their attack robustness. We further explore the best practices to operate watermarkers in adversarial environments. We believe our study sheds light on current LLM watermarking techniques while WaterPark serves as a valuable testbed to facilitate future research.
Authors: Apurva Kalia, Yan Zhou Chen, Dilip Krishnan, Soha Hassoun
Abstract: Motivation: A major challenge in metabolomics is annotation: assigning molecular structures to mass spectral fragmentation patterns. Despite recent advances in molecule-to-spectra and in spectra-to-molecular fingerprint prediction (FP), annotation rates remain low. Results: We introduce in this paper a novel paradigm (JESTR) for annotation. Unlike prior approaches that explicitly construct molecular fingerprints or spectra, JESTR leverages the insight that molecules and their corresponding spectra are views of the same data and effectively embeds their representations in a joint space. Candidate structures are ranked based on cosine similarity between the embeddings of query spectrum and each candidate. We evaluate JESTR against mol-to-spec and spec-to-FP annotation tools on three datasets. On average, for rank@[1-5], JESTR outperforms other tools by 23.6%-71.6%. We further demonstrate the strong value of regularization with candidate molecules during training, boosting rank@1 performance by 11.4% and enhancing the model's ability to discern between target and candidate molecules. When comparing JESTR's performance against that of publicly available pretrained models of SIRIUS and CFM-ID on appropriate subsets of MassSpecGym benchmark dataset, JESTR outperforms these tools by 31% and 238%, respectively. Through JESTR, we offer a novel promising avenue towards accurate annotation, therefore unlocking valuable insights into the metabolome.
Authors: Afrah Farea, Mustafa Serdar Celebi
Abstract: We investigate learnable activation functions in Physics-Informed Neural Networks (PINNs) for solving Partial Differential Equations (PDEs): comparing traditional Multilayer Perceptrons (MLPs) with fixed and trainable activations against Kolmogorov-Arnold Networks (KANs) that employ learnable basis functions. While PINNs effectively incorporate physical laws into the learning process, they suffer from convergence and spectral bias problems, which limit their applicability to problems with rapid oscillations or sharp transitions. In this work, we study various activation and basis functions across diverse PDEs, including oscillatory, nonlinear wave, mixed-physics, and fluid dynamics problems. Using empirical Neural Tangent Kernel (NTK) analysis and Hessian eigenvalue decomposition, we assess convergence behavior, stability, and high-frequency approximation capacity. While KANs offer improved expressivity for capturing complex, high-frequency PDE solutions, they introduce new optimization challenges, especially in deeper networks. Our findings show that KANs face a curse of functional dimensionality, creating intractable optimization landscapes in deeper networks. Low spectral bias alone does not guarantee good performance; adaptive spectral bias approaches such as B-splines achieve optimal results by balancing global stability with local high-frequency resolution. Different PDE types require tailored strategies: smooth global activation functions excel for wave phenomena, while local adaptive activation functions suit problems with sharp transitions.
Authors: Akshat Sharma, Hangliang Ding, Jianping Li, Neel Dani, Minjia Zhang
Abstract: How to efficiently serve LLMs in practice has become exceptionally challenging due to their prohibitive memory and computation requirements. In this study, we investigate optimizing the KV cache, whose memory footprint poses a critical bottleneck in LLM inference, especially when dealing with long context tasks. To tackle the challenge, we introduce MiniKV, a KV cache optimization method that simultaneously preserves long context task accuracy while significantly reducing KV cache size via a novel 2-bit layer-discriminative KV cache. More importantly, we develop specialized CUDA kernels to make MiniKV compatible with FlashAttention. Experiments on a wide range of long context tasks show that MiniKV effectively achieves 86% KV cache compression ratio while recovering over 98.5% of accuracy, outperforming state-of-the-art methods while achieving excellent measured system performance improvements.
Authors: Jessica Gronsbell, Jianhui Gao, Yaqi Shi, Zachary R. McCaw, David Cheng
Abstract: From structural biology to epidemiology, predictions from machine learning (ML) models are increasingly used to complement costly gold-standard data to enable faster, more affordable, and scalable scientific inquiry. In response, prediction-based (PB) inference has emerged to accommodate statistical analysis using a large volume of predictions together with a small amount of gold-standard data. The goals of PB inference are two-fold: (i) to mitigate bias from errors in predictions and (ii) to improve efficiency relative to traditional inference using only the gold-standard data. While early PB inference methods focused on bias, their ability to enhance efficiency remains unclear. We revisit a popular PB inference method and show that a simple modification can be applied to guarantee improvements in efficiency beyond yielding valid inferences when the ML predictions are imperfect. The utility of this approach in leveraging prediction-based outcomes to enhance efficiency is demonstrated through extensive simulation studies and an application to the UK Biobank data. We further contextualize the problem of PB inference through historical literature from economics and statistics to highlight perspectives from classical methods in this contemporary problem.
Authors: Zhiqiang Shen, Ammar Sherif, Zeyuan Yin, Shitong Shao
Abstract: Recent advances in dataset distillation have led to solutions in two main directions. The conventional batch-to-batch matching mechanism is ideal for small-scale datasets and includes bi-level optimization methods on models and syntheses, such as FRePo, RCIG, and RaT-BPTT, as well as other methods like distribution matching, gradient matching, and weight trajectory matching. Conversely, batch-to-global matching typifies decoupled methods, which are particularly advantageous for large-scale datasets. This approach has garnered substantial interest within the community, as seen in SRe$^2$L, G-VBSM, WMDD, and CDA. A primary challenge with the second approach is the lack of diversity among syntheses within each class since samples are optimized independently and the same global supervision signals are reused across different synthetic images. In this study, we propose a new Diversity-driven EarlyLate Training (DELT) scheme to enhance the diversity of images in batch-to-global matching with less computation. Our approach is conceptually simple yet effective, it partitions predefined IPC samples into smaller subtasks and employs local optimizations to distill each subset into distributions from distinct phases, reducing the uniformity induced by the unified optimization process. These distilled images from the subtasks demonstrate effective generalization when applied to the entire task. We conduct extensive experiments on CIFAR, Tiny-ImageNet, ImageNet-1K, and its sub-datasets. Our approach outperforms the previous state-of-the-art by 2$\sim$5% on average across different datasets and IPCs (images per class), increasing diversity per class by more than 5% while reducing synthesis time by up to 39.3% for enhancing the training efficiency. Code is available at: https://github.com/VILA-Lab/DELT.
Authors: Rajat Singh, Srikanta Bedathur
Abstract: Equivalence checking of SQL queries is an intractable problem often encountered in settings ranging from grading SQL submissions to debugging query optimizers. Despite recent work toward developing practical solutions, only simple queries written using a small subset of SQL are supported, leaving the equivalence checking of sophisticated SQL queries at the mercy of intensive, potentially error-prone, manual analysis. In this paper, we explore how LLMs can be used to reason with SQL queries to address this challenging problem. Towards this, we introduce a novel, realistic, and sufficiently complex benchmark called SQLEquiQuest for SQL query equivalence checking that reflects real-world settings. We establish strong baselines for SQL equivalence checking by leveraging the ability of LLMs to reason with SQL queries. We conduct a detailed evaluation of several state-of-the-art LLMs using various prompting strategies and carefully constructed in-context learning examples, including logical plans generated by SQL query processors. Our empirical evaluation shows that LLMs go well beyond the current capabilities of formal models for SQL equivalence, going from a mere 30% supported query pairs to full coverage, achieving up to 82% accuracy on Spider+DIN. However, a critical limitation of LLMs revealed by our analysis is that they exhibit a strong bias for equivalence predictions, with consistently poor performance over non-equivalent pairs, opening a new direction for potential future research.
Authors: Lea Bogensperger, Matthias J. Ehrhardt, Thomas Pock, Mohammad Sadegh Salehi, Hok Shing Wong
Abstract: We consider a bilevel learning framework for learning linear operators. In this framework, the learnable parameters are optimized via a loss function that also depends on the minimizer of a convex optimization problem (denoted lower-level problem). We utilize an iterative algorithm called `piggyback' to compute the gradient of the loss and minimizer of the lower-level problem. Given that the lower-level problem is solved numerically, the loss function and thus its gradient can only be computed inexactly. To estimate the accuracy of the computed hypergradient, we derive an a-posteriori error bound, which provides guides for setting the tolerance for the lower-level problem, as well as the piggyback algorithm. To efficiently solve the upper-level optimization, we also propose an adaptive method for choosing a suitable step-size. To illustrate the proposed method, we consider a few learned regularizer problems, such as training an input-convex neural network.
Authors: Zhaomeng Chen, Junting Duan, Victor Chernozhukov, Vasilis Syrgkanis
Abstract: This paper proposes the automatic Doubly Robust Random Forest (DRRF) algorithm for estimating the conditional expectation of a moment functional in the presence of high-dimensional nuisance functions. DRRF extends the automatic debiasing framework based on the Riesz representer to the conditional setting and enables nonparametric, forest-based estimation (Athey et al., 2019; Oprescu et al., 2019). In contrast to existing methods, DRRF does not require prior knowledge of the form of the debiasing term or impose restrictive parametric or semi-parametric assumptions on the target quantity. Additionally, it is computationally efficient in making predictions at multiple query points. We establish consistency and asymptotic normality results for the DRRF estimator under general assumptions, allowing for the construction of valid confidence intervals. Through extensive simulations in heterogeneous treatment effect (HTE) estimation, we demonstrate the superior performance of DRRF over benchmark approaches in terms of estimation accuracy, robustness, and computational efficiency.
Authors: Jaesung R. Park, Jaewook J. Suh, Youngjoon Hong, Ernest K. Ryu
Abstract: In deep learning, the recently introduced state space models utilize HiPPO (High-order Polynomial Projection Operators) memory units to approximate continuous-time trajectories of input functions using ordinary differential equations (ODEs), and these techniques have shown empirical success in capturing long-range dependencies in long input sequences. However, the mathematical foundations of these ODEs, particularly the singular HiPPO-LegS (Legendre Scaled) ODE, and their corresponding numerical discretizations remain unsettled. In this work, we fill this gap by establishing that HiPPO-LegS ODE is well-posed despite its singularity, albeit without the freedom of arbitrary initial conditions. Further, we establish convergence of the associated numerical discretization schemes for Riemann integrable input functions.
