Authors: Haoyu Yang, Sanjoy Dey, Pablo Meyer
Abstract: Modeling disease progression through multiple stages is critical for clinical decision-making for chronic diseases, e.g., cancer, diabetes, chronic kidney diseases, and so on. Existing approaches often model the disease progression as a uniform trajectory pattern at the population level. However, chronic diseases are highly heterogeneous and often have multiple progression patterns depending on a patient's individual genetics and environmental effects due to lifestyles. We propose a personalized disease progression model to jointly learn the heterogeneous progression patterns and groups of genetic profiles. In particular, an end-to-end pipeline is designed to simultaneously infer the characteristics of patients from genetic markers using a variational autoencoder and how it drives the disease progressions using an RNN-based state-space model based on clinical observations. Our proposed model shows improvement on real-world and synthetic clinical data.
Authors: Hongming Zhang, Ruixin Hong, Dong Yu
Abstract: Autoregressive decoding algorithms that use only past information often cannot guarantee the best performance. Recently, people discovered that looking-ahead algorithms such as Monte Carlo Tree Search (MCTS) with external reward models (RMs) can significantly improve models' output by allowing them to think ahead and leverage future outputs and associated rewards to guide the current generation. Such techniques can help the reinforcement fine-tuning phase by sampling better trajectories and the inference phase by selecting the better output. However, their high computational cost limits their applications, especially in streaming scenarios. To address this issue, we propose equipping the policy model with token-level self-reward modeling (TRM) capability to eliminate the need for external models and extra communication. We name the new architecture as Reward Transformer. In addition, we propose a streaming-looking-ahead (SLA) algorithm to further boost search efficiency with better parallelization. Experiments show that SLA achieves an overall win rate of 79.7\% against the baseline greedy decoding algorithm on three general-domain datasets with a frozen policy model while maintaining streaming efficiency. If we combine SLA with reinforcement fine-tuning techniques such as DPO, SLA achieves an overall win rate of 89.4\%.
Authors: Xiaohang Tang, Sangwoong Yoon, Seongho Son, Huizhuo Yuan, Quanquan Gu, Ilija Bogunovic
Abstract: Self-play alignment algorithms have been developed as effective methods for fine-tuning large language models (LLMs), formulating preference optimization as a two-player game. However, the regularization with respect to the reference policy, which is crucial for mitigating over-optimization, has been insufficiently investigated in self-play alignment. In this paper, we show that our regularization method can improve the unregularized self-play significantly. To study the impact of different regularizations in self-play alignment, we propose Regularized Self-Play Policy Optimization (RSPO). This generalized framework regularizes the self-play by simply adding a chosen regularization term into the loss while maintaining provable last-iterate convergence to the Nash Equilibrium of the corresponding regularized game. Surprisingly, empirical evaluations using the Mistral-7B-Instruct base model reveal that forward KL divergence regularization reduces response length in RSPO, whereas reverse KL divergence markedly improves raw win rates. RSPO with a linear combination of forward and reverse KL divergence regularization substantially increases the length-controlled win rate in AlpacaEval-2, elevating the unregularized self-play alignment method (SPPO) from $28.53\%$ to $35.44\%$. Finally, we show that RSPO also improves the response diversity.
Authors: Chengsong Huang, Langlin Huang, Jixuan Leng, Jiacheng Liu, Jiaxin Huang
Abstract: Increasing test-time computation is a straightforward approach to enhancing the quality of responses in Large Language Models (LLMs). While Best-of-N sampling and Self-Consistency with majority voting are simple and effective, they require a fixed number of sampling responses for each query, regardless of its complexity. This could result in wasted computation for simpler questions and insufficient exploration for more challenging ones. In this work, we argue that model confidence of responses can be used for improving the efficiency of test-time scaling. Unfortunately, LLMs are known to be overconfident and provide unreliable confidence estimation. To address this limitation, we introduce Self-Calibration by distilling Self-Consistency-derived confidence into the model itself. This enables reliable confidence estimation at test time with one forward pass. We then design confidence-based efficient test-time scaling methods to handle queries of various difficulty, such as Early-Stopping for Best-of-N and Self-Consistency with calibrated confidence. Experiments on three LLMs across six datasets demonstrate the effectiveness of our approach. Specifically, applying confidence-based Early Stopping to Best-of-N improves MathQA accuracy from 81.0 to 83.6 with a sample budget of 16 responses, indicating the efficacy of confidence-based sampling strategy at inference time.
Authors: Akshay Sathiya, Rohit Pandey
Abstract: Combinatorial optimization problems are prevalent across a wide variety of domains. These problems are often nuanced, their optimal solutions might not be efficiently obtainable, and they may require lots of time and compute resources to solve (they are NP-hard). It follows that the best course of action for solving these problems is to use general optimization algorithm paradigms to quickly and easily develop algorithms that are customized to these problems and can produce good solutions in a reasonable amount of time. In this paper, we present optimizn, a Python library for developing customized optimization algorithms under general optimization algorithm paradigms (simulated annealing, branch and bound). Additionally, optimizn offers continuous training, with which users can run their algorithms on a regular cadence, retain the salient aspects of previous runs, and use them in subsequent runs to potentially produce solutions that get closer and closer to optimality. An earlier version of this paper was peer reviewed and published internally at Microsoft.
Authors: Hongyi Cai, Yuqian Fu, Hongming Fu, Bo Zhao
Abstract: Instruction tuning is crucial for optimizing Large Language Models (LLMs), yet mainstream data selection methods heavily rely on LLMs as instruction quality scorers, leading to high computational costs and reduced data diversity. To address these limitations, we propose MergeIT, a novel LLM-based Merging strategy for better Instruction Tuning that shifts the focus from selection to synthesis. MergeIT operates in two stages: first, topic-aware filtering clusters and refines the dataset, preserving diversity while eliminating redundancy without relying on LLM-based scoring. Second, LLM-based merging synthesizes semantically similar instructions into more informative and compact training data, enhancing data richness while further reducing dataset size. Experimental results demonstrate that MergeIT enables efficient, diverse, and scalable instruction selection and synthesis, establishing LLM-based merging as a promising alternative to conventional scoring-based selection methods for instruction tuning. Our source code and datasets are now available at https://github.com/XcloudFance/MergeIT
Authors: Pierre Houdouin, Manuel Ruiz, Lucas Saludjian
Abstract: With the digitalization of power grids, physical equations become insufficient to describe the network's behavior, and realistic but time-consuming simulators must be used. Numerical experiments, such as safety validation, that involve simulating a large number of scenarios become computationally intractable. A popular solution to reduce the computational burden is to learn a surrogate model of the simulator with Machine Learning (ML) and then conduct the experiment directly on the fast-to-evaluate surrogate model. Among the various ML possibilities for building surrogate models, Gaussian processes (GPs) emerged as a popular solution due to their flexibility, data efficiency, and interpretability. Their probabilistic nature enables them to provide both predictions and uncertainty quantification (UQ). This paper starts with a discussion on the interest of using GPs to approximate power grid simulators and fasten numerical experiments. Such simulators, however, often violate the GP's underlying Gaussian assumption, leading to poor approximations. To address this limitation, an approach that consists in adding an adaptive residual uncertainty term to the UQ is proposed. It enables the GP to remain accurate and reliable despite the simulator's non-Gaussian behaviors. This approach is successfully applied to the certification of the proper functioning of a congestion management controller, with over 98% of simulations avoided.
Authors: Anna Beer, Lena Krieger, Pascal Weber, Martin Ritzert, Ira Assent, Claudia Plant
Abstract: In density-based clustering, clusters are areas of high object density separated by lower object density areas. This notion supports arbitrarily shaped clusters and automatic detection of noise points that do not belong to any cluster. However, it is challenging to adequately evaluate the quality of density-based clustering results. Even though some existing cluster validity indices (CVIs) target arbitrarily shaped clusters, none of them captures the quality of the labeled noise. In this paper, we propose DISCO, a Density-based Internal Score for Clustering Outcomes, which is the first CVI that also evaluates the quality of noise labels. DISCO reliably evaluates density-based clusters of arbitrary shape by assessing compactness and separation. It also introduces a direct assessment of noise labels for any given clustering. Our experiments show that DISCO evaluates density-based clusterings more consistently than its competitors. It is additionally the first method to evaluate the complete labeling of density-based clustering methods, including noise labels.
Authors: Keqiang Yan, Xiner Li, Hongyi Ling, Kenna Ashen, Carl Edwards, Raymundo Arr\'oyave, Marinka Zitnik, Heng Ji, Xiaofeng Qian, Xiaoning Qian, Shuiwang Ji
Abstract: We consider the problem of crystal materials generation using language models (LMs). A key step is to convert 3D crystal structures into 1D sequences to be processed by LMs. Prior studies used the crystallographic information framework (CIF) file stream, which fails to ensure SE(3) and periodic invariance and may not lead to unique sequence representations for a given crystal structure. Here, we propose a novel method, known as Mat2Seq, to tackle this challenge. Mat2Seq converts 3D crystal structures into 1D sequences and ensures that different mathematical descriptions of the same crystal are represented in a single unique sequence, thereby provably achieving SE(3) and periodic invariance. Experimental results show that, with language models, Mat2Seq achieves promising performance in crystal structure generation as compared with prior methods.
Authors: Akhil Jalan, Yassir Jedra, Arya Mazumdar, Soumendu Sundar Mukherjee, Purnamrita Sarkar
Abstract: We study transfer learning for matrix completion in a Missing Not-at-Random (MNAR) setting that is motivated by biological problems. The target matrix $Q$ has entire rows and columns missing, making estimation impossible without side information. To address this, we use a noisy and incomplete source matrix $P$, which relates to $Q$ via a feature shift in latent space. We consider both the active and passive sampling of rows and columns. We establish minimax lower bounds for entrywise estimation error in each setting. Our computationally efficient estimation framework achieves this lower bound for the active setting, which leverages the source data to query the most informative rows and columns of $Q$. This avoids the need for incoherence assumptions required for rate optimality in the passive sampling setting. We demonstrate the effectiveness of our approach through comparisons with existing algorithms on real-world biological datasets.
Authors: Reza Bayat, Ali Rahimi-Kalahroudi, Mohammad Pezeshki, Sarath Chandar, Pascal Vincent
Abstract: A key challenge in AI alignment is guiding large language models (LLMs) to follow desired behaviors at test time. Activation steering, which modifies internal model activations during inference, offers a potential solution. However, prior work in dense activation spaces struggles with superposition, wherein multiple features become entangled, limiting interpretability and precise control. In contrast, sparse representations provide an untapped opportunity for more interpretable behavior modulation. In this work, we introduce sparse activation steering (SAS), a method that leverages sparse autoencoders (SAEs) to steer LLM behavior in sparse spaces. By isolating behavior-specific features through a contrastive prompt-pairing approach, we define a set of features that can selectively reinforce or suppress behaviors. Experiments on Gemma 2 LLMs show that SAS vectors enable nuanced behavioral modulation and finer-grained control. Furthermore, scaling SAEs improves monosemanticity of SAS vectors, suggesting more reliable and interpretable interventions.
Authors: Hanjiang Hu, Alexander Robey, Changliu Liu
Abstract: Large language models (LLMs) are highly vulnerable to jailbreaking attacks, wherein adversarial prompts are designed to elicit harmful responses. While existing defenses effectively mitigate single-turn attacks by detecting and filtering unsafe inputs, they fail against multi-turn jailbreaks that exploit contextual drift over multiple interactions, gradually leading LLMs away from safe behavior. To address this challenge, we propose a safety steering framework grounded in safe control theory, ensuring invariant safety in multi-turn dialogues. Our approach models the dialogue with LLMs using state-space representations and introduces a novel neural barrier function (NBF) to detect and filter harmful queries emerging from evolving contexts proactively. Our method achieves invariant safety at each turn of dialogue by learning a safety predictor that accounts for adversarial queries, preventing potential context drift toward jailbreaks. Extensive experiments under multiple LLMs show that our NBF-based safety steering outperforms safety alignment baselines, offering stronger defenses against multi-turn jailbreaks while maintaining a better trade-off between safety and helpfulness under different multi-turn jailbreak methods. Our code is available at https://github.com/HanjiangHu/NBF-LLM .
Authors: Jian Gao, Weidong Cao, Junyi Yang, Xuan Zhang
Abstract: The massive and large-scale design of foundational semiconductor integrated circuits (ICs) is crucial to sustaining the advancement of many emerging and future technologies, such as generative AI, 5G/6G, and quantum computing. Excitingly, recent studies have shown the great capabilities of foundational models in expediting the design of digital ICs. Yet, applying generative AI techniques to accelerate the design of analog ICs remains a significant challenge due to critical domain-specific issues, such as the lack of a comprehensive dataset and effective representation methods for analog circuits. This paper proposes, $\textbf{AnalogGenie}$, a $\underline{\textbf{Gen}}$erat$\underline{\textbf{i}}$ve $\underline{\textbf{e}}$ngine for automatic design/discovery of $\underline{\textbf{Analog}}$ circuit topologies--the most challenging and creative task in the conventional manual design flow of analog ICs. AnalogGenie addresses two key gaps in the field: building a foundational comprehensive dataset of analog circuit topology and developing a scalable sequence-based graph representation universal to analog circuits. Experimental results show the remarkable generation performance of AnalogGenie in broadening the variety of analog ICs, increasing the number of devices within a single design, and discovering unseen circuit topologies far beyond any prior arts. Our work paves the way to transform the longstanding time-consuming manual design flow of analog ICs to an automatic and massive manner powered by generative AI. Our source code is available at https://github.com/xz-group/AnalogGenie.