Authors: Haolin Li, Yuyang Miao, Zahra Sharif Khodaei, M. H. Aliabadi
Abstract: PINN models have demonstrated capabilities in addressing fluid PDE problems, and their potential in solid mechanics is beginning to emerge. This study identifies two key challenges when using PINN to solve general solid mechanics problems. These challenges become evident when comparing the limitations of PINN with the well-established numerical methods commonly used in solid mechanics, such as the finite element method (FEM). Specifically: a) PINN models generate solutions over an infinite domain, which conflicts with the finite boundaries typical of most solid structures; and b) the solution space utilised by PINN is Euclidean, which is inadequate for addressing the complex geometries often present in solid structures. This work presents a PINN architecture for general solid mechanics problems, referred to as the Finite-PINN model. The model is designed to effectively tackle two key challenges, while retaining as much of the original PINN framework as possible. To this end, the Finite-PINN incorporates finite geometric encoding into the neural network inputs, thereby transforming the solution space from a conventional Euclidean space into a hybrid Euclidean-topological space. The model is comprehensively trained using both strong-form and weak-form loss formulations, enabling its application to a wide range of forward and inverse problems in solid mechanics. For forward problems, the Finite-PINN model efficiently approximates solutions to solid mechanics problems when the geometric information of a given structure has been preprocessed. For inverse problems, it effectively reconstructs full-field solutions from very sparse observations by embedding both physical laws and geometric information within its architecture.
Authors: Fengpei Li, Haoxian Chen, Jiahe Lin, Arkin Gupta, Xiaowei Tan, Honglei Zhao, Gang Xu, Yuriy Nevmyvaka, Agostino Capponi, Henry Lam
Abstract: For many complex simulation tasks spanning areas such as healthcare, engineering, and finance, Monte Carlo (MC) methods are invaluable due to their unbiased estimates and precise error quantification. Nevertheless, Monte Carlo simulations often become computationally prohibitive, especially for nested, multi-level, or path-dependent evaluations lacking effective variance reduction techniques. While machine learning (ML) surrogates appear as natural alternatives, naive replacements typically introduce unquantifiable biases. We address this challenge by introducing Prediction-Enhanced Monte Carlo (PEMC), a framework that leverages modern ML models as learned predictors, using cheap and parallelizable simulation as features, to output unbiased evaluation with reduced variance and runtime. PEMC can also be viewed as a "modernized" view of control variates, where we consider the overall computation-cost-aware variance reduction instead of per-replication reduction, while bypassing the closed-form mean function requirement and maintaining the advantageous unbiasedness and uncertainty quantifiability of Monte Carlo. We illustrate PEMC's broader efficacy and versatility through three examples: first, equity derivatives such as variance swaps under stochastic local volatility models; second, interest rate derivatives such as swaption pricing under the Heath-Jarrow-Morton (HJM) interest-rate model. Finally, we showcase PEMC in a socially significant context - ambulance dispatch and hospital load balancing - where accurate mortality rate estimates are key for ethically sensitive decision-making. Across these diverse scenarios, PEMC consistently reduces variance while preserving unbiasedness, highlighting its potential as a powerful enhancement to standard Monte Carlo baselines.
Authors: James E. Gallagher, Edward J. Oughton
Abstract: This paper presents the Visual Optical Recognition Telemetry EXtraction (VORTEX) system for extracting and analyzing drone telemetry data from First Person View (FPV) Uncrewed Aerial System (UAS) footage. VORTEX employs MMOCR, a PyTorch-based Optical Character Recognition (OCR) toolbox, to extract telemetry variables from drone Heads Up Display (HUD) recordings, utilizing advanced image preprocessing techniques, including CLAHE enhancement and adaptive thresholding. The study optimizes spatial accuracy and computational efficiency through systematic investigation of temporal sampling rates (1s, 5s, 10s, 15s, 20s) and coordinate processing methods. Results demonstrate that the 5-second sampling rate, utilizing 4.07% of available frames, provides the optimal balance with a point retention rate of 64% and mean speed accuracy within 4.2% of the 1-second baseline while reducing computational overhead by 80.5%. Comparative analysis of coordinate processing methods reveals that while UTM Zone 33N projection and Haversine calculations provide consistently similar results (within 0.1% difference), raw WGS84 coordinates underestimate distances by 15-30% and speeds by 20-35%. Altitude measurements showed unexpected resilience to sampling rate variations, with only 2.1% variation across all intervals. This research is the first of its kind, providing quantitative benchmarks for establishing a robust framework for drone telemetry extraction and analysis using open-source tools and spatial libraries.
Authors: Maxence Boels, Yang Liu, Prokar Dasgupta, Alejandro Granados, Sebastien Ourselin
Abstract: While existing approaches excel at recognising current surgical phases, they provide limited foresight and intraoperative guidance into future procedural steps. Similarly, current anticipation methods are constrained to predicting short-term and single events, neglecting the dense, repetitive, and long sequential nature of surgical workflows. To address these needs and limitations, we propose SWAG (Surgical Workflow Anticipative Generation), a framework that combines phase recognition and anticipation using a generative approach. This paper investigates two distinct decoding methods - single-pass (SP) and auto-regressive (AR) - to generate sequences of future surgical phases at minute intervals over long horizons. We propose a novel embedding approach using class transition probabilities to enhance the accuracy of phase anticipation. Additionally, we propose a generative framework using remaining time regression to classification (R2C). SWAG was evaluated on two publicly available datasets, Cholec80 and AutoLaparo21. Our single-pass model with class transition probability embeddings (SP*) achieves 32.1% and 41.3% F1 scores over 20 and 30 minutes on Cholec80 and AutoLaparo21, respectively. Moreover, our approach competes with existing methods on phase remaining time regression, achieving weighted mean absolute errors of 0.32 and 0.48 minutes for 2- and 3-minute horizons. SWAG demonstrates versatility across generative decoding frame works and classification and regression tasks to create temporal continuity between surgical workflow recognition and anticipation. Our method provides steps towards intraoperative surgical workflow generation for anticipation. Project: https://maxboels.github.io/swag.
Authors: Phi Vu Tran
Abstract: While modern visual recognition systems have made significant advancements, many continue to struggle with the open problem of learning from few exemplars. This paper focuses on the task of object detection in the setting where object classes follow a natural long-tailed distribution. Existing methods for long-tailed detection resort to external ImageNet labels to augment the low-shot training instances. However, such dependency on a large labeled database has limited utility in practical scenarios. We propose a versatile and scalable approach to leverage optional unlabeled images, which are easy to collect without the burden of human annotations. Our SimLTD framework is straightforward and intuitive, and consists of three simple steps: (1) pre-training on abundant head classes; (2) transfer learning on scarce tail classes; and (3) fine-tuning on a sampled set of both head and tail classes. Our approach can be viewed as an improved head-to-tail model transfer paradigm without the added complexities of meta-learning or knowledge distillation, as was required in past research. By harnessing supplementary unlabeled images, without extra image labels, SimLTD establishes new record results on the challenging LVIS v1 benchmark across both supervised and semi-supervised settings.
Authors: Bohdan Turbal, Anastasiia Mazur, Jiaxu Zhao, Mykola Pechenizkiy
Abstract: We investigate the adversarial robustness of LLMs in transfer learning scenarios. Through comprehensive experiments on multiple datasets (MBIB Hate Speech, MBIB Political Bias, MBIB Gender Bias) and various model architectures (BERT, RoBERTa, GPT-2, Gemma, Phi), we reveal that transfer learning, while improving standard performance metrics, often leads to increased vulnerability to adversarial attacks. Our findings demonstrate that larger models exhibit greater resilience to this phenomenon, suggesting a complex interplay between model size, architecture, and adaptation methods. Our work highlights the crucial need for considering adversarial robustness in transfer learning scenarios and provides insights into maintaining model security without compromising performance. These findings have significant implications for the development and deployment of LLMs in real-world applications where both performance and robustness are paramount.
Authors: Ashay Athalye, Nishanth Kumar, Tom Silver, Yichao Liang, Jiuguang Wang, Tom\'as Lozano-P\'erez, Leslie Pack Kaelbling
Abstract: Our aim is to learn to solve long-horizon decision-making problems in complex robotics domains given low-level skills and a handful of short-horizon demonstrations containing sequences of images. To this end, we focus on learning abstract symbolic world models that facilitate zero-shot generalization to novel goals via planning. A critical component of such models is the set of symbolic predicates that define properties of and relationships between objects. In this work, we leverage pretrained vision language models (VLMs) to propose a large set of visual predicates potentially relevant for decision-making, and to evaluate those predicates directly from camera images. At training time, we pass the proposed predicates and demonstrations into an optimization-based model-learning algorithm to obtain an abstract symbolic world model that is defined in terms of a compact subset of the proposed predicates. At test time, given a novel goal in a novel setting, we use the VLM to construct a symbolic description of the current world state, and then use a search-based planning algorithm to find a sequence of low-level skills that achieves the goal. We demonstrate empirically across experiments in both simulation and the real world that our method can generalize aggressively, applying its learned world model to solve problems with a wide variety of object types, arrangements, numbers of objects, and visual backgrounds, as well as novel goals and much longer horizons than those seen at training time.
Authors: Yuchuan Tian, Jing Han, Chengcheng Wang, Yuchen Liang, Chao Xu, Hanting Chen
Abstract: Diffusion models have shown exceptional performance in visual generation tasks. Recently, these models have shifted from traditional U-Shaped CNN-Attention hybrid structures to fully transformer-based isotropic architectures. While these transformers exhibit strong scalability and performance, their reliance on complicated self-attention operation results in slow inference speeds. Contrary to these works, we rethink one of the simplest yet fastest module in deep learning, 3x3 Convolution, to construct a scaled-up purely convolutional diffusion model. We first discover that an Encoder-Decoder Hourglass design outperforms scalable isotropic architectures for Conv3x3, but still under-performing our expectation. Further improving the architecture, we introduce sparse skip connections to reduce redundancy and improve scalability. Based on the architecture, we introduce conditioning improvements including stage-specific embeddings, mid-block condition injection, and conditional gating. These improvements lead to our proposed Diffusion CNN (DiC), which serves as a swift yet competitive diffusion architecture baseline. Experiments on various scales and settings show that DiC surpasses existing diffusion transformers by considerable margins in terms of performance while keeping a good speed advantage. Project page: https://github.com/YuchuanTian/DiC
Authors: Yifu Qiu, Varun Embar, Yizhe Zhang, Navdeep Jaitly, Shay B. Cohen, Benjamin Han
Abstract: Recent advancements in long-context language models (LCLMs) promise to transform Retrieval-Augmented Generation (RAG) by simplifying pipelines. With their expanded context windows, LCLMs can process entire knowledge bases and perform retrieval and reasoning directly -- a capability we define as In-Context Retrieval and Reasoning (ICR^2). However, existing benchmarks like LOFT often overestimate LCLM performance by providing overly simplified contexts. To address this, we introduce ICR^2, a benchmark that evaluates LCLMs in more realistic scenarios by including confounding passages retrieved with strong retrievers. We then propose three methods to enhance LCLM performance: (1) retrieve-then-generate fine-tuning, (2) retrieval-attention-probing, which uses attention heads to filter and de-noise long contexts during decoding, and (3) joint retrieval head training alongside the generation head. Our evaluation of five well-known LCLMs on LOFT and ICR^2 demonstrates significant gains with our best approach applied to Mistral-7B: +17 and +15 points by Exact Match on LOFT, and +13 and +2 points on ICR^2, compared to vanilla RAG and supervised fine-tuning, respectively. It even outperforms GPT-4-Turbo on most tasks despite being a much smaller model.
Authors: Kangjie Zheng, Junwei Yang, Siyue Liang, Bin Feng, Zequn Liu, Wei Ju, Zhiping Xiao, Ming Zhang
Abstract: Masked Language Models (MLMs) have achieved remarkable success in many self-supervised representation learning tasks. MLMs are trained by randomly masking portions of the input sequences with [MASK] tokens and learning to reconstruct the original content based on the remaining context. This paper explores the impact of [MASK] tokens on MLMs. Analytical studies show that masking tokens can introduce the corrupted semantics problem, wherein the corrupted context may convey multiple, ambiguous meanings. This problem is also a key factor affecting the performance of MLMs on downstream tasks. Based on these findings, we propose a novel enhanced-context MLM, ExLM. Our approach expands [MASK] tokens in the input context and models the dependencies between these expanded states. This enhancement increases context capacity and enables the model to capture richer semantic information, effectively mitigating the corrupted semantics problem during pre-training. Experimental results demonstrate that ExLM achieves significant performance improvements in both text modeling and SMILES modeling tasks. Further analysis confirms that ExLM enriches semantic representations through context enhancement, and effectively reduces the semantic multimodality commonly observed in MLMs.
Authors: Jaturong Kongmanee
Abstract: This research aims to unravel how large language models (LLMs) iteratively refine token predictions through internal processing. We utilized a logit lens technique to analyze the model's token predictions derived from intermediate representations. Specifically, we focused on (1) how LLMs access and utilize information from input contexts, and (2) how positioning of relevant information affects the model's token prediction refinement process. On a multi-document question answering task with varying input context lengths, we found that the depth of prediction refinement (defined as the number of intermediate layers an LLM uses to transition from an initial correct token prediction to its final, stable correct output), as a function of the position of relevant information, exhibits an approximately inverted U-shaped curve. We also found that the gap between these two layers, on average, diminishes when relevant information is positioned at the beginning or end of the input context. This suggested that the model requires more refinements when processing longer contexts with relevant information situated in the middle. Furthermore, our findings indicate that not all layers are equally essential for determining final correct outputs. Our analysis provides insights into how token predictions are distributed across different conditions, and establishes important connections to existing hypotheses and previous findings in AI safety research and development.
Authors: Tam Le, Truyen Nguyen, Hideitsu Hino, Kenji Fukumizu
Abstract: We investigate the Sobolev IPM problem for probability measures supported on a graph metric space. Sobolev IPM is an important instance of integral probability metrics (IPM), and is obtained by constraining a critic function within a unit ball defined by the Sobolev norm. In particular, it has been used to compare probability measures and is crucial for several theoretical works in machine learning. However, to our knowledge, there are no efficient algorithmic approaches to compute Sobolev IPM effectively, which hinders its practical applications. In this work, we establish a relation between Sobolev norm and weighted $L^p$-norm, and leverage it to propose a \emph{novel regularization} for Sobolev IPM. By exploiting the graph structure, we demonstrate that the regularized Sobolev IPM provides a \emph{closed-form} expression for fast computation. This advancement addresses long-standing computational challenges, and paves the way to apply Sobolev IPM for practical applications, even in large-scale settings. Additionally, the regularized Sobolev IPM is negative definite. Utilizing this property, we design positive-definite kernels upon the regularized Sobolev IPM, and provide preliminary evidences of their advantages for comparing probability measures on a given graph for document classification and topological data analysis.
Authors: Deqian Kong, Minglu Zhao, Dehong Xu, Bo Pang, Shu Wang, Edouardo Honig, Zhangzhang Si, Chuan Li, Jianwen Xie, Sirui Xie, Ying Nian Wu
Abstract: We propose a novel class of language models, Latent Thought Models (LTMs), which incorporate explicit latent thought vectors that follow an explicit prior model in latent space. These latent thought vectors guide the autoregressive generation of ground tokens through a Transformer decoder. Training employs a dual-rate optimization process within the classical variational Bayes framework: fast learning of local variational parameters for the posterior distribution of latent vectors (inference-time computation), and slow learning of global decoder parameters. Empirical studies reveal that LTMs possess additional scaling dimensions beyond traditional Large Language Models (LLMs), such as the number of iterations in inference-time computation and number of latent thought vectors. Higher sample efficiency can be achieved by increasing training compute per token, with further gains possible by trading model size for more inference steps. Designed based on these scaling properties, LTMs demonstrate superior sample and parameter efficiency compared to autoregressive models and discrete diffusion models. They significantly outperform these counterparts in validation perplexity and zero-shot language modeling tasks. Additionally, LTMs exhibit emergent few-shot in-context reasoning capabilities that scale with model size, and achieve competitive performance in conditional and unconditional text generation.
Authors: Zeren Luo, Zifan Peng, Yule Liu, Zhen Sun, Mingchen Li, Jingyi Zheng, Xinlei He
Abstract: Recent advancements in Large Language Models (LLMs) have significantly enhanced the capabilities of AI-Powered Search Engines (AIPSEs), offering precise and efficient responses by integrating external databases with pre-existing knowledge. However, we observe that these AIPSEs raise risks such as quoting malicious content or citing malicious websites, leading to harmful or unverified information dissemination. In this study, we conduct the first safety risk quantification on seven production AIPSEs by systematically defining the threat model, risk type, and evaluating responses to various query types. With data collected from PhishTank, ThreatBook, and LevelBlue, our findings reveal that AIPSEs frequently generate harmful content that contains malicious URLs even with benign queries (e.g., with benign keywords). We also observe that directly querying a URL will increase the number of main risk-inclusive responses, while querying with natural language will slightly mitigate such risk. Compared to traditional search engines, AIPSEs outperform in both utility and safety. We further perform two case studies on online document spoofing and phishing to show the ease of deceiving AIPSEs in the real-world setting. To mitigate these risks, we develop an agent-based defense with a GPT-4.1-based content refinement tool and a URL detector. Our evaluation shows that our defense can effectively reduce the risk, with only a minor cost of reducing available information by approximately 10.7%. Our research highlights the urgent need for robust safety measures in AIPSEs.
Authors: Elizaveta Semenova, Alisa Sheinkman, Timothy James Hitge, Siobhan Mackenzie Hall, Jon Cockayne
Abstract: Surrogate models are widely used to approximate complex systems across science and engineering to reduce computational costs. Despite their widespread adoption, the field lacks standardisation across key stages of the modelling pipeline, including data sampling, model selection, evaluation, and downstream analysis. This fragmentation limits reproducibility and cross-domain utility -- a challenge further exacerbated by the rapid proliferation of AI-driven surrogate models. We argue for the urgent need to establish a structured reporting standard, the Surrogate Model Reporting Specification (SMRS), that systematically captures essential design and evaluation choices while remaining agnostic to implementation specifics. By promoting a standardised yet flexible framework, we aim to improve the reliability of surrogate modelling, foster interdisciplinary knowledge transfer, and, as a result, accelerate scientific progress in the AI era.
Authors: Michael Kearns, Mirah Shi
Abstract: Algorithmic trading in modern financial markets is widely acknowledged to exhibit strategic, game-theoretic behaviors whose complexity can be difficult to model. A recent series of papers (Chriss, 2024b,c,a, 2025) has made progress in the setting of trading for position building. Here parties wish to buy or sell a fixed number of shares in a fixed time period in the presence of both temporary and permanent market impact, resulting in exponentially large strategy spaces. While these papers primarily consider the existence and structural properties of equilibrium strategies, in this work we focus on the algorithmic aspects of the proposed model. We give an efficient algorithm for computing best responses, and show that while the temporary impact only setting yields a potential game, best response dynamics do not generally converge for the general setting, for which no fast algorithm for (Nash) equilibrium computation is known. This leads us to consider the broader notion of Coarse Correlated Equilibria (CCE), which we show can be computed efficiently via an implementation of Follow the Perturbed Leader (FTPL). We illustrate the model and our results with an experimental investigation, where FTPL exhibits interesting behavior in different regimes of the relative weighting between temporary and permanent market impact.
Authors: Sebastin Santy, Prasanta Bhattacharya, Manoel Horta Ribeiro, Kelsey Allen, Sewoong Oh
Abstract: Progress in AI has relied on human-generated data, from annotator marketplaces to the wider Internet. However, the widespread use of large language models now threatens the quality and integrity of human-generated data on these very platforms. We argue that this issue goes beyond the immediate challenge of filtering AI-generated content -- it reveals deeper flaws in how data collection systems are designed. Existing systems often prioritize speed, scale, and efficiency at the cost of intrinsic human motivation, leading to declining engagement and data quality. We propose that rethinking data collection systems to align with contributors' intrinsic motivations -- rather than relying solely on external incentives -- can help sustain high-quality data sourcing at scale while maintaining contributor trust and long-term participation.
Authors: Yang Li, Feng Ruan
Abstract: The classical kernel ridge regression problem aims to find the best fit for the output $Y$ as a function of the input data $X\in \mathbb{R}^d$, with a fixed choice of regularization term imposed by a given choice of a reproducing kernel Hilbert space, such as a Sobolev space. Here we consider a generalization of the kernel ridge regression problem, by introducing an extra matrix parameter $U$, which aims to detect the scale parameters and the feature variables in the data, and thereby improve the efficiency of kernel ridge regression. This naturally leads to a nonlinear variational problem to optimize the choice of $U$. We study various foundational mathematical aspects of this variational problem, and in particular how this behaves in the presence of multiscale structures in the data.
Authors: Nuno Gon\c{c}alves, Marcos Treviso, Andr\'e F. T. Martins
Abstract: The computational cost of softmax-based attention in transformers limits their applicability to long-context tasks. Adaptive sparsity, of which $\alpha$-entmax attention is an example, offers a flexible data-dependent alternative, but existing implementations are inefficient and do not leverage the sparsity to obtain runtime and memory gains. In this work, we propose AdaSplash, which combines the efficiency of GPU-optimized algorithms with the sparsity benefits of $\alpha$-entmax. We first introduce a hybrid Halley-bisection algorithm, resulting in a 7-fold reduction in the number of iterations needed to compute the $\alpha$-entmax transformation. Then, we implement custom Triton kernels to efficiently handle adaptive sparsity. Experiments with RoBERTa and ModernBERT for text classification and single-vector retrieval, along with GPT-2 for language modeling, show that our method achieves substantial improvements in runtime and memory efficiency compared to existing $\alpha$-entmax implementations. It approaches -- and in some cases surpasses -- the efficiency of highly optimized softmax implementations like FlashAttention-2, enabling long-context training while maintaining strong task performance.