Authors: Naveen Mysore
Abstract: Reinforcement learning (RL) methods frequently assume that each new observation completely reflects the environment's state, thereby guaranteeing Markovian (one-step) transitions. In practice, partial observability or sensor/actuator noise often invalidates this assumption. This paper proposes a systematic methodology for detecting such violations, combining a partial correlation-based causal discovery process (PCMCI) with a novel Markov Violation score (MVS). The MVS measures multi-step dependencies that emerge when noise or incomplete state information disrupts the Markov property. Classic control tasks (CartPole, Pendulum, Acrobot) serve as examples to illustrate how targeted noise and dimension omissions affect both RL performance and measured Markov consistency. Surprisingly, even substantial observation noise sometimes fails to induce strong multi-lag dependencies in certain domains (e.g., Acrobot). In contrast, dimension-dropping investigations show that excluding some state variables (e.g., angular velocities in CartPole and Pendulum) significantly reduces returns and increases MVS, while removing other dimensions has minimal impact. These findings emphasize the importance of locating and safeguarding the most causally essential dimensions in order to preserve effective single-step learning. By integrating partial correlation tests with RL performance outcomes, the proposed approach precisely identifies when and where the Markov assumption is violated. This framework offers a principled mechanism for developing robust policies, informing representation learning, and addressing partial observability in real-world RL scenarios. All code and experimental logs are accessible for reproducibility (https://github.com/ucsb/markovianess).
Authors: Wenrui Fan, L. M. Riza Rizky, Jiayang Zhang, Chen Chen, Haiping Lu, Kevin Teh, Dinesh Selvarajah, Shuo Zhou
Abstract: Neuropathic pain, affecting up to 10% of adults, remains difficult to treat due to limited therapeutic efficacy and tolerability. Although resting-state functional MRI (rs-fMRI) is a promising non-invasive measurement of brain biomarkers to predict drug response in therapeutic development, the complexity of fMRI demands machine learning models with substantial capacity. However, extreme data scarcity in neuropathic pain research limits the application of high-capacity models. To address the challenge of data scarcity, we propose FMM$_{TC}$, a Foundation-Model-boosted Multimodal learning framework for fMRI-based neuropathic pain drug response prediction, which leverages both internal multimodal information in pain-specific data and external knowledge from large pain-agnostic data. Specifically, to maximize the value of limited pain-specific data, FMM$_{TC}$ integrates complementary information from two rs-fMRI modalities: Time series and functional Connectivity. FMM$_{TC}$ is further boosted by an fMRI foundation model with its external knowledge from extensive pain-agnostic fMRI datasets enriching limited pain-specific information. Evaluations with an in-house dataset and a public dataset from OpenNeuro demonstrate FMM$_{TC}$'s superior representation ability, generalizability, and cross-dataset adaptability over existing unimodal fMRI models that only consider one of the rs-fMRI modalities. The ablation study validates the effectiveness of multimodal learning and foundation-model-powered external knowledge transfer in FMM$_{TC}$. An integrated gradient-based interpretation study explains how FMM$_{TC}$'s cross-dataset dynamic behaviors enhance its adaptability. In conclusion, FMM$_{TC}$ boosts clinical trials in neuropathic pain therapeutic development by accurately predicting drug responses to improve the participant stratification efficiency.
Authors: Sharan Vaswani, Reza Babanezhad
Abstract: Armijo line-search (Armijo-LS) is a standard method to set the step-size for gradient descent (GD). For smooth functions, Armijo-LS alleviates the need to know the global smoothness constant $L$ and adapts to the local smoothness, enabling GD to converge faster. However, existing theoretical analyses of GD with Armijo-LS (GD-LS) do not characterize this fast convergence. We show that if the objective function satisfies a certain non-uniform smoothness condition, GD-LS converges provably faster than GD with a constant $1/L$ step-size (denoted as GD(1/L)). Our results imply that for convex losses corresponding to logistic regression and multi-class classification, GD-LS can converge to the optimum at a linear rate and, hence, improve over the sublinear convergence of GD(1/L). Furthermore, for non-convex losses satisfying gradient domination (for example, those corresponding to the softmax policy gradient in RL or generalized linear models with a logistic link function), GD-LS can match the fast convergence of algorithms tailored for these specific settings. Finally, we prove that under the interpolation assumption, for convex losses, stochastic GD with a stochastic line-search can match the fast convergence of GD-LS.
Authors: Lukasz Sztukiewicz, Ignacy St\k{e}pka, Micha{\l} Wili\'nski, Jerzy Stefanowski
Abstract: The widespread adoption of machine learning systems has raised critical concerns about fairness and bias, making mitigating harmful biases essential for AI development. In this paper, we investigate the relationship between fairness improvement and the removal of harmful biases in neural networks applied to computer vision tasks. First, we introduce a set of novel XAI-based metrics that analyze saliency maps to assess shifts in a model's decision-making process. Then, we demonstrate that successful debiasing methods systematically redirect model focus away from protected attributes. Additionally, we show that techniques originally developed for artifact removal can be effectively repurposed for fairness. These findings underscore the importance of ensuring that models are fair for the right reasons, contributing to the development of more ethical and trustworthy AI systems.
Authors: Marius F. R. Juston, William R. Norris, Dustin Nottage, Ahmet Soylemezoglu
Abstract: This paper discusses the weight parametrization of two standard 1-Lipschitz network structure methodologies, the Almost-Orthogonal-Layers (AOL) and the SDP-based Lipschitz Layers (SLL), and derives their impact on the initialization for deep 1-Lipschitz feedforward networks in addition to discussing underlying issues surrounding this initialization. These networks are mainly used in certifiably robust classification applications to combat adversarial attacks by limiting the effects of perturbations on the output classification result. An exact and an upper bound for the parameterized weight variance was calculated assuming a standard Normal distribution initialization; additionally, an upper bound was computed assuming a Generalized Normal Distribution, generalizing the proof for Uniform, Laplace, and Normal distribution weight initializations. It is demonstrated that the weight variance holds no bearing on the output variance distribution and that only the dimension of the weight matrices matters. Additionally, this paper demonstrates that the weight initialization always causes deep 1-Lipschitz networks to decay to zero.
Authors: Patrick Huber, Akshat Shrivastava, Ernie Chang, Chinnadhurai Sankar, Ahmed Aly, Adithya Sagar
Abstract: Sparse Mixture of Expert (MoE) models are popular foundational architectures at large scale, however, under-explored at smaller sizes. Here, we show how to enable Compact Sparse Mixture of Experts (CoSMoEs) for on-device inference. Specifically, we tackle the three main on-device dimensions: Quality, Memory and Latency. Along the quality axis, we show that in a fair evaluation (removing confounding factors) MoE architectures outperform FLOP-aligned dense models at on-device scale. We introduce weight-decomposed experts, further improving the MoE model performance. Regarding model memory and latency, we significantly improve model offloading efficiency and, in turn, reduce model inference latency.
Authors: Asghar A. Jadoon, D. Thomas Seidl, Reese E. Jones, Jan N. Fuhg
Abstract: The black-box nature of neural networks limits the ability to encode or impose specific structural relationships between inputs and outputs. While various studies have introduced architectures that ensure the network's output adheres to a particular form in relation to certain inputs, the majority of these approaches impose constraints on only a single set of inputs. This paper introduces a novel neural network architecture, termed the Input Specific Neural Network (ISNN), which extends this concept by allowing scalar-valued outputs to be subject to multiple constraints. Specifically, the ISNN can enforce convexity in some inputs, non-decreasing monotonicity combined with convexity with respect to others, and simple non-decreasing monotonicity or arbitrary relationships with additional inputs. The paper presents two distinct ISNN architectures, along with equations for the first and second derivatives of the output with respect to the inputs. These networks are broadly applicable. In this work, we restrict their usage to solving problems in computational mechanics. In particular, we show how they can be effectively applied to fitting data-driven constitutive models. We then embed our trained data-driven constitutive laws into a finite element solver where significant time savings can be achieved by using explicit manual differentiation using the derived equations as opposed to automatic differentiation. We also show how ISNNs can be used to learn structural relationships between inputs and outputs via a binary gating mechanism. Particularly, ISNNs are employed to model an anisotropic free energy potential to get the homogenized macroscopic response in a decoupled multiscale setting, where the network learns whether or not the potential should be modeled as polyconvex, and retains only the relevant layers while using the minimum number of inputs.
Authors: Jahan C. Penny-Dimri, Magdalena Bachmann, William R. Cooke, Sam Mathewlynn, Samuel Dockree, John Tolladay, Jannik Kossen, Lin Li, Yarin Gal, Gabriel Davis Jones
Abstract: Large language models (LLMs) hold substantial promise for clinical decision support. However, their widespread adoption in medicine, particularly in healthcare, is hindered by their propensity to generate false or misleading outputs, known as hallucinations. In high-stakes domains such as women's health (obstetrics & gynaecology), where errors in clinical reasoning can have profound consequences for maternal and neonatal outcomes, ensuring the reliability of AI-generated responses is critical. Traditional methods for quantifying uncertainty, such as perplexity, fail to capture meaning-level inconsistencies that lead to misinformation. Here, we evaluate semantic entropy (SE), a novel uncertainty metric that assesses meaning-level variation, to detect hallucinations in AI-generated medical content. Using a clinically validated dataset derived from UK RCOG MRCOG examinations, we compared SE with perplexity in identifying uncertain responses. SE demonstrated superior performance, achieving an AUROC of 0.76 (95% CI: 0.75-0.78), compared to 0.62 (0.60-0.65) for perplexity. Clinical expert validation further confirmed its effectiveness, with SE achieving near-perfect uncertainty discrimination (AUROC: 0.97). While semantic clustering was successful in only 30% of cases, SE remains a valuable tool for improving AI safety in women's health. These findings suggest that SE could enable more reliable AI integration into clinical practice, particularly in resource-limited settings where LLMs could augment care. This study highlights the potential of SE as a key safeguard in the responsible deployment of AI-driven tools in women's health, leading to safer and more effective digital health interventions.
Authors: Marzi Heidari, Abdullah Alchihabi, Hao Yan, Yuhong Guo
Abstract: In this work, we introduce a novel problem setup termed as Heterogeneous Semi-Supervised Learning (HSSL), which presents unique challenges by bridging the semi-supervised learning (SSL) task and the unsupervised domain adaptation (UDA) task, and expanding standard semi-supervised learning to cope with heterogeneous training data. At its core, HSSL aims to learn a prediction model using a combination of labeled and unlabeled training data drawn separately from heterogeneous domains that share a common set of semantic categories; this model is intended to differentiate the semantic categories of test instances sampled from both the labeled and unlabeled domains. In particular, the labeled and unlabeled domains have dissimilar label distributions and class feature distributions. This heterogeneity, coupled with the assorted sources of the test data, introduces significant challenges to standard SSL and UDA methods. Therefore, we propose a novel method, Unified Framework for Heterogeneous Semi-supervised Learning (Uni-HSSL), to address HSSL by directly learning a fine-grained classifier from the heterogeneous data, which adaptively handles the inter-domain heterogeneity while leveraging both the unlabeled data and the inter-domain semantic class relationships for cross-domain knowledge transfer and adaptation. We conduct comprehensive experiments and the experimental results validate the efficacy and superior performance of the proposed Uni-HSSL over state-of-the-art semi-supervised learning and unsupervised domain adaptation methods.
Authors: Junhui Shen, Aaron J. Davis, Ding Lu, Zhaojun Bai
Abstract: Principal Component Analysis (PCA) is a foundational technique in machine learning for dimensionality reduction of high-dimensional datasets. However, PCA could lead to biased outcomes that disadvantage certain subgroups of the underlying datasets. To address the bias issue, a Fair PCA (FPCA) model was introduced by Samadi et al. (2018) for equalizing the reconstruction loss between subgroups. The semidefinite relaxation (SDR) based approach proposed by Samadi et al. (2018) is computationally expensive even for suboptimal solutions. To improve efficiency, several alternative variants of the FPCA model have been developed. These variants often shift the focus away from equalizing the reconstruction loss. In this paper, we identify a hidden convexity in the FPCA model and introduce an algorithm for convex optimization via eigenvalue optimization. Our approach achieves the desired fairness in reconstruction loss without sacrificing performance. As demonstrated in real-world datasets, the proposed FPCA algorithm runs $8\times$ faster than the SDR-based algorithm, and only at most 85% slower than the standard PCA.
Authors: Chunyang Liao, Deanna Needell, Hayden Schaeffer
Abstract: Operator learning is the approximation of operators between infinite dimensional Banach spaces using machine learning approaches. While most progress in this area has been driven by variants of deep neural networks such as the Deep Operator Network and Fourier Neural Operator, the theoretical guarantees are often in the form of a universal approximation property. However, the existence theorems do not guarantee that an accurate operator network is obtainable in practice. Motivated by the recent kernel-based operator learning framework, we propose a random feature operator learning method with theoretical guarantees and error bounds. The random feature method can be viewed as a randomized approximation of a kernel method, which significantly reduces the computation requirements for training. We provide a generalization error analysis for our proposed random feature operator learning method along with comprehensive numerical results. Compared to kernel-based method and neural network methods, the proposed method can obtain similar or better test errors across benchmarks examples with significantly reduced training times. An additional advantages it that our implementation is simple and does require costly computational resources, such as GPU.