Authors: Ching-Hua Lee, Chouchang Yang, Jaejin Cho, Yashas Malur Saidutta, Rakshith Sharma Srinivasa, Yilin Shen, Hongxia Jin
Abstract: Denoising diffusion probabilistic models (DDPMs) can be utilized to recover a clean signal from its degraded observation(s) by conditioning the model on the degraded signal. The degraded signals are themselves contaminated versions of the clean signals; due to this correlation, they may encompass certain useful information about the target clean data distribution. However, existing adoption of the standard Gaussian as the prior distribution in turn discards such information when shaping the prior, resulting in sub-optimal performance. In this paper, we propose to improve conditional DDPMs for signal restoration by leveraging a more informative prior that is jointly learned with the diffusion model. The proposed framework, called RestoreGrad, seamlessly integrates DDPMs into the variational autoencoder (VAE) framework, taking advantage of the correlation between the degraded and clean signals to encode a better diffusion prior. On speech and image restoration tasks, we show that RestoreGrad demonstrates faster convergence (5-10 times fewer training steps) to achieve better quality of restored signals over existing DDPM baselines and improved robustness to using fewer sampling steps in inference time (2-2.5 times fewer), advocating the advantages of leveraging jointly learned prior for efficiency improvements in the diffusion process.
Authors: Sida Li, Nikolaos Ignatiadis
Abstract: Prediction-Powered Inference (PPI) is a powerful framework for enhancing statistical estimates by combining limited gold-standard data with machine learning (ML) predictions. While prior work has demonstrated PPI's benefits for individual statistical problems, modern applications require answering numerous parallel statistical questions. We introduce Prediction-Powered Adaptive Shrinkage (PAS), a method that bridges PPI with empirical Bayes shrinkage to improve the estimation of multiple means. PAS debiases noisy ML predictions within each task and then borrows strength across tasks by using those same predictions as a reference point for shrinkage. The amount of shrinkage is determined by minimizing an unbiased estimate of risk, and we prove that this tuning strategy is asymptotically optimal. Experiments on both synthetic and real-world datasets show that PAS adapts to the reliability of the ML predictions and outperforms traditional and modern baselines in large-scale applications.
Authors: Ivan Skorokhodov, Sharath Girish, Benran Hu, Willi Menapace, Yanyu Li, Rameen Abdal, Sergey Tulyakov, Aliaksandr Siarohin
Abstract: Latent diffusion models have emerged as the leading approach for generating high-quality images and videos, utilizing compressed latent representations to reduce the computational burden of the diffusion process. While recent advancements have primarily focused on scaling diffusion backbones and improving autoencoder reconstruction quality, the interaction between these components has received comparatively less attention. In this work, we perform a spectral analysis of modern autoencoders and identify inordinate high-frequency components in their latent spaces, which are especially pronounced in the autoencoders with a large bottleneck channel size. We hypothesize that this high-frequency component interferes with the coarse-to-fine nature of the diffusion synthesis process and hinders the generation quality. To mitigate the issue, we propose scale equivariance: a simple regularization strategy that aligns latent and RGB spaces across frequencies by enforcing scale equivariance in the decoder. It requires minimal code changes and only up to 20K autoencoder fine-tuning steps, yet significantly improves generation quality, reducing FID by 19% for image generation on ImageNet-1K $256^2$ and FVD by at least 44% for video generation on Kinetics-700 $17 \times 256^2$. The source code is available at https://github.com/snap-research/diffusability.
Authors: Seokhun Park, Insung Kong, Yongchan Choi, Chanmoo Park, Yongdai Kim
Abstract: Interpretability for machine learning models is becoming more and more important as machine learning models become more complex. The functional ANOVA model, which decomposes a high-dimensional function into a sum of lower dimensional functions (commonly referred to as components), is one of the most popular tools for interpretable AI, and recently, various neural networks have been developed for estimating each component in the functional ANOVA model. However, such neural networks are highly unstable when estimating each component since the components themselves are not uniquely defined. That is, there are multiple functional ANOVA decompositions for a given function. In this paper, we propose a novel neural network which guarantees a unique functional ANOVA decomposition and thus is able to estimate each component stably and accurately. We call our proposed neural network ANOVA Tensor Product Neural Network (ANOVA-TPNN) since it is motivated by the tensor product basis expansion. Theoretically, we prove that ANOVA-TPNN can approximate any smooth function well. Empirically, we show that ANOVA-TPNN provide much more stable estimation of each component and thus much more stable interpretation when training data and initial values of the model parameters vary than existing neural networks do.
Authors: Weikang Qiu, Zheng Huang, Haoyu Hu, Aosong Feng, Yujun Yan, Rex Ying
Abstract: Decoding functional magnetic resonance imaging (fMRI) signals into text has been a key challenge in the neuroscience community, with the potential to advance brain-computer interfaces and uncover deeper insights into brain mechanisms. However, existing approaches often struggle with suboptimal predictive performance, limited task variety, and poor generalization across subjects. In response to this, we propose MindLLM, a model designed for subject-agnostic and versatile fMRI-to-text decoding. MindLLM consists of an fMRI encoder and an off-the-shelf LLM. The fMRI encoder employs a neuroscience-informed attention mechanism, which is capable of accommodating subjects with varying input shapes and thus achieves high-performance subject-agnostic decoding. Moreover, we introduce Brain Instruction Tuning (BIT), a novel approach that enhances the model's ability to capture diverse semantic representations from fMRI signals, facilitating more versatile decoding. We evaluate MindLLM on comprehensive fMRI-to-text benchmarks. Results demonstrate that our model outperforms the baselines, improving downstream tasks by 12.0%, unseen subject generalization by 24.5%, and novel task adaptation by 25.0%. Furthermore, the attention patterns in MindLLM provide interpretable insights into its decision-making process.
Authors: Ravi Ghadia, Avinash Kumar, Gaurav Jain, Prashant Nair, Poulami Das
Abstract: Autoregressive Transformers rely on Key-Value (KV) caching to accelerate inference. However, the linear growth of the KV cache with context length leads to excessive memory consumption and bandwidth constraints. This bottleneck is particularly problematic in real-time applications -- such as chatbots and interactive assistants -- where low latency and high memory efficiency are critical. Existing methods drop distant tokens or compress states in a lossy manner, sacrificing accuracy by discarding vital context or introducing bias. We propose MorphKV, an inference-time technique that maintains a constant-sized KV cache while preserving accuracy. MorphKV balances long-range dependencies and local coherence during text generation. It eliminates early-token bias while retaining high-fidelity context by adaptively ranking tokens through correlation-aware selection. Unlike heuristic retention or lossy compression, MorphKV iteratively refines the KV cache via lightweight updates guided by attention patterns of recent tokens. This approach captures inter-token correlation with greater accuracy, crucial for tasks like content creation and code generation. Our studies on long-response tasks show 52.9$\%$ memory savings and 18.2$\%$ higher accuracy on average compared to state-of-the-art prior works, enabling efficient real-world deployment.
Authors: Niccol\`o Turcato, Matteo Iovino, Aris Synodinos, Alberto Dalla Libera, Ruggero Carli, Pietro Falco
Abstract: Recent advancements in Large Language Models (LLMs) and Visual Language Models (VLMs) have significantly impacted robotics, enabling high-level semantic motion planning applications. Reinforcement Learning (RL), a complementary paradigm, enables agents to autonomously optimize complex behaviors through interaction and reward signals. However, designing effective reward functions for RL remains challenging, especially in real-world tasks where sparse rewards are insufficient and dense rewards require elaborate design. In this work, we propose Autonomous Reinforcement learning for Complex Human-Informed Environments (ARCHIE), an unsupervised pipeline leveraging GPT-4, a pre-trained LLM, to generate reward functions directly from natural language task descriptions. The rewards are used to train RL agents in simulated environments, where we formalize the reward generation process to enhance feasibility. Additionally, GPT-4 automates the coding of task success criteria, creating a fully automated, one-shot procedure for translating human-readable text into deployable robot skills. Our approach is validated through extensive simulated experiments on single-arm and bi-manual manipulation tasks using an ABB YuMi collaborative robot, highlighting its practicality and effectiveness. Tasks are demonstrated on the real robot setup.
Authors: Dimitri von R\"utte, Janis Fluri, Yuhui Ding, Antonio Orvieto, Bernhard Sch\"olkopf, Thomas Hofmann
Abstract: While state-of-the-art language models achieve impressive results through next-token prediction, they have inherent limitations such as the inability to revise already generated tokens. This has prompted exploration of alternative approaches such as discrete diffusion. However, masked diffusion, which has emerged as a popular choice due to its simplicity and effectiveness, reintroduces this inability to revise words. To overcome this, we generalize masked diffusion, deriving a new family of general interpolating discrete diffusion (GIDD) which offers greater flexibility in the design of the noising processes. Leveraging a novel diffusion ELBO, we achieve compute-matched state-of-the-art performance in diffusion language modeling. Exploiting GIDD's flexibility, we explore a hybrid approach combining masking and uniform noise, leading to improved sample quality and unlocking the ability for the model to correct its own mistakes, an area where autoregressive models notoriously have struggled. Code: https://github.com/dvruette/gidd/
Authors: Tianyu Cui, Song-Jun Xu, Artem Moskalev, Shuwei Li, Tommaso Mansi, Mangal Prakash, Rui Liao
Abstract: Inferring Gene Regulatory Networks (GRNs) from gene expression data is crucial for understanding biological processes. While supervised models are reported to achieve high performance for this task, they rely on costly ground truth (GT) labels and risk learning gene-specific biases, such as class imbalances of GT interactions, rather than true regulatory mechanisms. To address these issues, we introduce InfoSEM, an unsupervised generative model that leverages textual gene embeddings as informative priors, improving GRN inference without GT labels. InfoSEM can also integrate GT labels as an additional prior when available, avoiding biases and further enhancing performance. Additionally, we propose a biologically motivated benchmarking framework that better reflects real-world applications such as biomarker discovery and reveals learned biases of existing supervised methods. InfoSEM outperforms existing models by 38.5% across four datasets using textual embeddings prior and further boosts performance by 11.1% when integrating labeled data as priors.
Authors: Wenjie Qiu, Yi-Chen Li, Xuqin Zhang, Tianyi Zhang, Yihang Zhang, Zongzhang Zhang, Yang Yu
Abstract: Learning reward models from human preference datasets and subsequently optimizing language models via reinforcement learning has emerged as a fundamental paradigm for aligning LLMs with human preferences. The performance of the reward model plays a crucial role in the effectiveness of alignment. Previous reward models operate at a coarse-grained level, requiring the generation of a complete response to obtain a reward value. The sparse reward may present challenges for downstream reinforcement learning. While recent efforts have attempted to learn token-level reward models, the lack of explicit semantic information makes it difficult to model the credit of every individual token. In this paper, we propose assigning scores to every sentence, introducing an intermediate-grained reward model. By segmenting the complete response into sentences and applying differential operations to reward output at the start and end positions of each sentence, we can effectively model the rewards of sentences. Moreover, a novel attention mechanism is introduced to aggregate the scores of all sentences into a response-level score, which allows it to be trained using the Bradley-Terry model. On common benchmarks, our method outperforms the response-level reward model by 2.7% on RewardBench (for reward modeling evaluation) and surpasses all baselines on AlpacaEval (for alignment evaluation).
Authors: Claudius Kienle, Benjamin Alt, Finn Schneider, Tobias Pertlwieser, Rainer J\"akel, Rania Rayyes
Abstract: Despite the widespread adoption of industrial robots in automotive assembly, wire harness installation remains a largely manual process, as it requires precise and flexible manipulation. To address this challenge, we design a novel AI-based framework that automates cable connector mating by integrating force control with deep visuotactile learning. Our system optimizes search-and-insertion strategies using first-order optimization over a multimodal transformer architecture trained on visual, tactile, and proprioceptive data. Additionally, we design a novel automated data collection and optimization pipeline that minimizes the need for machine learning expertise. The framework optimizes robot programs that run natively on standard industrial controllers, permitting human experts to audit and certify them. Experimental validations on a center console assembly task demonstrate significant improvements in cycle times and robustness compared to conventional robot programming approaches. Videos are available under https://claudius-kienle.github.io/AppMuTT.
Authors: Wa\"iss Azizian, Franck Iutzeler, J\'er\^ome Malick, Panayotis Mertikopoulos
Abstract: In this paper, we examine the time it takes for stochastic gradient descent (SGD) to reach the global minimum of a general, non-convex loss function. We approach this question through the lens of randomly perturbed dynamical systems and large deviations theory, and we provide a tight characterization of the global convergence time of SGD via matching upper and lower bounds. These bounds are dominated by the most "costly" set of obstacles that the algorithm may need to overcome in order to reach a global minimizer from a given initialization, coupling in this way the global geometry of the underlying loss landscape with the statistics of the noise entering the process. Finally, motivated by applications to the training of deep neural networks, we also provide a series of refinements and extensions of our analysis for loss functions with shallow local minima.
Authors: Suho Yoo, Hyunjong Ok, Jaeho Lee
Abstract: Language models pretrained on text-only corpora often struggle with tasks that require auditory commonsense knowledge. Previous work addresses this problem by augmenting the language model to retrieve knowledge from external audio databases. This approach has several limitations, such as the potential lack of relevant audio in databases and the high costs associated with constructing the databases. To address these issues, we propose Imagine to Hear, a novel approach that dynamically generates auditory knowledge using generative models. Our framework detects multiple audio-related textual spans from the given prompt and generates corresponding auditory knowledge. We develop several mechanisms to efficiently process multiple auditory knowledge, including a CLAP-based rejection sampler and a language-audio fusion module. Our experiments show that our method achieves state-of-the-art performance on AuditoryBench without relying on external databases, highlighting the effectiveness of our generation-based approach.
Authors: Chung-En Sun, Ge Yan, Tsui-Wei Weng
Abstract: Recent studies have shown that Large Language Models (LLMs) augmented with chain-of-thought (CoT) reasoning demonstrate impressive problem-solving abilities. However, in this work, we identify a recurring issue where these models occasionally generate overly short reasoning, leading to degraded performance on even simple mathematical problems. Specifically, we investigate how reasoning length is embedded in the hidden representations of reasoning models and its impact on accuracy. Our analysis reveals that reasoning length is governed by a linear direction in the representation space, allowing us to induce overly short reasoning by steering the model along this direction. Building on this insight, we introduce \textbf{\textit{ThinkEdit}}, a simple yet effective weight-editing approach to mitigate the issue of overly short reasoning. We first identify a small subset of attention heads (approximately 4%) that predominantly drive short reasoning behavior. We then edit the output projection weights of these heads to remove the short reasoning direction. With changes to only 0.2% of the model's parameters, \textbf{\textit{ThinkEdit}} effectively reduces overly short reasoning and yields notable accuracy gains for short reasoning outputs (+6.39%), along with an overall improvement across multiple math benchmarks (+3.34%). Our findings provide new mechanistic insights into how reasoning length is controlled within LLMs and highlight the potential of fine-grained model interventions to improve reasoning quality. Our code is available at: https://github.com/Trustworthy-ML-Lab/ThinkEdit\
Authors: Parshin Shojaee, Ngoc-Hieu Nguyen, Kazem Meidani, Amir Barati Farimani, Khoa D Doan, Chandan K Reddy
Abstract: Scientific equation discovery is a fundamental task in the history of scientific progress, enabling the derivation of laws governing natural phenomena. Recently, Large Language Models (LLMs) have gained interest for this task due to their potential to leverage embedded scientific knowledge for hypothesis generation. However, evaluating the true discovery capabilities of these methods remains challenging, as existing benchmarks often rely on common equations that are susceptible to memorization by LLMs, leading to inflated performance metrics that do not reflect discovery. In this paper, we introduce LLM-SRBench, a comprehensive benchmark with 239 challenging problems across four scientific domains specifically designed to evaluate LLM-based scientific equation discovery methods while preventing trivial memorization. Our benchmark comprises two main categories: LSR-Transform, which transforms common physical models into less common mathematical representations to test reasoning beyond memorized forms, and LSR-Synth, which introduces synthetic, discovery-driven problems requiring data-driven reasoning. Through extensive evaluation of several state-of-the-art methods, using both open and closed LLMs, we find that the best-performing system so far achieves only 31.5% symbolic accuracy. These findings highlight the challenges of scientific equation discovery, positioning LLM-SRBench as a valuable resource for future research.
Authors: Kun Wang, Guibin Zhang, Zhenhong Zhou, Jiahao Wu, Miao Yu, Shiqian Zhao, Chenlong Yin, Jinhu Fu, Yibo Yan, Hanjun Luo, Liang Lin, Zhihao Xu, Haolang Lu, Xinye Cao, Xinyun Zhou, Weifei Jin, Fanci Meng, Shicheng Xu, Junyuan Mao, Yu Wang, Hao Wu, Minghe Wang, Fan Zhang, Junfeng Fang, Wenjie Qu, Yue Liu, Chengwei Liu, Yifan Zhang, Qiankun Li, Chongye Guo, Yalan Qin, Zhaoxin Fan, Kai Wang, Yi Ding, Donghai Hong, Jiaming Ji, Yingxin Lai, Zitong Yu, Xinfeng Li, Yifan Jiang, Yanhui Li, Xinyu Deng, Junlin Wu, Dongxia Wang, Yihao Huang, Yufei Guo, Jen-tse Huang, Qiufeng Wang, Xiaolong Jin, Wenxuan Wang, Dongrui Liu, Yanwei Yue, Wenke Huang, Guancheng Wan, Heng Chang, Tianlin Li, Yi Yu, Chenghao Li, Jiawei Li, Lei Bai, Jie Zhang, Qing Guo, Jingyi Wang, Tianlong Chen, Joey Tianyi Zhou, Xiaojun Jia, Weisong Sun, Cong Wu, Jing Chen, Xuming Hu, Yiming Li, Xiao Wang, Ningyu Zhang, Luu Anh Tuan, Guowen Xu, Jiaheng Zhang, Tianwei Zhang, Xingjun Ma, Jindong Gu, Liang Pang, Xiang Wang, Bo An, Jun Sun, Mohit Bansal, Shirui Pan, Lingjuan Lyu, Yuval Elovici, Bhavya Kailkhura, Yaodong Yang, Hongwei Li, Wenyuan Xu, Yizhou Sun, Wei Wang, Qing Li, Ke Tang, Yu-Gang Jiang, Felix Juefei-Xu, Hui Xiong, Xiaofeng Wang, Dacheng Tao, Philip S. Yu, Qingsong Wen, Yang Liu
Abstract: The remarkable success of Large Language Models (LLMs) has illuminated a promising pathway toward achieving Artificial General Intelligence for both academic and industrial communities, owing to their unprecedented performance across various applications. As LLMs continue to gain prominence in both research and commercial domains, their security and safety implications have become a growing concern, not only for researchers and corporations but also for every nation. Currently, existing surveys on LLM safety primarily focus on specific stages of the LLM lifecycle, e.g., deployment phase or fine-tuning phase, lacking a comprehensive understanding of the entire "lifechain" of LLMs. To address this gap, this paper introduces, for the first time, the concept of "full-stack" safety to systematically consider safety issues throughout the entire process of LLM training, deployment, and eventual commercialization. Compared to the off-the-shelf LLM safety surveys, our work demonstrates several distinctive advantages: (I) Comprehensive Perspective. We define the complete LLM lifecycle as encompassing data preparation, pre-training, post-training, deployment and final commercialization. To our knowledge, this represents the first safety survey to encompass the entire lifecycle of LLMs. (II) Extensive Literature Support. Our research is grounded in an exhaustive review of over 800+ papers, ensuring comprehensive coverage and systematic organization of security issues within a more holistic understanding. (III) Unique Insights. Through systematic literature analysis, we have developed reliable roadmaps and perspectives for each chapter. Our work identifies promising research directions, including safety in data generation, alignment techniques, model editing, and LLM-based agent systems. These insights provide valuable guidance for researchers pursuing future work in this field.
Authors: Junfan Xia, Bin Jiang
Abstract: Machine learning potentials have achieved great success in accelerating atomistic simulations. Many of them relying on atom-centered local descriptors are natural for parallelization. More recent message passing neural network (MPNN) models have demonstrated their superior accuracy and become increasingly popular. However, efficiently parallelizing MPNN models across multiple nodes remains challenging, limiting their practical applications in large-scale simulations. Here, we propose an efficient parallel algorithm for MPNN models, in which additional data communication is minimized among local atoms only in each MP layer without redundant computation, thus scaling linearly with the layer number. Integrated with our recursively embedded atom neural network model, this algorithm demonstrates excellent strong scaling and weak scaling behaviors in several benchmark systems. This approach enables massive molecular dynamics simulations on MPNN models as fast as on strictly local models for over 100 million atoms, vastly extending the applicability of the MPNN potential to an unprecedented scale. This general parallelization framework can empower various MPNN models to efficiently simulate very large and complex systems.
Authors: Hugo Gobato Souto, Francisco Louzada Neto
Abstract: This paper introduces the Difference-in-Differences Bayesian Causal Forest (DiD-BCF), a novel non-parametric model addressing key challenges in DiD estimation, such as staggered adoption and heterogeneous treatment effects. DiD-BCF provides a unified framework for estimating Average (ATE), Group-Average (GATE), and Conditional Average Treatment Effects (CATE). A core innovation, its Parallel Trends Assumption (PTA)-based reparameterization, enhances estimation accuracy and stability in complex panel data settings. Extensive simulations demonstrate DiD-BCF's superior performance over established benchmarks, particularly under non-linearity, selection biases, and effect heterogeneity. Applied to U.S. minimum wage policy, the model uncovers significant conditional treatment effect heterogeneity related to county population, insights obscured by traditional methods. DiD-BCF offers a robust and versatile tool for more nuanced causal inference in modern DiD applications.