Authors: Guanghan Wang, Yair Schiff, Subham Sekhar Sahoo, Volodymyr Kuleshov
Abstract: Part of the success of diffusion models stems from their ability to perform iterative refinement, i.e., repeatedly correcting outputs during generation. However, modern masked discrete diffusion lacks this capability: when a token is generated, it cannot be updated again, even when it introduces an error. Here, we address this limitation by introducing the remasking diffusion model (ReMDM) sampler, a method that can be applied to pretrained masked diffusion models in a principled way and that is derived from a discrete diffusion model with a custom remasking backward process. Most interestingly, ReMDM endows discrete diffusion with a form of inference-time compute scaling. By increasing the number of sampling steps, ReMDM generates natural language outputs that approach the quality of autoregressive models, whereas when the computation budget is limited, ReMDM better maintains quality. ReMDM also improves sample quality of masked diffusion models for discretized images, and in scientific domains such as molecule design, ReMDM facilitates diffusion guidance and pushes the Pareto frontier of controllability relative to classical masking and uniform noise diffusion. We provide the code along with a blog post on the project page: https://remdm.github.io.
URLs: https://remdm.github.io.
Authors: Zhaoxi Jiang, Fei Wang
Abstract: Deep operator networks (DeepONets) represent a powerful class of data-driven methods for operator learning, demonstrating strong approximation capabilities for a wide range of linear and nonlinear operators. They have shown promising performance in learning operators that govern partial differential equations (PDEs), including diffusion-reaction systems and Burgers' equations. However, the accuracy of DeepONets is often constrained by computational limitations and optimization challenges inherent in training deep neural networks. Furthermore, the computational cost associated with training these networks is typically very high. To address these challenges, we leverage randomized neural networks (RaNNs), in which the parameters of the hidden layers remain fixed following random initialization. RaNNs compute the output layer parameters using the least-squares method, significantly reducing training time and mitigating optimization errors. In this work, we integrate DeepONets with RaNNs to propose RaNN-DeepONets, a hybrid architecture designed to balance accuracy and efficiency. Furthermore, to mitigate the need for extensive data preparation, we introduce the concept of physics-informed RaNN-DeepONets. Instead of relying on data generated through other time-consuming numerical methods, we incorporate PDE information directly into the training process. We evaluate the proposed model on three benchmark PDE problems: diffusion-reaction dynamics, Burgers' equation, and the Darcy flow problem. Through these tests, we assess its ability to learn nonlinear operators with varying input types. When compared to the standard DeepONet framework, RaNN-DeepONets achieves comparable accuracy while reducing computational costs by orders of magnitude. These results highlight the potential of RaNN-DeepONets as an efficient alternative for operator learning in PDE-based systems.
Authors: Ahmad Faraz Khan, Samuel Fountain, Ahmed M. Abdelmoniem, Ali R. Butt, Ali Anwar
Abstract: Federated Learning (FL) is an approach for privacy-preserving Machine Learning (ML), enabling model training across multiple clients without centralized data collection. With an aggregator server coordinating training, aggregating model updates, and storing metadata across rounds. In addition to training, a substantial part of FL systems are the non-training workloads such as scheduling, personalization, clustering, debugging, and incentivization. Most existing systems rely on the aggregator to handle non-training workloads and use cloud services for data storage. This results in high latency and increased costs as non-training workloads rely on large volumes of metadata, including weight parameters from client updates, hyperparameters, and aggregated updates across rounds, making the situation even worse. We propose FLStore, a serverless framework for efficient FL non-training workloads and storage. FLStore unifies the data and compute planes on a serverless cache, enabling locality-aware execution via tailored caching policies to reduce latency and costs. Per our evaluations, compared to cloud object store based aggregator server FLStore reduces per request average latency by 71% and costs by 92.45%, with peak improvements of 99.7% and 98.8%, respectively. Compared to an in-memory cloud cache based aggregator server, FLStore reduces average latency by 64.6% and costs by 98.83%, with peak improvements of 98.8% and 99.6%, respectively. FLStore integrates seamlessly with existing FL frameworks with minimal modifications, while also being fault-tolerant and highly scalable.
Authors: Hajar Kazemi Naeini, Roya Shomali, Abolhassan Pishahang, Hamidreza Hasanzadeh, Mahdieh Mohammadi, Saeid Asadi, Ahmad Gholizadeh Lonbar
Abstract: The advancement of smart grid technologies necessitates the integration of cutting-edge computational methods to enhance predictive energy optimization. This study proposes a multi-faceted approach by incorporating (1) Deep Reinforcement Learning (DRL) agents trained using data from Digital Twins (DTs) to optimize energy consumption in real time, (2) Physics-Informed Neural Networks (PINNs) to seamlessly embed physical laws within the optimization process, ensuring model accuracy and interpretability, and (3) Blockchain (BC) technology to facilitate secure and transparent communication across the smart grid infrastructure. The model was trained and validated using comprehensive datasets, including smart meter energy consumption data, renewable energy outputs, dynamic pricing, and user preferences collected from IoT devices. The proposed framework achieved superior predictive performance with a Mean Absolute Error (MAE) of 0.237 kWh, Root Mean Square Error (RMSE) of 0.298 kWh, and an R-squared (R2) value of 0.978, indicating a 97.8% explanation of data variance. Classification metrics further demonstrated the model's robustness, achieving 97.7% accuracy, 97.8% precision, 97.6% recall, and an F1 Score of 97.7%. Comparative analysis with traditional models like Linear Regression, Random Forest, SVM, LSTM, and XGBoost revealed the superior accuracy and real-time adaptability of the proposed method. In addition to enhancing energy efficiency, the model reduced energy costs by 35%, maintained a 96% user comfort index, and increased renewable energy utilization to 40%. This study demonstrates the transformative potential of integrating PINNs, DT, and Blockchain technologies to optimize energy consumption in smart grids, paving the way for sustainable, secure, and efficient energy management systems.
Authors: Quanyu Dai, Jiaren Xiao, Zhaocheng Du, Jieming Zhu, Chengxiao Luo, Xiao-Ming Wu, Zhenhua Dong
Abstract: In online advertising, uncertainty calibration aims to adjust a ranking model's probability predictions to better approximate the true likelihood of an event, e.g., a click or a conversion. However, existing calibration approaches may lack the ability to effectively model complex nonlinear relations, consider context features, and achieve balanced performance across different data subsets. To tackle these challenges, we introduce a novel model called Monotonic Calibration Networks, featuring three key designs: a monotonic calibration function (MCF), an order-preserving regularizer, and a field-balance regularizer. The nonlinear MCF is capable of naturally modeling and universally approximating the intricate relations between uncalibrated predictions and the posterior probabilities, thus being much more expressive than existing methods. MCF can also integrate context features using a flexible model architecture, thereby achieving context awareness. The order-preserving and field-balance regularizers promote the monotonic relationship between adjacent bins and the balanced calibration performance on data subsets, respectively. Experimental results on both public and industrial datasets demonstrate the superior performance of our method in generating well-calibrated probability predictions.
Authors: Rui Lu, Yang Yue, Andrew Zhao, Simon Du, Gao Huang
Abstract: Multitask Representation Learning (MRL) has emerged as a prevalent technique to improve sample efficiency in Reinforcement Learning (RL). Empirical studies have found that training agents on multiple tasks simultaneously within online and transfer learning environments can greatly improve efficiency. Despite its popularity, a comprehensive theoretical framework that elucidates its operational efficacy remains incomplete. Prior analyses have predominantly assumed that agents either possess a pre-known representation function or utilize functions from a linear class, where both are impractical. The complexity of real-world applications typically requires the use of sophisticated, non-linear functions such as neural networks as representation function, which are not pre-existing but must be learned. Our work tries to fill the gap by extending the analysis to \textit{unknown non-linear} representations, giving a comprehensive analysis for its mechanism in online and transfer learning setting. We consider the setting that an agent simultaneously playing $M$ contextual bandits (or MDPs), developing a shared representation function $\phi$ from a non-linear function class $\Phi$ using our novel Generalized Functional Upper Confidence Bound algorithm (GFUCB). We formally prove that this approach yields a regret upper bound that outperforms the lower bound associated with learning $M$ separate tasks, marking the first demonstration of MRL's efficacy in a general function class. This framework also explains the contribution of representations to transfer learning when faced with new, yet related tasks, and identifies key conditions for successful transfer. Empirical experiments further corroborate our theoretical findings.
Authors: Rickard Br\"annvall
Abstract: Federated learning is a distributed machine learning approach where multiple clients collaboratively train a model without sharing their local data, which contributes to preserving privacy. A challenge in federated learning is managing heterogeneous data distributions across clients, which can hinder model convergence and performance due to the need for the global model to generalize well across diverse local datasets. We propose to use local characteristic statistics, by which we mean some statistical properties calculated independently by each client using only their local training dataset. These statistics, such as means, covariances, and higher moments, are used to capture the characteristics of the local data distribution. They are not shared with other clients or a central node. During training, these local statistics help the model learn how to condition on the local data distribution, and during inference, they guide the client's predictions. Our experiments show that this approach allows for efficient handling of heterogeneous data across the federation, has favorable scaling compared to approaches that directly try to identify peer nodes that share distribution characteristics, and maintains privacy as no additional information needs to be communicated.
Authors: Renato Cordeiro de Amorim, Vladimir Makarenkov
Abstract: Clustering is a fundamental technique in machine learning and data analysis, widely used across various domains. Internal clustering validation measures, such as the Average Silhouette Width, Calinski-Harabasz, and Davies-Bouldin indices, play a crucial role in assessing clustering quality when external ground truth labels are unavailable. However, these measures can be affected by feature relevance, potentially leading to unreliable evaluations in high-dimensional or noisy data sets. In this paper, we introduce a Feature Importance Rescaling (FIR) method designed to enhance internal clustering validation by adjusting feature contributions based on their dispersion. Our method systematically attenuates noise features making clustering compactness and separation clearer, and by consequence aligning internal validation measures more closely with the ground truth. Through extensive experiments on synthetic data sets under different configurations, we demonstrate that FIR consistently improves the correlation between internal validation indices and the ground truth, particularly in settings with noisy or irrelevant features. The results show that FIR increases the robustness of clustering evaluation, reduces variability in performance across different data sets, and remains effective even when clusters exhibit significant overlap. These findings highlight the potential of FIR as a valuable enhancement for internal clustering validation, making it a practical tool for unsupervised learning tasks where labelled data is not available.
Authors: Song Xia, Yi Yu, Wenhan Yang, Meiwen Ding, Zhuo Chen, Lingyu Duan, Alex C. Kot, Xudong Jiang
Abstract: By locally encoding raw data into intermediate features, collaborative inference enables end users to leverage powerful deep learning models without exposure of sensitive raw data to cloud servers. However, recent studies have revealed that these intermediate features may not sufficiently preserve privacy, as information can be leaked and raw data can be reconstructed via model inversion attacks (MIAs). Obfuscation-based methods, such as noise corruption, adversarial representation learning, and information filters, enhance the inversion robustness by obfuscating the task-irrelevant redundancy empirically. However, methods for quantifying such redundancy remain elusive, and the explicit mathematical relation between this redundancy minimization and inversion robustness enhancement has not yet been established. To address that, this work first theoretically proves that the conditional entropy of inputs given intermediate features provides a guaranteed lower bound on the reconstruction mean square error (MSE) under any MIA. Then, we derive a differentiable and solvable measure for bounding this conditional entropy based on the Gaussian mixture estimation and propose a conditional entropy maximization (CEM) algorithm to enhance the inversion robustness. Experimental results on four datasets demonstrate the effectiveness and adaptability of our proposed CEM; without compromising feature utility and computing efficiency, plugging the proposed CEM into obfuscation-based defense mechanisms consistently boosts their inversion robustness, achieving average gains ranging from 12.9\% to 48.2\%. Code is available at \href{https://github.com/xiasong0501/CEM}{https://github.com/xiasong0501/CEM}.
URLs: https://github.com/xiasong0501/CEM, https://github.com/xiasong0501/CEM
Authors: Qihui Zhou, Peiqi Yin, Pengfei Zuo, James Cheng
Abstract: Processing long contexts has become a critical capability for modern large language models (LLMs). However, serving long-context LLMs comes with significant inference costs due to the high memory overhead of the key-value (KV) cache. Existing work leverages dynamic sparse attention algorithms (DSAes) to mitigate the KV cache overhead, but these algorithms rely on top-$k$ KV cache selection, which results in a trade-off between accuracy and efficiency. A larger $k$ improves accuracy but decreases efficiency, while a smaller $k$ boosts efficiency but compromises accuracy. To overcome this trade-off, this paper presents PSA, a $\underline{P}$rogressive $\underline{S}$parse $\underline{A}$ttention mechanism that integrates algorithmic innovations with system co-design to achieve both high inference accuracy and improved efficiency in LLM serving. The PSA algorithm adaptively adjusts the KV cache budget of different tokens and layers according to their real attention weight distributions, rather than relying on a fixed budget $k$. This enables high accuracy while minimizing KV cache usage. To further enhance execution efficiency, we introduce a pipelined iteration scheme that reduces CPU-GPU interleaving and synchronization overhead during PSA computation. Additionally, we implement unified GPU memory management that optimizes PSA's memory utilization by accounting for uneven memory requirements across different model layers. Extensive experimental results demonstrate that PSA reduces KV cache usage for attention computation by up to 2.4$\times$ and 8.8$\times$, and increases end-to-end serving throughput by up to 1.4$\times$ and 2.0$\times$, compared to state-of-the-art DSAes and systems without sparse attention, respectively.