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: Yapei Chang, Yekyung Kim, Michael Krumdick, Amir Zadeh, Chuan Li, Chris Tanner, Mohit Iyyer
Abstract: Reward models are central to aligning LLMs with human preferences, but they are costly to train, requiring large-scale human-labeled preference data and powerful pretrained LLM backbones. Meanwhile, the increasing availability of high-quality synthetic instruction-following datasets raises the question: can simpler, reference-based metrics serve as viable alternatives to reward models during RL-based alignment? In this paper, we show first that BLEU, a basic string-matching metric, surprisingly matches strong reward models in agreement with human preferences on general instruction-following datasets. Based on this insight, we develop BLEUBERI, a method that first identifies challenging instructions and then applies Group Relative Policy Optimization (GRPO) using BLEU directly as the reward function. We demonstrate that BLEUBERI-trained models are competitive with models trained via reward model-guided RL across four challenging instruction-following benchmarks and three different base language models. A human evaluation further supports that the quality of BLEUBERI model outputs is on par with those from reward model-aligned models. Moreover, BLEUBERI models generate outputs that are more factually grounded than competing methods. Overall, we show that given access to high-quality reference outputs (easily obtained via existing instruction-following datasets or synthetic data generation), string matching-based metrics are cheap yet effective proxies for reward models during alignment. We release our code and data at https://github.com/lilakk/BLEUBERI.
Authors: Jiahao Yu, Qizhen Ying, Leyang Wang, Ziyue Jiang, Song Liu
Abstract: This paper introduces a novel iterative method for missing data imputation that sequentially reduces the mutual information between data and their corresponding missing mask. Inspired by GAN-based approaches, which train generators to decrease the predictability of missingness patterns, our method explicitly targets the reduction of mutual information. Specifically, our algorithm iteratively minimizes the KL divergence between the joint distribution of the imputed data and missing mask, and the product of their marginals from the previous iteration. We show that the optimal imputation under this framework corresponds to solving an ODE, whose velocity field minimizes a rectified flow training objective. We further illustrate that some existing imputation techniques can be interpreted as approximate special cases of our mutual-information-reducing framework. Comprehensive experiments on synthetic and real-world datasets validate the efficacy of our proposed approach, demonstrating superior imputation performance.
Authors: Yan Wang, Ling Ding, Tien N Nguyen, Shaohua Wang, Yanan Zheng
Abstract: Large Language Models for code often entail significant computational complexity, which grows significantly with the length of the input code sequence. We propose LeanCode for code simplification to reduce training and prediction time, leveraging code contexts in utilizing attention scores to represent the tokens' importance. We advocate for the selective removal of tokens based on the average context-aware attention scores rather than average scores across all inputs. LeanCode uses the attention scores of `CLS' tokens within the encoder for classification tasks, such as code search. It also employs the encoder-decoder attention scores to determine token significance for sequence-to-sequence tasks like code summarization. Our evaluation shows LeanCode's superiority over the SOTAs DietCode and Slimcode, with improvements of 60% and 16% for code search, and 29% and 27% for code summarization, respectively.
Authors: Feilong Tang, Chengzhi Liu, Zhongxing Xu, Ming Hu, Zelin Peng, Zhiwei Yang, Jionglong Su, Minquan Lin, Yifan Peng, Xuelian Cheng, Imran Razzak, Zongyuan Ge
Abstract: Recent advancements in multimodal large language models (MLLMs) have significantly improved performance in visual question answering. However, they often suffer from hallucinations. In this work, hallucinations are categorized into two main types: initial hallucinations and snowball hallucinations. We argue that adequate contextual information can be extracted directly from the token interaction process. Inspired by causal inference in the decoding strategy, we propose to leverage causal masks to establish information propagation between multimodal tokens. The hypothesis is that insufficient interaction between those tokens may lead the model to rely on outlier tokens, overlooking dense and rich contextual cues. Therefore, we propose to intervene in the propagation process by tackling outlier tokens to enhance in-context inference. With this goal, we present FarSight, a versatile plug-and-play decoding strategy to reduce attention interference from outlier tokens merely by optimizing the causal mask. The heart of our method is effective token propagation. We design an attention register structure within the upper triangular matrix of the causal mask, dynamically allocating attention to capture attention diverted to outlier tokens. Moreover, a positional awareness encoding method with a diminishing masking rate is proposed, allowing the model to attend to further preceding tokens, especially for video sequence tasks. With extensive experiments, FarSight demonstrates significant hallucination-mitigating performance across different MLLMs on both image and video benchmarks, proving its effectiveness.
Authors: Jintian Shao
Abstract: During the finetuning stage of text generation tasks, standard cross-entropy loss treats all tokens equally. This can lead models to overemphasize high-frequency, low-information tokens, neglecting lower-frequency tokens crucial for specificity and informativeness in generated content. This paper introduces a novel loss function, Power-Law Decay Loss (PDL), specifically designed to optimize the finetuning process for text generation. The core motivation for PDL stems from observations in information theory and linguistics: the informativeness of a token is often inversely proportional to its frequency of occurrence. PDL re-weights the contribution of each token in the standard cross-entropy loss based on its frequency in the training corpus, following a power-law decay. Specifically, the weights for high-frequency tokens are reduced, while low-frequency, information-dense tokens are assigned higher weights. This mechanism guides the model during finetuning to focus more on learning and generating tokens that convey specific and unique information, thereby enhancing the quality, diversity, and informativeness of the generated text. We theoretically elaborate on the motivation and construction of PDL and discuss its potential applications and advantages across various text generation finetuning tasks, such as abstractive summarization, dialogue systems, and style transfer.
Authors: Pradyumna Shyama Prasad, Minh Nhat Nguyen
Abstract: Can LLMs accurately adjust their confidence when facing opposition? Building on previous studies measuring calibration on static fact-based question-answering tasks, we evaluate Large Language Models (LLMs) in a dynamic, adversarial debate setting, uniquely combining two realistic factors: (a) a multi-turn format requiring models to update beliefs as new information emerges, and (b) a zero-sum structure to control for task-related uncertainty, since mutual high-confidence claims imply systematic overconfidence. We organized 60 three-round policy debates among ten state-of-the-art LLMs, with models privately rating their confidence (0-100) in winning after each round. We observed five concerning patterns: (1) Systematic overconfidence: models began debates with average initial confidence of 72.9% vs. a rational 50% baseline. (2) Confidence escalation: rather than reducing confidence as debates progressed, debaters increased their win probabilities, averaging 83% by the final round. (3) Mutual overestimation: in 61.7% of debates, both sides simultaneously claimed >=75% probability of victory, a logical impossibility. (4) Persistent self-debate bias: models debating identical copies increased confidence from 64.1% to 75.2%; even when explicitly informed their chance of winning was exactly 50%, confidence still rose (from 50.0% to 57.1%). (5) Misaligned private reasoning: models' private scratchpad thoughts sometimes differed from their public confidence ratings, raising concerns about faithfulness of chain-of-thought reasoning. These results suggest LLMs lack the ability to accurately self-assess or update their beliefs in dynamic, multi-turn tasks; a major concern as LLMs are now increasingly deployed without careful review in assistant and agentic roles. Code for our experiments is available at https://github.com/pradyuprasad/llms_overconfidence
Authors: Javier Mar\'in
Abstract: We present Adjacent Possible Exploration (APE), a selective fine-tuning method for adapting large language models that systematically explores parameter modifications while maintaining model stability. Inspired by evolutionary optimization principles, APE evaluates multiple candidate parameter updates through fine-tuning on small data subsets and accepts only those exceeding a performance threshold. Unlike standard fine-tuning that follows single gradient directions, APE implements a filtered selection process that prevents destabilizing parameter changes while enabling systematic improvement. Our method achieves 33.9\% BLEU improvement and 36.2\% perplexity reduction on news summarization tasks while using minimal computational resources. The approach provides a practical framework for controlled model adaptation that balances performance gains with representational stability.
Authors: Jian Wang, Boyan Zhu, Chak Tou Leong, Yongqi Li, Wenjie Li
Abstract: Large reasoning models (LRMs) have exhibited the capacity of enhancing reasoning performance via internal test-time scaling. Building upon this, a promising direction is to further scale test-time compute to unlock even greater reasoning capabilities. However, as we push these scaling boundaries, systematically understanding the practical limits and achieving optimal resource allocation becomes a critical challenge. In this paper, we investigate the scaling plateau of test-time scaling and introduce the Test-Time Scaling Performance Model (TTSPM). We theoretically analyze two fundamental paradigms for such extended scaling, parallel scaling and sequential scaling, from a probabilistic modeling perspective. Our primary contribution is the derivation of the saturation point on the scaling budget for both strategies, identifying thresholds beyond which additional computation yields diminishing returns. Remarkably, despite their distinct mechanisms, both paradigms converge to a unified mathematical structure in their upper bounds. We empirically validate our theoretical findings on challenging reasoning benchmarks, including AIME, MATH-500, and GPQA, demonstrating the practical utility of these bounds for test-time resource allocation. We hope that this work provides insights into the cost-benefit trade-offs of test-time scaling, guiding the development of more resource-efficient inference strategies for large reasoning models.
Authors: Charles McGhee, Mark J. F. Gales, Kate M. Knill
Abstract: Acoustic-to-Articulatory Inversion (AAI) attempts to model the inverse mapping from speech to articulation. Exact articulatory prediction from speech alone may be impossible, as speakers can choose different forms of articulation seemingly without reference to their vocal tract structure. However, once a speaker has selected an articulatory form, their productions vary minimally. Recent works in AAI have proposed adapting Self-Supervised Learning (SSL) models to single-speaker datasets, claiming that these single-speaker models provide a universal articulatory template. In this paper, we investigate whether SSL-adapted models trained on single and multi-speaker data produce articulatory targets which are consistent across speaker identities for English and Russian. We do this through the use of a novel evaluation method which extracts articulatory targets using minimal pair sets. We also present a training method which can improve interspeaker consistency using only speech data.
Authors: H. L. Dao
Abstract: In this work, we introduce the first type of non-Euclidean neural quantum state (NQS) ansatz, in the form of the hyperbolic GRU (a variant of recurrent neural networks (RNNs)), to be used in the Variational Monte Carlo method of approximating the ground state energy for quantum many-body systems. In particular, we examine the performances of NQS ansatzes constructed from both conventional or Euclidean RNN/GRU and from hyperbolic GRU in the prototypical settings of the one- and two-dimensional transverse field Ising models (TFIM) and the one-dimensional Heisenberg $J_1J_2$ and $J_1J_2J_3$ systems. By virtue of the fact that, for all of the experiments performed in this work, hyperbolic GRU can yield performances comparable to or better than Euclidean RNNs, which have been extensively studied in these settings in the literature, our work is a proof-of-concept for the viability of hyperbolic GRU as the first type of non-Euclidean NQS ansatz for quantum many-body systems. Furthermore, in settings where the Hamiltonian displays a clear hierarchical interaction structure, such as the 1D Heisenberg $J_1J_2$ & $J_1J_2J_3$ systems with the 1st, 2nd and even 3rd nearest neighbor interactions, our results show that hyperbolic GRU definitively outperforms its Euclidean version in all instances. The fact that these results are reminiscent of the established ones from natural language processing where hyperbolic GRU almost always outperforms Euclidean RNNs when the training data exhibit a tree-like or hierarchical structure leads us to hypothesize that hyperbolic GRU NQS ansatz would likely outperform Euclidean RNN/GRU NQS ansatz in quantum spin systems that involve different degrees of nearest neighbor interactions. Finally, with this work, we hope to initiate future studies of other types of non-Euclidean NQS beyond hyperbolic GRU.
Authors: Agnideep Aich, Ashit Baran Aich, Bruce Wade
Abstract: Traditional classifiers often assume feature independence or rely on overly simplistic relationships, leading to poor performance in settings where real-world dependencies matter. We introduce the Deep Copula Classifier (DCC), a generative model that separates the learning of each feature's marginal distribution from the modeling of their joint dependence structure via neural network-parameterized copulas. For each class, lightweight neural networks are used to flexibly and adaptively capture feature interactions, making DCC particularly effective when classification is driven by complex dependencies. We establish that DCC converges to the Bayes-optimal classifier under standard conditions and provide explicit convergence rates of O(n^{-r/(2r + d)}) for r-smooth copula densities. Beyond theoretical guarantees, we outline several practical extensions, including high-dimensional scalability through vine and factor copula architectures, semi-supervised learning via entropy regularization, and online adaptation using streaming gradient methods. By unifying statistical rigor with the representational power of neural networks, DCC offers a mathematically grounded and interpretable framework for dependency-aware classification.
Authors: Hidetaka Kamigaito, Ying Zhang, Jingun Kwon, Katsuhiko Hayashi, Manabu Okumura, Taro Watanabe
Abstract: Transformers deliver outstanding performance across a wide range of tasks and are now a dominant backbone architecture for large language models (LLMs). Their task-solving performance is improved by increasing parameter size, as shown in the recent studies on parameter scaling laws. Although recent mechanistic-interpretability studies have deepened our understanding of the internal behavior of Transformers by analyzing their residual stream, the relationship between these internal mechanisms and the parameter scaling laws remains unclear. To bridge this gap, we focus on layers and their size, which mainly decide the parameter size of Transformers. For this purpose, we first theoretically investigate the layers within the residual stream through a bias-diversity decomposition. The decomposition separates (i) bias, the error of each layer's output from the ground truth, and (ii) diversity, which indicates how much the outputs of each layer differ from each other. Analyzing Transformers under this theory reveals that performance improves when individual layers make predictions close to the correct answer and remain mutually diverse. We show that diversity becomes especially critical when individual layers' outputs are far from the ground truth. Finally, we introduce an information-theoretic diversity and show our main findings that adding layers enhances performance only when those layers behave differently, i.e., are diverse. We also reveal the performance gains from increasing the number of layers exhibit submodularity: marginal improvements diminish as additional layers increase, mirroring the logarithmic convergence predicted by the parameter scaling laws. Experiments on multiple semantic-understanding tasks with various LLMs empirically confirm the theoretical properties derived in this study.
Authors: Yi Gu
Abstract: T-distributed stochastic neighbor embedding (t-SNE) is a well-known algorithm for visualizing high-dimensional data by finding low-dimensional representations. In this paper, we study the convergence of t-SNE with generalized kernels and extend the results of Auffinger and Fletcher in 2023. Our work starts by giving a concrete formulation of generalized input and output kernels. Then we prove that under certain conditions, the t-SNE algorithm converges to an equilibrium distribution for a wide range of input and output kernels as the number of data points diverges.
Authors: Tushar Gautam, Robert M. Kirby, Jacob Hochhalter, Shandian Zhe
Abstract: Fatigue-induced crack growth is a leading cause of structural failure across critical industries such as aerospace, civil engineering, automotive, and energy. Accurate prediction of stress intensity factors (SIFs) -- the key parameters governing crack propagation in linear elastic fracture mechanics -- is essential for assessing fatigue life and ensuring structural integrity. While machine learning (ML) has shown great promise in SIF prediction, its advancement has been severely limited by the lack of rich, transparent, well-organized, and high-quality datasets. To address this gap, we introduce SIFBench, an open-source, large-scale benchmark database designed to support ML-based SIF prediction. SIFBench contains over 5 million different crack and component geometries derived from high-fidelity finite element simulations across 37 distinct scenarios, and provides a unified Python interface for seamless data access and customization. We report baseline results using a range of popular ML models -- including random forests, support vector machines, feedforward neural networks, and Fourier neural operators -- alongside comprehensive evaluation metrics and template code for model training, validation, and assessment. By offering a standardized and scalable resource, SIFBench substantially lowers the entry barrier and fosters the development and application of ML methods in damage tolerance design and predictive maintenance.
Authors: Minjun Zhu, Qiujie Xie, Yixuan Weng, Jian Wu, Zhen Lin, Linyi Yang, Yue Zhang
Abstract: The emergence of Artificial Intelligence (AI) Scientist represents a paradigm shift in scientific discovery, with large language models (LLMs) taking the lead as the primary executor in the entire scientific workflow from idea generation to experiment implementation. Recent AI Scientist studies demonstrate sufficient capabilities for independent scientific discovery, with the generated research reports gaining acceptance at the ICLR 2025 workshop and ACL 2025, arguing that a human-level AI Scientist, capable of uncovering phenomena previously unknown to humans, may be imminent. Despite this substantial progress, AI Scientist has yet to produce a groundbreaking achievement in the domain of computer science on par with automated scientific tools. Based on extensive quantitative evidence from existing benchmarks in complex engineering tasks and a systematic evaluation assess 28 research papers generated by five advanced AI Scientist systems, we argue that \textbf{the fundamental bottleneck for AI Scientists lies in their capability to execute the requisite verification procedures.} Current AI Scientist systems lack the execution capabilities needed to execute rigorous experiments and produce high-quality scientific papers. To better illustrate the root cause of this \textbf{implementation gap}, we provide an in-depth discussion on the fundamental limitations of AI Scientist. This position paper aims to call for the participants in the community to bridge the implementation gap.
Authors: Shuo Yan, Yuliang Yan, Bin Ma, Chenao Li, Haochun Tang, Jiahua Lu, Minhua Lin, Yuyuan Feng, Hui Xiong, Enyan Dai
Abstract: Recently, extensive deep learning architectures and pretraining strategies have been explored to support downstream protein applications. Additionally, domain-specific models incorporating biological knowledge have been developed to enhance performance in specialized tasks. In this work, we introduce $\textbf{Protap}$, a comprehensive benchmark that systematically compares backbone architectures, pretraining strategies, and domain-specific models across diverse and realistic downstream protein applications. Specifically, Protap covers five applications: three general tasks and two novel specialized tasks, i.e., enzyme-catalyzed protein cleavage site prediction and targeted protein degradation, which are industrially relevant yet missing from existing benchmarks. For each application, Protap compares various domain-specific models and general architectures under multiple pretraining settings. Our empirical studies imply that: (i) Though large-scale pretraining encoders achieve great results, they often underperform supervised encoders trained on small downstream training sets. (ii) Incorporating structural information during downstream fine-tuning can match or even outperform protein language models pretrained on large-scale sequence corpora. (iii) Domain-specific biological priors can enhance performance on specialized downstream tasks. Code and datasets are publicly available at https://github.com/Trust-App-AI-Lab/protap.
Authors: Tom Danino, Nahum Shimkin
Abstract: Multi-agent reinforcement learning (MARL) methods have achieved state-of-the-art results on a range of multi-agent tasks. Yet, MARL algorithms typically require significantly more environment interactions than their single-agent counterparts to converge, a problem exacerbated by the difficulty in exploring over a large joint action space and the high variance intrinsic to MARL environments. To tackle these issues, we propose a novel algorithm that combines a decomposed centralized critic with decentralized ensemble learning, incorporating several key contributions. The main component in our scheme is a selective exploration method that leverages ensemble kurtosis. We extend the global decomposed critic with a diversity-regularized ensemble of individual critics and utilize its excess kurtosis to guide exploration toward high-uncertainty states and actions. To improve sample efficiency, we train the centralized critic with a novel truncated variation of the TD($\lambda$) algorithm, enabling efficient off-policy learning with reduced variance. On the actor side, our suggested algorithm adapts the mixed samples approach to MARL, mixing on-policy and off-policy loss functions for training the actors. This approach balances between stability and efficiency and outperforms purely off-policy learning. The evaluation shows our method outperforms state-of-the-art baselines on standard MARL benchmarks, including a variety of SMAC II maps.
Authors: Hasan Fallahgoul
Abstract: Recent advances in machine learning have shown promising results for financial prediction using large, over-parameterized models. This paper provides theoretical foundations and empirical validation for understanding when and how these methods achieve predictive success. I examine two key aspects of high-dimensional learning in finance. First, I prove that within-sample standardization in Random Fourier Features implementations fundamentally alters the underlying Gaussian kernel approximation, replacing shift-invariant kernels with training-set dependent alternatives. Second, I establish information-theoretic lower bounds that identify when reliable learning is impossible no matter how sophisticated the estimator. A detailed quantitative calibration of the polynomial lower bound shows that with typical parameter choices, e.g., 12,000 features, 12 monthly observations, and R-square 2-3%, the required sample size to escape the bound exceeds 25-30 years of data--well beyond any rolling-window actually used. Thus, observed out-of-sample success must originate from lower-complexity artefacts rather than from the intended high-dimensional mechanism.
Authors: Hieu Trung Nguyen, Bao Nguyen, Viet Anh Nguyen
Abstract: Model pruning in transformer-based language models, traditionally viewed as a means of achieving computational savings, can enhance the model's reasoning capabilities. In this work, we uncover a surprising phenomenon: the selective pruning of certain attention heads leads to improvements in reasoning performance, particularly on challenging tasks. Motivated by this observation, we propose SPRINT, a novel contrastive learning framework that dynamically selects the optimal head and layer to prune during inference. By aligning question embeddings with head embeddings, SPRINT identifies those pruned-head configurations that result in more accurate reasoning. Extensive experiments demonstrate that our method significantly outperforms traditional best-of-$N$ and random head selection strategies on the MATH500 and GSM8K datasets.