Authors: Abdullah M. Zyarah, Alaa M. Abdul-Hadi, Dhireesha Kudithipudi
Abstract: The unprecedented dissemination of edge devices is accompanied by a growing demand for neuromorphic chips that can process time-series data natively without cloud support. Echo state network (ESN) is a class of recurrent neural networks that can be used to identify unique patterns in time-series data and predict future events. It is known for minimal computing resource requirements and fast training, owing to the use of linear optimization solely at the readout stage. In this work, a custom-design neuromorphic chip based on ESN targeting edge devices is proposed. The proposed system supports various learning mechanisms, including structural plasticity and synaptic plasticity, locally on-chip. This provides the network with an additional degree of freedom to continuously learn, adapt, and alter its structure and sparsity level, ensuring high performance and continuous stability. We demonstrate the performance of the proposed system as well as its robustness to noise against real-world time-series datasets while considering various topologies of data movement. An average accuracy of 95.95% and 85.24% are achieved on human activity recognition and prosthetic finger control, respectively. We also illustrate that the proposed system offers a throughput of 6x10^4 samples/sec with a power consumption of 47.7mW on a 65nm IBM process.
Authors: Fan Wan, Yuchen Li, Xueqi Qiu, Rui Sun, Leyuan Zhang, Xingyu Miao, Tianyu Zhang, Haoran Duan, Yang Long
Abstract: The proliferation of Internet of Things devices and advances in communication technology have unleashed an explosion of personal data, amplifying privacy concerns amid stringent regulations like GDPR and CCPA. Federated Learning offers a privacy preserving solution by enabling collaborative model training across decentralized devices without centralizing sensitive data. However, statistical heterogeneity from non-independent and identically distributed datasets and system heterogeneity due to client dropouts particularly those with monopolistic classes severely degrade the global model's performance. To address these challenges, we propose the Asynchronous Personalized Federated Learning framework, which empowers clients to develop personalized models using a server side semantic generator. This generator, trained via data free knowledge transfer under global model supervision, enhances client data diversity by producing both seen and unseen samples, the latter enabled by Zero-Shot Learning to mitigate dropout-induced data loss. To counter the risks of synthetic data impairing training, we introduce a decoupled model interpolation method, ensuring robust personalization. Extensive experiments demonstrate that AP FL significantly outperforms state of the art FL methods in tackling non-IID distributions and client dropouts, achieving superior accuracy and resilience across diverse real-world scenarios.
Authors: Jing Wang, Yu-Jie Zhang, Peng Zhao, Zhi-Hua Zhou
Abstract: We study the stochastic linear bandits with heavy-tailed noise. Two principled strategies for handling heavy-tailed noise, truncation and median-of-means, have been introduced to heavy-tailed bandits. Nonetheless, these methods rely on specific noise assumptions or bandit structures, limiting their applicability to general settings. The recent work [Huang et al.2024] develops a soft truncation method via the adaptive Huber regression to address these limitations. However, their method suffers undesired computational cost: it requires storing all historical data and performing a full pass over these data at each round. In this paper, we propose a \emph{one-pass} algorithm based on the online mirror descent framework. Our method updates using only current data at each round, reducing the per-round computational cost from $\widetilde{\mathcal{O}}(t \log T)$ to $\widetilde{\mathcal{O}}(1)$ with respect to current round $t$ and the time horizon $T$, and achieves a near-optimal and variance-aware regret of order $\widetilde{\mathcal{O}}\big(d T^{\frac{1-\epsilon}{2(1+\epsilon)}} \sqrt{\sum_{t=1}^T \nu_t^2} + d T^{\frac{1-\epsilon}{2(1+\epsilon)}}\big)$ where $d$ is the dimension and $\nu_t^{1+\epsilon}$ is the $(1+\epsilon)$-th central moment of reward at round $t$.
Authors: Magnus Cunow, Gerrit Gro{\ss}mann
Abstract: Autoencoders are effective deep learning models that can function as generative models and learn latent representations for downstream tasks. The use of graph autoencoders - with both encoder and decoder implemented as message passing networks - is intriguing due to their ability to generate permutation-invariant graph representations. However, this approach faces difficulties because decoding a graph structure from a single vector is challenging, and comparing input and output graphs requires an effective permutation-invariant similarity measure. As a result, many studies rely on approximate methods. In this work, we explore the effect of graph matching precision on the training behavior and generation capabilities of a Variational Autoencoder (VAE). Our contribution is two-fold: (1) we propose a transformer-based message passing graph decoder as an alternative to a graph neural network decoder, that is more robust and expressive by leveraging global attention mechanisms. (2) We show that the precision of graph matching has significant impact on training behavior and is essential for effective de novo (molecular) graph generation. Code is available at https://github.com/mcunow/graph-matching
Authors: Thomas Rimbot, Martin Jaggi, Luis Barba
Abstract: In this work, we investigate the application of Machine Learning techniques to sport climbing. Expanding upon previous projects, we develop a visualization tool for move sequence evaluation on a given boulder. Then, we look into move sequence prediction from simple holds sequence information using three different Transformer models. While the results are not conclusive, they are a first step in this kind of approach and lay the ground for future work.
Authors: Junqi He, Yujie Zhang, Jialu Wang, Tao Wang, Pan Zhang, Chengjie Cai, Jinxing Yang, Xiao Lin, Xiaohui Yang
Abstract: Two-dimensional (2D) materials and heterostructures exhibit unique physical properties, necessitating efficient and accurate characterization methods. Leveraging advancements in artificial intelligence, we introduce a deep learning-based method for efficiently characterizing heterostructures and 2D materials, specifically MoS2-MoSe2 lateral heterostructures and MoS2 flakes with varying shapes and thicknesses. By utilizing YOLO models, we achieve an accuracy rate of over 94.67% in identifying these materials. Additionally, we explore the application of transfer learning across different materials, which further enhances model performance. This model exhibits robust generalization and anti-interference ability, ensuring reliable results in diverse scenarios. To facilitate practical use, we have developed an application that enables real-time analysis directly from optical microscope images, making the process significantly faster and more cost-effective than traditional methods. This deep learning-driven approach represents a promising tool for the rapid and accurate characterization of 2D materials, opening new avenues for research and development in material science.
Authors: Yicong Dong, Rundong He, Guangyao Chen, Wentao Zhang, Zhongyi Han, Jieming Shi, Yilong Yin
Abstract: Graph Neural Networks (GNNs) have achieved significant success in machine learning, with wide applications in social networks, bioinformatics, knowledge graphs, and other fields. Most research assumes ideal closed-set environments. However, in real-world open-set environments, graph learning models face challenges in robustness and reliability due to unseen classes. This highlights the need for Graph Open-Set Recognition (GOSR) methods to address these issues and ensure effective GNN application in practical scenarios. Research in GOSR is in its early stages, with a lack of a comprehensive benchmark spanning diverse tasks and datasets to evaluate methods. Moreover, traditional methods, Graph Out-of-Distribution Detection (GOODD), GOSR, and Graph Anomaly Detection (GAD) have mostly evolved in isolation, with little exploration of their interconnections or potential applications to GOSR. To fill these gaps, we introduce \textbf{G-OSR}, a comprehensive benchmark for evaluating GOSR methods at both the node and graph levels, using datasets from multiple domains to ensure fair and standardized comparisons of effectiveness and efficiency across traditional, GOODD, GOSR, and GAD methods. The results offer critical insights into the generalizability and limitations of current GOSR methods and provide valuable resources for advancing research in this field through systematic analysis of diverse approaches.
Authors: Andy Gray, Alma Rahat, Tom Crick, Stephen Lindsay
Abstract: Comparative Judgement (CJ) provides an alternative assessment approach by evaluating work holistically rather than breaking it into discrete criteria. This method leverages human ability to make nuanced comparisons, yielding more reliable and valid assessments. CJ aligns with real-world evaluations, where overall quality emerges from the interplay of various elements. However, rubrics remain widely used in education, offering structured criteria for grading and detailed feedback. This creates a gap between CJ's holistic ranking and the need for criterion-based performance breakdowns. This paper addresses this gap using a Bayesian approach. We build on Bayesian CJ (BCJ) by Gray et al., which directly models preferences instead of using likelihoods over total scores, allowing for expected ranks with uncertainty estimation. Their entropy-based active learning method selects the most informative pairwise comparisons for assessors. We extend BCJ to handle multiple independent learning outcome (LO) components, defined by a rubric, enabling both holistic and component-wise predictive rankings with uncertainty estimates. Additionally, we propose a method to aggregate entropies and identify the most informative comparison for assessors. Experiments on synthetic and real data demonstrate our method's effectiveness. Finally, we address a key limitation of BCJ, which is the inability to quantify assessor agreement. We show how to derive agreement levels, enhancing transparency in assessment.
Authors: Jingchu Gai, Yiheng Du, Bohang Zhang, Haggai Maron, Liwei Wang
Abstract: Graph spectra are an important class of structural features on graphs that have shown promising results in enhancing Graph Neural Networks (GNNs). Despite their widespread practical use, the theoretical understanding of the power of spectral invariants -- particularly their contribution to GNNs -- remains incomplete. In this paper, we address this fundamental question through the lens of homomorphism expressivity, providing a comprehensive and quantitative analysis of the expressive power of spectral invariants. Specifically, we prove that spectral invariant GNNs can homomorphism-count exactly a class of specific tree-like graphs which we refer to as parallel trees. We highlight the significance of this result in various contexts, including establishing a quantitative expressiveness hierarchy across different architectural variants, offering insights into the impact of GNN depth, and understanding the subgraph counting capabilities of spectral invariant GNNs. In particular, our results significantly extend Arvind et al. (2024) and settle their open questions. Finally, we generalize our analysis to higher-order GNNs and answer an open question raised by Zhang et al. (2024).
Authors: Francesco Capuano, Davorin Peceli, Gabriele Tiboni
Abstract: High Power Laser (HPL) systems operate in the femtosecond regime--one of the shortest timescales achievable in experimental physics. HPL systems are instrumental in high-energy physics, leveraging ultra-short impulse durations to yield extremely high intensities, which are essential for both practical applications and theoretical advancements in light-matter interactions. Traditionally, the parameters regulating HPL optical performance are tuned manually by human experts, or optimized by using black-box methods that can be computationally demanding. Critically, black box methods rely on stationarity assumptions overlooking complex dynamics in high-energy physics and day-to-day changes in real-world experimental settings, and thus need to be often restarted. Deep Reinforcement Learning (DRL) offers a promising alternative by enabling sequential decision making in non-static settings. This work investigates the safe application of DRL to HPL systems, and extends the current research by (1) learning a control policy directly from images and (2) addressing the need for generalization across diverse dynamics. We evaluate our method across various configurations and observe that DRL effectively enables cross-domain adaptability, coping with dynamics' fluctuations while achieving 90% of the target intensity in test environments.
Authors: Zhuo Ouyang, Kaiwen Hu, Qi Zhang, Yifei Wang, Yisen Wang
Abstract: Recently, contrastive learning has risen to be a promising paradigm for extracting meaningful data representations. Among various special designs, adding a projection head on top of the encoder during training and removing it for downstream tasks has proven to significantly enhance the performance of contrastive learning. However, despite its empirical success, the underlying mechanism of the projection head remains under-explored. In this paper, we develop an in-depth theoretical understanding of the projection head from the information-theoretic perspective. By establishing the theoretical guarantees on the downstream performance of the features before the projector, we reveal that an effective projector should act as an information bottleneck, filtering out the information irrelevant to the contrastive objective. Based on theoretical insights, we introduce modifications to projectors with training and structural regularizations. Empirically, our methods exhibit consistent improvement in the downstream performance across various real-world datasets, including CIFAR-10, CIFAR-100, and ImageNet-100. We believe our theoretical understanding on the role of the projection head will inspire more principled and advanced designs in this field. Code is available at https://github.com/PKU-ML/Projector_Theory.
Authors: Yuriy Dorn, Aleksandr Katrutsa, Ilgam Latypov, Anastasiia Soboleva
Abstract: Bandit optimization usually refers to the class of online optimization problems with limited feedback, namely, a decision maker uses only the objective value at the current point to make a new decision and does not have access to the gradient of the objective function. While this name accurately captures the limitation in feedback, it is somehow misleading since it does not have any connection with the multi-armed bandits (MAB) problem class. We propose two new classes of problems: the functional multi-armed bandit problem (FMAB) and the best function identification problem. They are modifications of a multi-armed bandit problem and the best arm identification problem, respectively, where each arm represents an unknown black-box function. These problem classes are a surprisingly good fit for modeling real-world problems such as competitive LLM training. To solve the problems from these classes, we propose a new reduction scheme to construct UCB-type algorithms, namely, the F-LCB algorithm, based on algorithms for nonlinear optimization with known convergence rates. We provide the regret upper bounds for this reduction scheme based on the base algorithms' convergence rates. We add numerical experiments that demonstrate the performance of the proposed scheme.
Authors: Kishalay Das, Subhojyoti Khastagir, Pawan Goyal, Seung-Cheol Lee, Satadeep Bhattacharjee, Niloy Ganguly
Abstract: Equivariant diffusion models have emerged as the prevailing approach for generating novel crystal materials due to their ability to leverage the physical symmetries of periodic material structures. However, current models do not effectively learn the joint distribution of atom types, fractional coordinates, and lattice structure of the crystal material in a cohesive end-to-end diffusion framework. Also, none of these models work under realistic setups, where users specify the desired characteristics that the generated structures must match. In this work, we introduce TGDMat, a novel text-guided diffusion model designed for 3D periodic material generation. Our approach integrates global structural knowledge through textual descriptions at each denoising step while jointly generating atom coordinates, types, and lattice structure using a periodic-E(3)-equivariant graph neural network (GNN). Extensive experiments using popular datasets on benchmark tasks reveal that TGDMat outperforms existing baseline methods by a good margin. Notably, for the structure prediction task, with just one generated sample, TGDMat outperforms all baseline models, highlighting the importance of text-guided diffusion. Further, in the generation task, TGDMat surpasses all baselines and their text-fusion variants, showcasing the effectiveness of the joint diffusion paradigm. Additionally, incorporating textual knowledge reduces overall training and sampling computational overhead while enhancing generative performance when utilizing real-world textual prompts from experts.
Authors: Denis Blessing, Xiaogang Jia, Gerhard Neumann
Abstract: Diffusion models optimized via variational inference (VI) have emerged as a promising tool for generating samples from unnormalized target densities. These models create samples by simulating a stochastic differential equation, starting from a simple, tractable prior, typically a Gaussian distribution. However, when the support of this prior differs greatly from that of the target distribution, diffusion models often struggle to explore effectively or suffer from large discretization errors. Moreover, learning the prior distribution can lead to mode-collapse, exacerbated by the mode-seeking nature of reverse Kullback-Leibler divergence commonly used in VI. To address these challenges, we propose end-to-end learnable Gaussian mixture priors (GMPs). GMPs offer improved control over exploration, adaptability to target support, and increased expressiveness to counteract mode collapse. We further leverage the structure of mixture models by proposing a strategy to iteratively refine the model by adding mixture components during training. Our experimental results demonstrate significant performance improvements across a diverse range of real-world and synthetic benchmark problems when using GMPs without requiring additional target evaluations.
Authors: Zirun Guo, Shulei Wang, Wang Lin, Weicai Yan, Yangyang Wu, Tao Jin
Abstract: Missing modality issues are common in real-world applications, arising from factors such as equipment failures and privacy concerns. When fine-tuning pre-trained models on downstream datasets with missing modalities, performance can degrade significantly. Current methods often aggregate various missing cases to train recovery modules or align multimodal features, resulting in suboptimal performance, high computational costs, and the risk of catastrophic forgetting in continual environments where data arrives sequentially. In this paper, we formulate the dynamic missing modality problem as a continual learning task and introduce the continual multimodal missing modality task. To address this challenge efficiently, we introduce three types of prompts: modality-specific, task-aware, and task-specific prompts. These prompts enable the model to learn intra-modality, inter-modality, intra-task, and inter-task features. Furthermore, we propose a contrastive task interaction strategy to explicitly learn prompts correlating different modalities. We conduct extensive experiments on three public datasets, where our method consistently outperforms state-of-the-art approaches.
Authors: Haofei Lu, Dongqi Han, Yifei Shen, Dongsheng Li
Abstract: Diffusion models have recently shown significant potential in solving decision-making problems, particularly in generating behavior plans -- also known as diffusion planning. While numerous studies have demonstrated the impressive performance of diffusion planning, the mechanisms behind the key components of a good diffusion planner remain unclear and the design choices are highly inconsistent in existing studies. In this work, we address this issue through systematic empirical experiments on diffusion planning in an offline reinforcement learning (RL) setting, providing practical insights into the essential components of diffusion planning. We trained and evaluated over 6,000 diffusion models, identifying the critical components such as guided sampling, network architecture, action generation and planning strategy. We revealed that some design choices opposite to the common practice in previous work in diffusion planning actually lead to better performance, e.g., unconditional sampling with selection can be better than guided sampling and Transformer outperforms U-Net as denoising network. Based on these insights, we suggest a simple yet strong diffusion planning baseline that achieves state-of-the-art results on standard offline RL benchmarks.
Authors: Junjie Sheng, Jiehao Wu, Haochuan Cui, Yiqiu Hu, Wenli Zhou, Lei Zhu, Qian Peng, Wenhao Li, Xiangfeng Wang
Abstract: Recent advancements in reinforcement learning (RL) have shown promise for optimizing virtual machine scheduling (VMS) in small-scale clusters. The utilization of RL to large-scale cloud computing scenarios remains notably constrained. This paper introduces a scalable RL framework, called Cluster Value Decomposition Reinforcement Learning (CVD-RL), to surmount the scalability hurdles inherent in large-scale VMS. The CVD-RL framework innovatively combines a decomposition operator with a look-ahead operator to adeptly manage representation complexities, while complemented by a Top-$k$ filter operator that refines exploration efficiency. Different from existing approaches limited to clusters of $10$ or fewer physical machines (PMs), CVD-RL extends its applicability to environments encompassing up to $50$ PMs. Furthermore, the CVD-RL framework demonstrates generalization capabilities that surpass contemporary SOTA methodologies across a variety of scenarios in empirical studies. This breakthrough not only showcases the framework's exceptional scalability and performance but also represents a significant leap in the application of RL for VMS within complex, large-scale cloud infrastructures. The code is available at https://anonymous.4open.science/r/marl4sche-D0FE.
Authors: Debmalya Mandal, Paulius Sasnauskas, Goran Radanovic
Abstract: Reinforcement learning from human feedback (RLHF) has evolved to be one of the main methods for fine-tuning large language models (LLMs). However, existing RLHF methods are non-robust, and their performance deteriorates if the downstream task differs significantly from the preference dataset used in fine-tuning. In order to mitigate this problem, we introduce a distributionally robust RLHF for fine-tuning LLMs. In particular, our goal is to ensure that a fine-tuned model retains its performance even when the distribution of prompts significantly differs from the distribution encountered during fine-tuning. We formulate distributionally robust optimization (DRO) version of two popular fine-tuning methods -- (1) reward-based RLHF and (2) reward-free DPO (direct preference optimization). We propose a minibatch gradient descent based algorithms for both of them, and theoretically prove convergence guarantees for the algorithms. Subsequently, we evaluate our algorithms on an out-of-distribution (OOD) task by first training the model on the Unified-Feedback dataset and evaluating its performance on two different datasets. The experimental results show that our robust training improves the accuracy of the learned reward models on average, and markedly on some tasks, such as reasoning. Furthermore, we show that the robust versions of policy optimization methods, similarly improve performance on OOD tasks.
Authors: Lukas Fesser, Melanie Weber
Abstract: Graph Neural Networks have emerged as the most popular architecture for graph-level learning, including graph classification and regression tasks, which frequently arise in areas such as biochemistry and drug discovery. Achieving good performance in practice requires careful model design. Due to gaps in our understanding of the relationship between model and data characteristics, this often requires manual architecture and hyperparameter tuning. This is particularly pronounced in graph-level tasks, due to much higher variation in the input data than in node-level tasks. To work towards closing these gaps, we begin with a systematic analysis of individual performance in graph-level tasks. Our results establish significant performance heterogeneity in both message-passing and transformer-based architectures. We then investigate the interplay of model and data characteristics as drivers of the observed heterogeneity. Our results suggest that graph topology alone cannot explain heterogeneity. Using the Tree Mover's Distance, which jointly evaluates topological and feature information, we establish a link between class-distance ratios and performance heterogeneity in graph classification. These insights motivate model and data preprocessing choices that account for heterogeneity between graphs. We propose a selective rewiring approach, which only targets graphs whose individual performance benefits from rewiring. We further show that the optimal network depth depends on the graph's spectrum, which motivates a heuristic for choosing the number of GNN layers. Our experiments demonstrate the utility of both design choices in practice.
Authors: Ramit Debnath, Taran Chandel, Fengyuan Han, Ronita Bardhan
Abstract: Heatwaves, intensified by climate change and rapid urbanisation, pose significant threats to urban systems, particularly in the Global South, where adaptive capacity is constrained. This study investigates the relationship between heatwaves and nighttime light (NTL) radiance, a proxy of nighttime economic activity, in four hyperdense cities: Delhi, Guangzhou, Cairo, and Sao Paulo. We hypothesised that heatwaves increase nighttime activity. Using a double machine learning (DML) framework, we analysed data from 2013 to 2019 to quantify the impact of heatwaves on NTL while controlling for local climatic confounders. Results revealed a statistically significant increase in NTL intensity during heatwaves, with Cairo, Delhi, and Guangzhou showing elevated NTL on the third day, while S\~ao Paulo exhibits a delayed response on the fourth day. Sensitivity analyses confirmed the robustness of these findings, indicating that prolonged heat stress prompts urban populations to shift activities to night. Heterogeneous responses across cities highlight the possible influence of urban morphology and adaptive capacity to heatwave impacts. Our findings provide a foundation for policymakers to develop data-driven heat adaptation strategies, ensuring that cities remain liveable and economically resilient in an increasingly warming world.
Authors: Eric Heim, Oren Wright, David Shriver
Abstract: One of the main barriers to adoption of Machine Learning (ML) is that ML models can fail unexpectedly. In this work, we aim to provide practitioners a guide to better understand why ML models fail and equip them with techniques they can use to reason about failure. Specifically, we discuss failure as either being caused by lack of reliability or lack of robustness. Differentiating the causes of failure in this way allows us to formally define why models fail from first principles and tie these definitions to engineering concepts and real-world deployment settings. Throughout the document we provide 1) a summary of important theoretic concepts in reliability and robustness, 2) a sampling current techniques that practitioners can utilize to reason about ML model reliability and robustness, and 3) examples that show how these concepts and techniques can apply to real-world settings.
Authors: Jake B. Perazzone, Shiqiang Wang, Mingyue Ji, Kevin Chan
Abstract: Federated learning (FL) is a useful tool that enables the training of machine learning models over distributed data without having to collect data centrally. When deploying FL in constrained wireless environments, however, intermittent connectivity of devices, heterogeneous connection quality, and non-i.i.d. data can severely slow convergence. In this paper, we consider FL with arbitrary device participation probabilities for each round and show that by weighing each device's update by the reciprocal of their per-round participation probability, we can guarantee convergence to a stationary point. Our bound applies to non-convex loss functions and non-i.i.d. datasets and recovers state-of-the-art convergence rates for both full and uniform partial participation, including linear speedup, with only a single-sided learning rate. Then, using the derived convergence bound, we develop a new online client selection and power allocation algorithm that utilizes the Lyapunov drift-plus-penalty framework to opportunistically minimize a function of the convergence bound and the average communication time under a transmit power constraint. We use optimization over manifold techniques to obtain a solution to the minimization problem. Thanks to the Lyapunov framework, one key feature of the algorithm is that knowledge of the channel distribution is not required and only the instantaneous channel state information needs to be known. Using the CIFAR-10 dataset with varying levels of data heterogeneity, we show through simulations that the communication time can be significantly decreased using our algorithm compared to uniformly random participation, especially for heterogeneous channel conditions.
Authors: Tu\u{g}rul Hasan Karabulut, \.Inci M. Bayta\c{s}
Abstract: Graph Neural Networks (GNNs) set the state-of-the-art in representation learning for graph-structured data. They are used in many domains, from online social networks to complex molecules. Most GNNs leverage the message-passing paradigm and achieve strong performances on various tasks. However, the message-passing mechanism used in most models suffers from over-smoothing as a GNN's depth increases. The over-smoothing degrades GNN's performance due to the increased similarity between the representations of unrelated nodes. This study proposes an adaptive channel-wise message-passing approach to alleviate the over-smoothing. The proposed model, Channel-Attentive GNN, learns how to attend to neighboring nodes and their feature channels. Thus, much diverse information can be transferred between nodes during message-passing. Experiments with widely used benchmark datasets show that the proposed model is more resistant to over-smoothing than baselines and achieves state-of-the-art performances for various graphs with strong heterophily. Our code is at https://github.com/ALLab-Boun/CHAT-GNN.
Authors: Xinliang Zhou, Chenyu Liu, Zhisheng Chen, Kun Wang, Yi Ding, Ziyu Jia, Qingsong Wen
Abstract: Brain foundation models (BFMs) have emerged as a transformative paradigm in computational neuroscience, offering a revolutionary framework for processing diverse neural signals across different brain-related tasks. These models leverage large-scale pre-training techniques, allowing them to generalize effectively across multiple scenarios, tasks, and modalities, thus overcoming the traditional limitations faced by conventional artificial intelligence (AI) approaches in understanding complex brain data. By tapping into the power of pretrained models, BFMs provide a means to process neural data in a more unified manner, enabling advanced analysis and discovery in the field of neuroscience. In this survey, we define BFMs for the first time, providing a clear and concise framework for constructing and utilizing these models in various applications. We also examine the key principles and methodologies for developing these models, shedding light on how they transform the landscape of neural signal processing. This survey presents a comprehensive review of the latest advancements in BFMs, covering the most recent methodological innovations, novel views of application areas, and challenges in the field. Notably, we highlight the future directions and key challenges that need to be addressed to fully realize the potential of BFMs. These challenges include improving the quality of brain data, optimizing model architecture for better generalization, increasing training efficiency, and enhancing the interpretability and robustness of BFMs in real-world applications.
Authors: Nicky Kriplani, Minh Pham, Gowthami Somepalli, Chinmay Hegde, Niv Cohen
Abstract: Recent works have shown that diffusion models are able to memorize training images and emit them at generation time. However, the metrics used to evaluate memorization and its mitigation techniques suffer from dataset-dependent biases and struggle to detect whether a given specific image has been memorized or not. This paper begins with a comprehensive exploration of issues surrounding memorization metrics in diffusion models. Then, to mitigate these issues, we introduce $\rm \style{font-variant: small-caps}{SolidMark}$, a novel evaluation method that provides a per-image memorization score. We then re-evaluate existing memorization mitigation techniques. We also show that $\rm \style{font-variant: small-caps}{SolidMark}$ is capable of evaluating fine-grained pixel-level memorization. Finally, we release a variety of models based on $\rm \style{font-variant: small-caps}{SolidMark}$ to facilitate further research for understanding memorization phenomena in generative models. All of our code is available at https://github.com/NickyDCFP/SolidMark.