Authors: Hicham Eddoubi, Jonas Ricker, Federico Cocchi, Lorenzo Baraldi, Angelo Sotgiu, Maura Pintor, Marcella Cornia, Lorenzo Baraldi, Asja Fischer, Rita Cucchiara, Battista Biggio
Abstract: AI-generated images have reached a quality level at which humans are incapable of reliably distinguishing them from real images. To counteract the inherent risk of fraud and disinformation, the detection of AI-generated images is a pressing challenge and an active research topic. While many of the presented methods claim to achieve high detection accuracy, they are usually evaluated under idealized conditions. In particular, the adversarial robustness is often neglected, potentially due to a lack of awareness or the substantial effort required to conduct a comprehensive robustness analysis. In this work, we tackle this problem by providing a simpler means to assess the robustness of AI-generated image detectors. We present RAID (Robust evaluation of AI-generated image Detectors), a dataset of 72k diverse and highly transferable adversarial examples. The dataset is created by running attacks against an ensemble of seven state-of-the-art detectors and images generated by four different text-to-image models. Extensive experiments show that our methodology generates adversarial images that transfer with a high success rate to unseen detectors, which can be used to quickly provide an approximate yet still reliable estimate of a detector's adversarial robustness. Our findings indicate that current state-of-the-art AI-generated image detectors can be easily deceived by adversarial examples, highlighting the critical need for the development of more robust methods. We release our dataset at https://huggingface.co/datasets/aimagelab/RAID and evaluation code at https://github.com/pralab/RAID.
URLs: https://huggingface.co/datasets/aimagelab/RAID, https://github.com/pralab/RAID.
Authors: Jiaheng Hu, Peter Stone, Roberto Mart\'in-Mart\'in
Abstract: Building capable household and industrial robots requires mastering the control of versatile, high-degree-of-freedom (DoF) systems such as mobile manipulators. While reinforcement learning (RL) holds promise for autonomously acquiring robot control policies, scaling it to high-DoF embodiments remains challenging. Direct RL in the real world demands both safe exploration and high sample efficiency, which are difficult to achieve in practice. Sim-to-real RL, on the other hand, is often brittle due to the reality gap. This paper introduces SLAC, a method that renders real-world RL feasible for complex embodiments by leveraging a low-fidelity simulator to pretrain a task-agnostic latent action space. SLAC trains this latent action space via a customized unsupervised skill discovery method designed to promote temporal abstraction, disentanglement, and safety, thereby facilitating efficient downstream learning. Once a latent action space is learned, SLAC uses it as the action interface for a novel off-policy RL algorithm to autonomously learn downstream tasks through real-world interactions. We evaluate SLAC against existing methods on a suite of bimanual mobile manipulation tasks, where it achieves state-of-the-art performance. Notably, SLAC learns contact-rich whole-body tasks in under an hour of real-world interactions, without relying on any demonstrations or hand-crafted behavior priors. More information, code, and videos at robo-rl.github.io
Authors: Muhammed Golec, Maha AlabdulJalil
Abstract: Large Language Models (LLM), which have developed in recent years, enable credit risk assessment through the analysis of financial texts such as analyst reports and corporate disclosures. This paper presents the first systematic review and taxonomy focusing on LLMbased approaches in credit risk estimation. We determined the basic model architectures by selecting 60 relevant papers published between 2020-2025 with the PRISMA research strategy. And we examined the data used for scenarios such as credit default prediction and risk analysis. Since the main focus of the paper is interpretability, we classify concepts such as explainability mechanisms, chain of thought prompts and natural language justifications for LLM-based credit models. The taxonomy organizes the literature under four main headings: model architectures, data types, explainability mechanisms and application areas. Based on this analysis, we highlight the main future trends and research gaps for LLM-based credit scoring systems. This paper aims to be a reference paper for artificial intelligence and financial researchers.
Authors: Elmira Mirzabeigi, Rezvan Salehi, Kourosh Parand
Abstract: BridgeNet is a novel hybrid framework that integrates convolutional neural networks with physics-informed neural networks to efficiently solve non-linear, high-dimensional Fokker-Planck equations (FPEs). Traditional PINNs, which typically rely on fully connected architectures, often struggle to capture complex spatial hierarchies and enforce intricate boundary conditions. In contrast, BridgeNet leverages adaptive CNN layers for effective local feature extraction and incorporates a dynamically weighted loss function that rigorously enforces physical constraints. Extensive numerical experiments across various test cases demonstrate that BridgeNet not only achieves significantly lower error metrics and faster convergence compared to conventional PINN approaches but also maintains robust stability in high-dimensional settings. This work represents a substantial advancement in computational physics, offering a scalable and accurate solution methodology with promising applications in fields ranging from financial mathematics to complex system dynamics.
Authors: Osayamen Jonathan Aimuyo, Byungsoo Oh, Rachee Singh
Abstract: The computational sparsity of Mixture-of-Experts (MoE) models enables sub-linear growth in compute cost as model size increases, thus offering a scalable path to training massive neural networks. However, existing implementations suffer from \emph{low GPU utilization}, \emph{significant latency overhead}, and a fundamental \emph{inability to leverage task locality}, primarily due to CPU-managed scheduling, host-initiated communication, and frequent kernel launches. To overcome these limitations, we develop FlashDMoE, a fully GPU-resident MoE operator that fuses expert computation and inter-GPU communication into a \emph{single persistent GPU kernel}. FlashDMoE enables fine-grained pipelining of dispatch, compute, and combine phases, eliminating launch overheads and reducing idle gaps. Unlike existing work, FlashDMoE obviates bulk-synchronous collectives for one-sided, device-initiated, inter-GPU (R)DMA transfers, thus unlocking \emph{payload efficiency}, where we eliminate bloated or redundant network payloads in sparsely activated layers. When evaluated on a single 8-H100 GPU node with MoE models having up to 128 experts and 16K token sequences, FlashDMoE achieves up to \textbf{9}$\times$ higher GPU utilization, \textbf{6}$\times$ lower latency, \textbf{5.7}$\times$ higher throughput, and \textbf{4}$\times$ better overlap efficiency compared to state-of-the-art baselines, despite using FP32 while baselines use FP16. FlashDMoE demonstrates that principled GPU kernel-hardware co-design is key to unlocking the performance ceiling of large-scale distributed ML workloads.
Authors: Alex Pan, Mary-Anne Williams
Abstract: The dominant evaluation of Large Language Models has centered on their ability to surface explicit facts from increasingly vast contexts. While today's best models demonstrate near-perfect recall on these tasks, this apparent success masks a fundamental failure in multi-step computation when information is embedded in a narrative. We introduce Verbose ListOps (VLO), a novel benchmark designed to isolate this failure. VLO programmatically weaves deterministic, nested computations into coherent stories, forcing models to track and update internal state rather than simply locate explicit values. Our experiments show that leading LLMs, capable of solving the raw ListOps equations with near-perfect accuracy, collapse in performance on VLO at just 10k tokens. The VLO framework is extensible to any verifiable reasoning task, providing a critical tool to move beyond simply expanding context windows and begin building models with the robust, stateful comprehension required for complex knowledge work.
Authors: Hondamunige Prasanna Silva, Federico Becattini, Lorenzo Seidenari
Abstract: Foundation models represent the most prominent and recent paradigm shift in artificial intelligence. Foundation models are large models, trained on broad data that deliver high accuracy in many downstream tasks, often without fine-tuning. For this reason, models such as CLIP , DINO or Vision Transfomers (ViT), are becoming the bedrock of many industrial AI-powered applications. However, the reliance on pre-trained foundation models also introduces significant security concerns, as these models are vulnerable to adversarial attacks. Such attacks involve deliberately crafted inputs designed to deceive AI systems, jeopardizing their reliability. This paper studies the vulnerabilities of vision foundation models, focusing specifically on CLIP and ViTs, and explores the transferability of adversarial attacks to downstream tasks. We introduce a novel attack, targeting the structure of transformer-based architectures in a task-agnostic fashion. We demonstrate the effectiveness of our attack on several downstream tasks: classification, captioning, image/text retrieval, segmentation and depth estimation. Code available at:https://github.com/HondamunigePrasannaSilva/attack-attention
URLs: https://github.com/HondamunigePrasannaSilva/attack-attention
Authors: Wenyuan Li, Shunlin Liang, Yuxiang Zhang, Liqin Liu, Keyan Chen, Yongzhe Chen, Han Ma, Jianglei Xu, Yichuan Ma, Shikang Guan, Zhenwei Shi
Abstract: Fine-grained crop type classification serves as the fundamental basis for large-scale crop mapping and plays a vital role in ensuring food security. It requires simultaneous capture of both phenological dynamics (obtained from multi-temporal satellite data like Sentinel-2) and subtle spectral variations (demanding nanometer-scale spectral resolution from hyperspectral imagery). Research combining these two modalities remains scarce currently due to challenges in hyperspectral data acquisition and crop types annotation costs. To address these issues, we construct a hierarchical hyperspectral crop dataset (H2Crop) by integrating 30m-resolution EnMAP hyperspectral data with Sentinel-2 time series. With over one million annotated field parcels organized in a four-tier crop taxonomy, H2Crop establishes a vital benchmark for fine-grained agricultural crop classification and hyperspectral image processing. We propose a dual-stream Transformer architecture that synergistically processes these modalities. It coordinates two specialized pathways: a spectral-spatial Transformer extracts fine-grained signatures from hyperspectral EnMAP data, while a temporal Swin Transformer extracts crop growth patterns from Sentinel-2 time series. The designed hierarchical classification head with hierarchical fusion then simultaneously delivers multi-level crop type classification across all taxonomic tiers. Experiments demonstrate that adding hyperspectral EnMAP data to Sentinel-2 time series yields a 4.2% average F1-scores improvement (peaking at 6.3%). Extensive comparisons also confirm our method's higher accuracy over existing deep learning approaches for crop type classification and the consistent benefits of hyperspectral data across varying temporal windows and crop change scenarios. Codes and dataset are available at https://github.com/flyakon/H2Crop.
Authors: Sabri Eyuboglu, Ryan Ehrlich, Simran Arora, Neel Guha, Dylan Zinsley, Emily Liu, Will Tennien, Atri Rudra, James Zou, Azalia Mirhoseini, Christopher Re
Abstract: Large language models are often used to answer queries grounded in large text corpora (e.g. codebases, legal documents, or chat histories) by placing the entire corpus in the context window and leveraging in-context learning (ICL). Although current models support contexts of 100K-1M tokens, this setup is costly to serve because the memory consumption of the KV cache scales with input length. We explore an alternative: training a smaller KV cache offline on each corpus. At inference time, we load this trained KV cache, which we call a Cartridge, and decode a response. Critically, the cost of training a Cartridge can be amortized across all the queries referencing the same corpus. However, we find that the naive approach of training the Cartridge with next-token prediction on the corpus is not competitive with ICL. Instead, we propose self-study, a training recipe in which we generate synthetic conversations about the corpus and train the Cartridge with a context-distillation objective. We find that Cartridges trained with self-study replicate the functionality of ICL, while being significantly cheaper to serve. On challenging long-context benchmarks, Cartridges trained with self-study match ICL performance while using 38.6x less memory and enabling 26.4x higher throughput. Self-study also extends the model's effective context length (e.g. from 128k to 484k tokens on MTOB) and surprisingly, leads to Cartridges that can be composed at inference time without retraining.