Authors: Adam N. Elmachtoub, Henry Lam, Haixiang Lan, Haofeng Zhang
Abstract: Data-driven optimization aims to translate a machine learning model into decision-making by optimizing decisions on estimated costs. Such a pipeline can be conducted by fitting a distributional model which is then plugged into the target optimization problem. While this fitting can utilize traditional methods such as maximum likelihood, a more recent approach uses estimation-optimization integration that minimizes decision error instead of estimation error. Although intuitive, the statistical benefit of the latter approach is not well understood yet is important to guide the prescriptive usage of machine learning. In this paper, we dissect the performance comparisons between these approaches in terms of the amount of model misspecification. In particular, we show how the integrated approach offers a ``universal double benefit'' on the top two dominating terms of regret when the underlying model is misspecified, while the traditional approach can be advantageous when the model is nearly well-specified. Our comparison is powered by finite-sample tail regret bounds that are derived via new higher-order expansions of regrets and the leveraging of a recent Berry-Esseen theorem.
Authors: Iyas AlTalafha, Yaprak Yalcin, Gulcihan Ozdemir
Abstract: Production Lines and Conveying Systems are the staple of modern manufacturing processes. Manufacturing efficiency is directly related to the efficiency of the means of production and conveying. Modelling in the industrial context has always been a challenge due to the complexity that comes along with modern manufacturing standards. Long Short-Term Memory is a pattern recognition Recurrent Neural Network, that is utilised on a simple pneumatic conveying system which transports a wooden block around the system. Recurrent Neural Networks (RNNs) capture temporal dependencies through feedback loops, while Long Short-Term Memory (LSTM) networks enhance this capability by using gated mechanisms to effectively learn long-term dependencies. Conveying systems, representing a major component of production lines, are chosen as the target to model to present an approach applicable in large scale production lines in a simpler format. In this paper data from sensors are used to train the LSTM in order to output an Automaton that models the conveying system. The automaton obtained from the proposed LSTM approach is compared with the automaton obtained from OTALA. The resultant LSTM automaton proves to be a more accurate representation of the conveying system, unlike the one obtained from OTALA.
Authors: Yifei He, Yang Liu, Chen Liang, Hany Hassan Awadalla
Abstract: Mixture-of-Experts (MoE) models have become a key approach for scaling large language models efficiently by activating only a subset of experts during training and inference. Typically, the number of activated experts presents a trade-off: fewer experts reduce computational costs, while more experts improve performance. Recent studies reveal that not all activated experts contribute equally to model performance, with some providing minimal utility, particularly when finetuning pretrained MoE models for specialized downstream tasks. The co-existence of significant and redundant parameters in experts provides us an opportunity to reduce the number of activated experts while maintaining model performance. In this work, we propose the concept of compressed experts, lightweight modules that serve as compact representations of full experts. Our approach preserves the most important experts while replacing other auxiliary activated experts with compressed experts. The reduction of active parameters significantly lowers inference costs while achieving comparable performance. Extensive experiments on models including Phi-MoE and OLMoE demonstrate that compressed experts recover over 90% of full expert performance across various tasks while reducing more than 30% active parameters and saving 20% in inference costs. This approach enables efficient deployment of MoE models in resource-constrained settings and facilitates scaling to larger models with manageable overhead. Our code is available at https://github.com/yifei-he/Compressed-Experts.
Authors: Zijian Li, Shunxing Fan, Yujia Zheng, Ignavier Ng, Shaoan Xie, Guangyi Chen, Xinshuai Dong, Ruichu Cai, Kun Zhang
Abstract: Disentangled representation learning aims to uncover latent variables underlying the observed data, and generally speaking, rather strong assumptions are needed to ensure identifiability. Some approaches rely on sufficient changes on the distribution of latent variables indicated by auxiliary variables such as domain indices, but acquiring enough domains is often challenging. Alternative approaches exploit structural sparsity assumptions on the mixing procedure, but such constraints are usually (partially) violated in practice. Interestingly, we find that these two seemingly unrelated assumptions can actually complement each other to achieve identifiability. Specifically, when conditioned on auxiliary variables, the sparse mixing procedure assumption provides structural constraints on the mapping from estimated to true latent variables and hence compensates for potentially insufficient distribution changes. Building on this insight, we propose an identifiability theory with less restrictive constraints regarding distribution changes and the sparse mixing procedure, enhancing applicability to real-world scenarios. Additionally, we develop an estimation framework incorporating a domain encoding network and a sparse mixing constraint and provide two implementations based on variational autoencoders and generative adversarial networks, respectively. Experiment results on synthetic and real-world datasets support our theoretical results.
Authors: Ali Shirali, Ariel Procaccia, Rediet Abebe
Abstract: Algorithmic predictions are increasingly informing societal resource allocations by identifying individuals for targeting. Policymakers often build these systems with the assumption that by gathering more observations on individuals, they can improve predictive accuracy and, consequently, allocation efficiency. An overlooked yet consequential aspect of prediction-driven allocations is that of timing. The planner has to trade off relying on earlier and potentially noisier predictions to intervene before individuals experience undesirable outcomes, or they may wait to gather more observations to make more precise allocations. We examine this tension using a simple mathematical model, where the planner collects observations on individuals to improve predictions over time. We analyze both the ranking induced by these predictions and optimal resource allocation. We show that though individual prediction accuracy improves over time, counter-intuitively, the average ranking loss can worsen. As a result, the planner's ability to improve social welfare can decline. We identify inequality as a driving factor behind this phenomenon. Our findings provide a nuanced perspective and challenge the conventional wisdom that it is preferable to wait for more accurate predictions to ensure the most efficient allocations.
Authors: Aidan Scannell, Mohammadreza Nakhaei, Kalle Kujanp\"a\"a, Yi Zhao, Kevin Sebastian Luck, Arno Solin, Joni Pajarinen
Abstract: In reinforcement learning (RL), world models serve as internal simulators, enabling agents to predict environment dynamics and future outcomes in order to make informed decisions. While previous approaches leveraging discrete latent spaces, such as DreamerV3, have demonstrated strong performance in discrete action settings and visual control tasks, their comparative performance in state-based continuous control remains underexplored. In contrast, methods with continuous latent spaces, such as TD-MPC2, have shown notable success in state-based continuous control benchmarks. In this paper, we demonstrate that modeling discrete latent states has benefits over continuous latent states and that discrete codebook encodings are more effective representations for continuous control, compared to alternative encodings, such as one-hot and label-based encodings. Based on these insights, we introduce DCWM: Discrete Codebook World Model, a self-supervised world model with a discrete and stochastic latent space, where latent states are codes from a codebook. We combine DCWM with decision-time planning to get our model-based RL algorithm, named DC-MPC: Discrete Codebook Model Predictive Control, which performs competitively against recent state-of-the-art algorithms, including TD-MPC2 and DreamerV3, on continuous control benchmarks. See our project website www.aidanscannell.com/dcmpc.
Authors: Yousuf Islam, Sumon Chandra Das, Md. Jalal Uddin Chowdhury
Abstract: The rapid evolution of machine learning (ML) has brought about groundbreaking developments in numerous industries, not the least of which is in the area of undersea communication. This domain is critical for applications like ocean exploration, environmental monitoring, resource management, and national security. Bangladesh, a maritime nation with abundant resources in the Bay of Bengal, can harness the immense potential of ML to tackle the unprecedented challenges associated with underwater communication. Beyond that, environmental conditions are unique to the region: in addition to signal attenuation, multipath propagation, noise interference, and limited bandwidth. In this study, we address the necessity to bring ML into communication via undersea; it investigates the latest technologies under the domain of ML in that respect, such as deep learning and reinforcement learning, especially concentrating on Bangladesh scenarios in the sense of implementation. This paper offers a contextualized regional perspective by incorporating region-specific needs, case studies, and recent research to propose a roadmap for deploying ML-driven solutions to improve safety at sea, promote sustainable resource use, and enhance disaster response systems. This research ultimately highlights the promise of ML-powered solutions for transforming undersea communication, leading to more efficient and cost-effective technologies that subsequently contribute to both economic growth and environmental sustainability.
Authors: Laziz Abdullaev, Tan Nguyen
Abstract: Transformer-based deep learning models have achieved state-of-the-art performance across numerous language and vision tasks. While the self-attention mechanism, a core component of transformers, has proven capable of handling complex data patterns, it has been observed that the representational capacity of the attention matrix degrades significantly across transformer layers, thereby hurting its overall performance. In this work, we leverage the connection between self-attention computations and low-pass non-local means (NLM) smoothing filters and propose the Twicing Attention, a novel attention mechanism that uses kernel twicing procedure in nonparametric regression to alleviate the low-pass behavior of associated NLM smoothing with compelling theoretical guarantees and enhanced adversarial robustness. This approach enables the extraction and reuse of meaningful information retained in the residuals following the imperfect smoothing operation at each layer. Our proposed method offers two key advantages over standard self-attention: 1) a provably slower decay of representational capacity and 2) improved robustness and accuracy across various data modalities and tasks. We empirically demonstrate the performance gains of our model over baseline transformers on multiple tasks and benchmarks, including image classification and language modeling, on both clean and corrupted data.
Authors: Hyunsu Kim, Giung Nam, Chulhee Yun, Hongseok Yang, Juho Lee
Abstract: Bayesian Neural Networks (BNNs) provide a promising framework for modeling predictive uncertainty and enhancing out-of-distribution robustness (OOD) by estimating the posterior distribution of network parameters. Stochastic Gradient Markov Chain Monte Carlo (SGMCMC) is one of the most powerful methods for scalable posterior sampling in BNNs, achieving efficiency by combining stochastic gradient descent with second-order Langevin dynamics. However, SGMCMC often suffers from limited sample diversity in practice, which affects uncertainty estimation and model performance. We propose a simple yet effective approach to enhance sample diversity in SGMCMC without the need for tempering or running multiple chains. Our approach reparameterizes the neural network by decomposing each of its weight matrices into a product of matrices, resulting in a sampling trajectory that better explores the target parameter space. This approach produces a more diverse set of samples, allowing faster mixing within the same computational budget. Notably, our sampler achieves these improvements without increasing the inference cost compared to the standard SGMCMC. Extensive experiments on image classification tasks, including OOD robustness, diversity, loss surface analyses, and a comparative study with Hamiltonian Monte Carlo, demonstrate the superiority of the proposed approach.
Authors: Zhiqi Bu, Ruixuan Liu
Abstract: Differential privacy (DP) is a privacy-preserving paradigm that protects the training data when training deep learning models. Critically, the performance of models is determined by the training hyperparameters, especially those of the learning rate schedule, thus requiring fine-grained hyperparameter tuning on the data. In practice, it is common to tune the learning rate hyperparameters through the grid search that (1) is computationally expensive as multiple runs are needed, and (2) increases the risk of data leakage as the selection of hyperparameters is data-dependent. In this work, we adapt the automatic learning rate schedule to DP optimization for any models and optimizers, so as to significantly mitigate or even eliminate the cost of hyperparameter tuning when applied together with automatic per-sample gradient clipping. Our hyperparameter-free DP optimization is almost as computationally efficient as the standard non-DP optimization, and achieves state-of-the-art DP performance on various language and vision tasks.
Authors: Tomas Geffner, Kieran Didi, Zuobai Zhang, Danny Reidenbach, Zhonglin Cao, Jason Yim, Mario Geiger, Christian Dallago, Emine Kucukbenli, Arash Vahdat, Karsten Kreis
Abstract: Recently, diffusion- and flow-based generative models of protein structures have emerged as a powerful tool for de novo protein design. Here, we develop Proteina, a new large-scale flow-based protein backbone generator that utilizes hierarchical fold class labels for conditioning and relies on a tailored scalable transformer architecture with up to 5x as many parameters as previous models. To meaningfully quantify performance, we introduce a new set of metrics that directly measure the distributional similarity of generated proteins with reference sets, complementing existing metrics. We further explore scaling training data to millions of synthetic protein structures and explore improved training and sampling recipes adapted to protein backbone generation. This includes fine-tuning strategies like LoRA for protein backbones, new guidance methods like classifier-free guidance and autoguidance for protein backbones, and new adjusted training objectives. Proteina achieves state-of-the-art performance on de novo protein backbone design and produces diverse and designable proteins at unprecedented length, up to 800 residues. The hierarchical conditioning offers novel control, enabling high-level secondary-structure guidance as well as low-level fold-specific generation.
Authors: Zhijiang Wan, Qianhao Yu, Jia Mao, Wenfeng Duan, Cheng Ding
Abstract: This study introduces OpenECG, a large-scale benchmark of 1.2 million 12-lead ECG recordings from nine centers, to evaluate ECG foundation models (ECG-FMs) trained on public datasets. We investigate three self-supervised learning methods (SimCLR, BYOL, MAE) with ResNet-50 and Vision Transformer architectures, assessing model generalization through leave-one-dataset-out experiments and data scaling analysis. Results show that pre-training on diverse datasets significantly improves generalization, with BYOL and MAE outperforming SimCLR, highlighting the efficacy of feature-consistency and generative learning over contrastive approaches. Data scaling experiments reveal that performance saturates at 60-70% of total data for BYOL and MAE, while SimCLR requires more data. These findings demonstrate that publicly available ECG data can match or surpass proprietary datasets in training robust ECG-FMs, paving the way for scalable, clinically meaningful AI-driven ECG analysis.
Authors: Yiyang Liu, James Chenhao Liang, Ruixiang Tang, Yugyung Lee, Majid Rabbani, Sohail Dianat, Raghuveer Rao, Lifu Huang, Dongfang Liu, Qifan Wang, Cheng Han
Abstract: Multimodal instruction tuning has proven to be an effective strategy for achieving zero-shot generalization by fine-tuning pre-trained Large Multimodal Models (LMMs) with instruction-following data. However, as the scale of LMMs continues to grow, fully fine-tuning these models has become highly parameter-intensive. Although Parameter-Efficient Fine-Tuning (PEFT) methods have been introduced to reduce the number of tunable parameters, a significant performance gap remains compared to full fine-tuning. Furthermore, existing PEFT approaches are often highly parameterized, making them difficult to interpret and control. In light of this, we introduce Multimodal Representation Tuning (MRT), a novel approach that focuses on directly editing semantically rich multimodal representations to achieve strong performance and provide intuitive control over LMMs. Empirical results show that our method surpasses current state-of-the-art baselines with significant performance gains (e.g., 1580.40 MME score) while requiring substantially fewer tunable parameters (e.g., 0.03% parameters). Additionally, we conduct experiments on editing instrumental tokens within multimodal representations, demonstrating that direct manipulation of these representations enables simple yet effective control over network behavior.
Authors: Toby Simonds, Akira Yoshiyama
Abstract: We introduce LADDER (Learning through Autonomous Difficulty-Driven Example Recursion), a framework enabling LLMs to autonomously improve their problem-solving capabilities through self-guided learning. By recursively generating and solving progressively simpler variants of complex problems, LADDER enables models to progressively learn through reinforcement learning how to solve harder problems. This self-improvement process is guided by verifiable reward signals, allowing the model to assess its solutions. Unlike prior approaches requiring curated datasets or human feedback, LADDER leverages the model's own capabilities to easier variants of sample questions. We demonstrate LADDER's effectiveness on mathematical integration tasks, where it improves a Llama 3B model's accuracy from 1\% to 82\% on undergraduate-level problems and enables a 7B parameter model to achieve state-of-the-art performance (70\%) on the MIT Integration Bee examination for it's model size. We also introduce TTRL (Test-Time Reinforcement Learning), a method that generates variants of test problems at inference time and applies reinforcement learning to further improve performance. By further creating and solving related problems during testing, TTRL enables the 7B model to achieve a score of 85\%, surpassing o1. These results showcase how strategic self-directed learning can achieve significant capability improvements without relying on architectural scaling or human supervision.
Authors: Xingbo Fu, Yinhan He, Jundong Li
Abstract: Pre-training powerful Graph Neural Networks (GNNs) with unlabeled graph data in a self-supervised manner has emerged as a prominent technique in recent years. However, inevitable objective gaps often exist between pre-training and downstream tasks. To bridge this gap, graph prompt tuning techniques design and learn graph prompts by manipulating input graphs or reframing downstream tasks as pre-training tasks without fine-tuning the pre-trained GNN models. While recent graph prompt tuning methods have proven effective in adapting pre-trained GNN models for downstream tasks, they overlook the crucial role of edges in graph prompt design, which can significantly affect the quality of graph representations for downstream tasks. In this study, we propose EdgePrompt, a simple yet effective graph prompt tuning method from the perspective of edges. Unlike previous studies that design prompt vectors on node features, EdgePrompt manipulates input graphs by learning additional prompt vectors for edges and incorporates the edge prompts through message passing in the pre-trained GNN models to better embed graph structural information for downstream tasks. Our method is compatible with prevalent GNN architectures pre-trained under various pre-training strategies and is universal for different downstream tasks. We provide comprehensive theoretical analyses of our method regarding its capability of handling node classification and graph classification as downstream tasks. Extensive experiments on ten graph datasets under four pre-training strategies demonstrate the superiority of our proposed method against six baselines. Our code is available at https://github.com/xbfu/EdgePrompt.
Authors: Anas Jnini, Lorenzo Breschi, Flavio Vella
Abstract: Divergence-free symmetric tensors (DFSTs) are fundamental in continuum mechanics, encoding conservation laws such as mass and momentum conservation. We introduce Riemann Tensor Neural Networks (RTNNs), a novel neural architecture that inherently satisfies the DFST condition to machine precision, providing a strong inductive bias for enforcing these conservation laws. We prove that RTNNs can approximate any sufficiently smooth DFST with arbitrary precision and demonstrate their effectiveness as surrogates for conservative PDEs, achieving improved accuracy across benchmarks. This work is the first to use DFSTs as an inductive bias in neural PDE surrogates and to explicitly enforce the conservation of both mass and momentum within a physics-constrained neural architecture.
Authors: Wei Liu, Tao Zhang, Chenhui Lin, Kaiwen Li, Rui Wang
Abstract: Independent microgrids are crucial for supplying electricity by combining distributed energy resources and loads in scenarios like isolated islands and field combat. Fast and accurate assessments of microgrid vulnerability against intentional attacks or natural disasters are essential for effective risk prevention and design optimization. However, conventional Monte Carlo simulation (MCS) methods are computationally expensive and time-consuming, while existing machine learning-based approaches often lack accuracy and explainability. To address these challenges, this study proposes a fast and explainable vulnerability assessment framework that integrates MCS with a graph attention network enhanced by self-attention pooling (GAT-S). MCS generates training data, while the GAT-S model learns the structural and electrical characteristics of the microgrid and further assesses its vulnerability intelligently. The GAT-S improves explainability and computational efficiency by dynamically assigning attention weights to critical nodes. Comprehensive experimental evaluations across various microgrid configurations demonstrate that the proposed framework provides accurate vulnerability assessments, achieving a mean squared error as low as 0.001, real-time responsiveness within 1 second, and delivering explainable results.
Authors: Jayden Teoh, Pradeep Varakantham, Peter Vamplew
Abstract: Real-world sequential decision-making tasks often require balancing trade-offs between multiple conflicting objectives, making Multi-Objective Reinforcement Learning (MORL) an increasingly prominent field of research. Despite recent advances, existing MORL literature has narrowly focused on performance within static environments, neglecting the importance of generalizing across diverse settings. Conversely, existing research on generalization in RL has always assumed scalar rewards, overlooking the inherent multi-objectivity of real-world problems. Generalization in the multi-objective context is fundamentally more challenging, as it requires learning a Pareto set of policies addressing varying preferences across multiple objectives. In this paper, we formalize the concept of generalization in MORL and how it can be evaluated. We then contribute a novel benchmark featuring diverse multi-objective domains with parameterized environment configurations to facilitate future studies in this area. Our baseline evaluations of state-of-the-art MORL algorithms on this benchmark reveals limited generalization capabilities, suggesting significant room for improvement. Our empirical findings also expose limitations in the expressivity of scalar rewards, emphasizing the need for multi-objective specifications to achieve effective generalization. We further analyzed the algorithmic complexities within current MORL approaches that could impede the transfer in performance from the single- to multiple-environment settings. This work fills a critical gap and lays the groundwork for future research that brings together two key areas in reinforcement learning: solving multi-objective decision-making problems and generalizing across diverse environments. We make our code available at https://github.com/JaydenTeoh/MORL-Generalization.
Authors: Harin Lee, Min-hwan Oh
Abstract: In our quest for a reinforcement learning (RL) algorithm that is both practical and provably optimal, we introduce EQO (Exploration via Quasi-Optimism). Unlike existing minimax optimal approaches, EQO avoids reliance on empirical variances and employs a simple bonus term proportional to the inverse of the state-action visit count. Central to EQO is the concept of quasi-optimism, where estimated values need not be fully optimistic, allowing for a simpler yet effective exploration strategy. The algorithm achieves the sharpest known regret bound for tabular RL under the mildest assumptions, proving that fast convergence can be attained with a practical and computationally efficient approach. Empirical evaluations demonstrate that EQO consistently outperforms existing algorithms in both regret performance and computational efficiency, providing the best of both theoretical soundness and practical effectiveness.
Authors: Hongzhi Luan, Changxin Tian, Zhaoxin Huan, Xiaolu Zhang, Kunlong Chen, Zhiqiang Zhang, Jun Zhou
Abstract: This paper poses two critical issues in evaluating base models (without post-training): (1) Unstable evaluation during training: in the early stages of pre-training, the models lack the capability to answer questions as required, leading to unstable evaluation results. This instability makes it difficult to provide solid conclusions to guide the training, especially for key experiments such as data ablation and scaling law. (2) Inconsistency between base and instruct models: base models generally exhibit poorer evaluation performance compared to corresponding instruct models. This gap poses a challenge for assessing whether a base model with better evaluation can truly lead to a better instruct model. To address these issues, we propose Base model Oriented Systematic Evaluation (BOSE), a method specifically designed to optimize the evaluation of base models. Specifically, BOSE introduces two key innovations: In-Context Light-instruction Prompt (ICLiP) for open-ended tasks and Blank-ppl for multi-choice tasks with candidate options, which transforms the standard perplexity (ppl) metric into a fill-in-the-blank format to mitigate early-stage evaluation fluctuations. Furthermore, we are the first to propose Kendall's rank correlation to quantitatively measure the evaluation stability and consistency. Experimental results demonstrate that BOSE significantly enhances both the stability of evaluations during pre-training and the consistency between base and instruct models, thereby providing more reliable guidance for the LLMs' training.
Authors: Jeffrey Gu, Serena Yeung-Levy
Abstract: Large pre-trained models, or foundation models, have shown impressive performance when adapted to a variety of downstream tasks, often out-performing specialized models. Hypernetworks, neural networks that generate some or all of the parameters of another neural network, have become an increasingly important technique for conditioning and generalizing implicit neural representations (INRs), which represent signals or objects such as audio or 3D shapes using a neural network. However, despite the potential benefits of incorporating foundation models in hypernetwork methods, this research direction has not been investigated, likely due to the dissimilarity of the weight generation task with other visual tasks. To address this gap, we (1) show how foundation models can improve hypernetworks with Transformer-based architectures, (2) provide an empirical analysis of the benefits of foundation models for hypernetworks through the lens of the generalizable INR task, showing that leveraging foundation models improves performance, generalizability, and data efficiency across a variety of algorithms and modalities. We also provide further analysis in examining the design space of foundation model-based hypernetworks, including examining the choice of foundation models, algorithms, and the effect of scaling foundation models.
Authors: Shubham Gupta, Srikanta Bedathur
Abstract: Dynamic interactions between entities are prevalent in domains like social platforms, financial systems, healthcare, and e-commerce. These interactions can be effectively represented as time-evolving graphs, where predicting future connections is a key task in applications such as recommendation systems. Temporal Graph Neural Networks (TGNNs) have achieved strong results for such predictive tasks but typically require extensive training data, which is often limited in real-world scenarios. One approach to mitigating data scarcity is leveraging pre-trained models from related datasets. However, direct knowledge transfer between TGNNs is challenging due to their reliance on node-specific memory structures, making them inherently difficult to adapt across datasets. To address this, we introduce a novel transfer approach that disentangles node representations from their associated features through a structured bipartite encoding mechanism. This decoupling enables more effective transfer of memory components and other learned inductive patterns from one dataset to another. Empirical evaluations on real-world benchmarks demonstrate that our method significantly enhances TGNN performance in low-data regimes, outperforming non-transfer baselines by up to 56\% and surpassing existing transfer strategies by 36\%
Authors: Tai Le Quy, Long Le Thanh, Lan Luong Thi Hong, Frank Hopfgartner
Abstract: Fair clustering has attracted remarkable attention from the research community. Many fairness measures for clustering have been proposed; however, they do not take into account the clustering quality w.r.t. the values of the protected attribute. In this paper, we introduce a new visual-based fairness measure for fair clustering through ROC curves, namely FACROC. This fairness measure employs AUCC as a measure of clustering quality and then computes the difference in the corresponding ROC curves for each value of the protected attribute. Experimental results on several popular datasets for fairness-aware machine learning and well-known (fair) clustering models show that FACROC is a beneficial method for visually evaluating the fairness of clustering models.
Authors: Qia Hu, Bo Jiao
Abstract: Graph sampling based Graph Convolutional Networks (GCNs) decouple the sampling from the forward and backward propagation during minibatch training, which exhibit good scalability in terms of layer depth and graph size. We propose HIS_GCNs, a hierarchical importance graph sampling based learning method. By constructing minibatches using sampled subgraphs, HIS_GCNs gives attention to the importance of both core and periphery nodes/edges. Specifically, it preserves the centrum of the core to most minibatches, which maintains connectivity between periphery nodes, and samples periphery edges without core node interference, in order to keep more long chains composed entirely of low-degree nodes in the same minibatch. In addition, we verify the effectiveness of HIS_GCNs in reducing node embedding variance and chain information loss. Experiments on GCNs and other Graph Neural Networks (GNNs) with node classification tasks on five large-scale graphs confirm superior performance of the proposed hierarchical importance sampling method in both accuracy and training time.
Authors: Muhammad Talha Sharif, Abdul Rehman
Abstract: Clinical trial eligibility matching is a critical yet often labor-intensive and error-prone step in medical research, as it ensures that participants meet precise criteria for safe and reliable study outcomes. Recent advances in Natural Language Processing (NLP) have shown promise in automating and improving this process by rapidly analyzing large volumes of unstructured clinical text and structured electronic health record (EHR) data. In this paper, we present a systematic overview of current NLP methodologies applied to clinical trial eligibility screening, focusing on data sources, annotation practices, machine learning approaches, and real-world implementation challenges. A comprehensive literature search (spanning Google Scholar, Mendeley, and PubMed from 2015 to 2024) yielded high-quality studies, each demonstrating the potential of techniques such as rule-based systems, named entity recognition, contextual embeddings, and ontology-based normalization to enhance patient matching accuracy. While results indicate substantial improvements in screening efficiency and precision, limitations persist regarding data completeness, annotation consistency, and model scalability across diverse clinical domains. The review highlights how explainable AI and standardized ontologies can bolster clinician trust and broaden adoption. Looking ahead, further research into advanced semantic and temporal representations, expanded data integration, and rigorous prospective evaluations is necessary to fully realize the transformative potential of NLP in clinical trial recruitment.
Authors: Kota Nakamura, Koki Kawabata, Shungo Tanaka, Yasuko Matsubara, Yasushi Sakurai
Abstract: Cybersecurity systems are continuously producing a huge number of time-stamped events in the form of high-order tensors, such as {count; time, port, flow duration, packet size, . . . }, and so how can we detect anomalies/intrusions in real time? How can we identify multiple types of intrusions and capture their characteristic behaviors? The tensor data consists of categorical and continuous attributes and the data distributions of continuous attributes typically exhibit skew. These data properties require handling skewed infinite and finite dimensional spaces simultaneously. In this paper, we propose a novel streaming method, namely CyberCScope. The method effectively decomposes incoming tensors into major trends while explicitly distinguishing between categorical and skewed continuous attributes. To our knowledge, it is the first to compute hybrid skewed infinite and finite dimensional decomposition. Based on this decomposition, it streamingly finds distinct time-evolving patterns, enabling the detection of multiple types of anomalies. Extensive experiments on large-scale real datasets demonstrate that CyberCScope detects various intrusions with higher accuracy than state-of-the-art baselines while providing meaningful summaries for the intrusions that occur in practice.
Authors: Zijian Dong, Yilei Wu, Chongyao Chen, Yingtian Zou, Yichi Zhang, Juan Helen Zhou
Abstract: In representation learning, uniformity refers to the uniform feature distribution in the latent space (i.e., unit hypersphere). Previous work has shown that improving uniformity contributes to the learning of under-represented classes. However, most of the previous work focused on classification; the representation space of imbalanced regression remains unexplored. Classification-based methods are not suitable for regression tasks because they cluster features into distinct groups without considering the continuous and ordered nature essential for regression. In a geometric aspect, we uniquely focus on ensuring uniformity in the latent space for imbalanced regression through two key losses: enveloping and homogeneity. The enveloping loss encourages the induced trace to uniformly occupy the surface of a hypersphere, while the homogeneity loss ensures smoothness, with representations evenly spaced at consistent intervals. Our method integrates these geometric principles into the data representations via a Surrogate-driven Representation Learning (SRL) framework. Experiments with real-world regression and operator learning tasks highlight the importance of uniformity in imbalanced regression and validate the efficacy of our geometry-based loss functions.
Authors: Dilfira Kudrat, Zongxia Xie, Yanru Sun, Tianyu Jia, Qinghua Hu
Abstract: Time-series forecasting has gained significant attention in machine learning due to its crucial role in various domains. However, most existing forecasting models rely heavily on point-wise loss functions like Mean Square Error, which treat each time step independently and neglect the structural dependencies inherent in time series data, making it challenging to capture complex temporal patterns accurately. To address these challenges, we propose a novel Patch-wise Structural (PS) loss, designed to enhance structural alignment by comparing time series at the patch level. Through leveraging local statistical properties, such as correlation, variance, and mean, PS loss captures nuanced structural discrepancies overlooked by traditional point-wise losses. Furthermore, it integrates seamlessly with point-wise loss, simultaneously addressing local structural inconsistencies and individual time-step errors. PS loss establishes a novel benchmark for accurately modeling complex time series data and provides a new perspective on time series loss function design. Extensive experiments demonstrate that PS loss significantly improves the performance of state-of-the-art models across diverse real-world datasets.
Authors: Rundong He, Yicong Dong, Lanzhe Guo, Yilong Yin, Tailin Wu
Abstract: Semi-supervised learning (SSL) effectively leverages unlabeled data and has been proven successful across various fields. Current safe SSL methods believe that unseen classes in unlabeled data harm the performance of SSL models. However, previous methods for assessing the impact of unseen classes on SSL model performance are flawed. They fix the size of the unlabeled dataset and adjust the proportion of unseen classes within the unlabeled data to assess the impact. This process contravenes the principle of controlling variables. Adjusting the proportion of unseen classes in unlabeled data alters the proportion of seen classes, meaning the decreased classification performance of seen classes may not be due to an increase in unseen class samples in the unlabeled data, but rather a decrease in seen class samples. Thus, the prior flawed assessment standard that ``unseen classes in unlabeled data can damage SSL model performance" may not always hold true. This paper strictly adheres to the principle of controlling variables, maintaining the proportion of seen classes in unlabeled data while only changing the unseen classes across five critical dimensions, to investigate their impact on SSL models from global robustness and local robustness. Experiments demonstrate that unseen classes in unlabeled data do not necessarily impair the performance of SSL models; in fact, under certain conditions, unseen classes may even enhance them.
Authors: Federico Costanza, Lachlan Simpson
Abstract: We introduce Riemannian Integrated Gradients (RIG); an extension of Integrated Gradients (IG) to Riemannian manifolds. We demonstrate that RIG restricts to IG when the Riemannian manifold is Euclidean space. We show that feature attribution can be phrased as an eigenvalue problem where attributions correspond to eigenvalues of a symmetric endomorphism.
Authors: Shashank Gupta, Chaitanya Ahuja, Tsung-Yu Lin, Sreya Dutta Roy, Harrie Oosterhuis, Maarten de Rijke, Satya Narayan Shukla
Abstract: Reinforcement learning ( RL)-based fine-tuning has emerged as a powerful approach for aligning diffusion models with black-box objectives. Proximal policy optimization (PPO) is the most popular choice of method for policy optimization. While effective in terms of performance, PPO is highly sensitive to hyper-parameters and involves substantial computational overhead. REINFORCE, on the other hand, mitigates some computational complexities such as high memory overhead and sensitive hyper-parameter tuning, but has suboptimal performance due to high-variance and sample inefficiency. While the variance of the REINFORCE can be reduced by sampling multiple actions per input prompt and using a baseline correction term, it still suffers from sample inefficiency. To address these challenges, we systematically analyze the efficiency-effectiveness trade-off between REINFORCE and PPO, and propose leave-one-out PPO ( LOOP), a novel RL for diffusion fine-tuning method. LOOP combines variance reduction techniques from REINFORCE, such as sampling multiple actions per input prompt and a baseline correction term, with the robustness and sample efficiency of PPO via clipping and importance sampling. Our results demonstrate that LOOP effectively improves diffusion models on various black-box objectives, and achieves a better balance between computational efficiency and performance.
Authors: Jing Peng, Meiqi Yang, Qiong Zhang, Xiaoxiao Li
Abstract: Multivariate time series data play a pivotal role in a wide range of real-world applications. However, the presence of block missing data introduces significant challenges, often compromising the performance of predictive models. Traditional two-step approaches, which first impute missing values and then perform forecasting, are prone to error accumulation, particularly in complex multivariate settings characterized by high missing ratios and intricate dependency structures. In this work, we introduce S4M, an end-to-end time series forecasting framework that seamlessly integrates missing data handling into the Structured State Space Sequence (S4) model architecture. Unlike conventional methods that treat imputation as a separate preprocessing step, S4M leverages the latent space of S4 models to directly recognize and represent missing data patterns, thereby more effectively capturing the underlying temporal and multivariate dependencies. Our framework comprises two key components: the Adaptive Temporal Prototype Mapper (ATPM) and the Missing-Aware Dual Stream S4 (MDS-S4). The ATPM employs a prototype bank to derive robust and informative representations from historical data patterns, while the MDS-S4 processes these representations alongside missingness masks as dual input streams to enable accurate forecasting. Through extensive empirical evaluations on diverse real-world datasets, we demonstrate that S4M consistently achieves state-of-the-art performance. These results underscore the efficacy of our integrated approach in handling missing data, showcasing its robustness and superiority over traditional imputation-based methods. Our findings highlight the potential of S4M to advance reliable time series forecasting in practical applications, offering a promising direction for future research and deployment. Code is available at https://github.com/WINTERWEEL/S4M.git.
Authors: Ali Ebrahimpour-Boroojeny, Hari Sundaram, Varun Chandrasekaran
Abstract: Machine unlearning, where users can request the deletion of a forget dataset, is becoming increasingly important because of numerous privacy regulations. Initial works on ``exact'' unlearning (e.g., retraining) incur large computational overheads. However, while computationally inexpensive, ``approximate'' methods have fallen short of reaching the effectiveness of exact unlearning: models produced fail to obtain comparable accuracy and prediction confidence on both the forget and test (i.e., unseen) dataset. Exploiting this observation, we propose a new unlearning method, Adversarial Machine UNlearning (AMUN), that outperforms prior state-of-the-art (SOTA) methods for image classification. AMUN lowers the confidence of the model on the forget samples by fine-tuning the model on their corresponding adversarial examples. Adversarial examples naturally belong to the distribution imposed by the model on the input space; fine-tuning the model on the adversarial examples closest to the corresponding forget samples (a) localizes the changes to the decision boundary of the model around each forget sample and (b) avoids drastic changes to the global behavior of the model, thereby preserving the model's accuracy on test samples. Using AMUN for unlearning a random $10\%$ of CIFAR-10 samples, we observe that even SOTA membership inference attacks cannot do better than random guessing.
Authors: Xueping Gong, Jiheng Zhang
Abstract: Dynamic pricing is crucial in sectors like e-commerce and transportation, balancing exploration of demand patterns and exploitation of pricing strategies. Existing methods often require precise knowledge of the demand function, e.g., the H{\"o}lder smoothness level and Lipschitz constant, limiting practical utility. This paper introduces an adaptive approach to address these challenges without prior parameter knowledge. By partitioning the demand function's domain and employing a linear bandit structure, we develop an algorithm that manages regret efficiently, enhancing flexibility and practicality. Our Parameter-Adaptive Dynamic Pricing (PADP) algorithm outperforms existing methods, offering improved regret bounds and extensions for contextual information. Numerical experiments validate our approach, demonstrating its superiority in handling unknown demand parameters.
Authors: Padmanaba Srinivasan, William Knottenbelt
Abstract: Offline reinforcement learning (RL) methods aim to learn optimal policies with access only to trajectories in a fixed dataset. Policy constraint methods formulate policy learning as an optimization problem that balances maximizing reward with minimizing deviation from the behavior policy. Closed form solutions to this problem can be derived as weighted behavioral cloning objectives that, in theory, must compute an intractable partition function. Reinforcement learning has gained popularity in language modeling to align models with human preferences; some recent works consider paired completions that are ranked by a preference model following which the likelihood of the preferred completion is directly increased. We adapt this approach of paired comparison. By reformulating the paired-sample optimization problem, we fit the maximum-mode of the Q function while maximizing behavioral consistency of policy actions. This yields our algorithm, Behavior Preference Regression for offline RL (BPR). We empirically evaluate BPR on the widely used D4RL Locomotion and Antmaze datasets, as well as the more challenging V-D4RL suite, which operates in image-based state spaces. BPR demonstrates state-of-the-art performance over all domains. Our on-policy experiments suggest that BPR takes advantage of the stability of on-policy value functions with minimal perceptible performance degradation on Locomotion datasets.
Authors: Anders Aamand, Justin Y. Chen, Siddharth Gollapudi, Sandeep Silwal, Hao Wu
Abstract: An influential paper of Hsu et al. (ICLR'19) introduced the study of learning-augmented streaming algorithms in the context of frequency estimation. A fundamental problem in the streaming literature, the goal of frequency estimation is to approximate the number of occurrences of items appearing in a long stream of data using only a small amount of memory. Hsu et al. develop a natural framework to combine the worst-case guarantees of popular solutions such as CountMin and CountSketch with learned predictions of high frequency elements. They demonstrate that learning the underlying structure of data can be used to yield better streaming algorithms, both in theory and practice. We simplify and generalize past work on learning-augmented frequency estimation. Our first contribution is a learning-augmented variant of the Misra-Gries algorithm which improves upon the error of learned CountMin and learned CountSketch and achieves the state-of-the-art performance of randomized algorithms (Aamand et al., NeurIPS'23) with a simpler, deterministic algorithm. Our second contribution is to adapt learning-augmentation to a high-dimensional generalization of frequency estimation corresponding to finding important directions (top singular vectors) of a matrix given its rows one-by-one in a stream. We analyze a learning-augmented variant of the Frequent Directions algorithm, extending the theoretical and empirical understanding of learned predictions to matrix streaming.
Authors: Xingzhuo Guo, Yu Zhang, Baixu Chen, Haoran Xu, Jianmin Wang, Mingsheng Long
Abstract: Diffusion models have emerged as powerful generative frameworks by progressively adding noise to data through a forward process and then reversing this process to generate realistic samples. While these models have achieved strong performance across various tasks and modalities, their application to temporal predictive learning remains underexplored. Existing approaches treat predictive learning as a conditional generation problem, but often fail to fully exploit the temporal dynamics inherent in the data, leading to challenges in generating temporally coherent sequences. To address this, we introduce Dynamical Diffusion (DyDiff), a theoretically sound framework that incorporates temporally aware forward and reverse processes. Dynamical Diffusion explicitly models temporal transitions at each diffusion step, establishing dependencies on preceding states to better capture temporal dynamics. Through the reparameterization trick, Dynamical Diffusion achieves efficient training and inference similar to any standard diffusion model. Extensive experiments across scientific spatiotemporal forecasting, video prediction, and time series forecasting demonstrate that Dynamical Diffusion consistently improves performance in temporal predictive tasks, filling a crucial gap in existing methodologies. Code is available at this repository: https://github.com/thuml/dynamical-diffusion.
Authors: Md Abrar Jahin, Shahriar Soudeep, M. F. Mridha, Raihan Kabir, Md Rashedul Islam, Yutaka Watanobe
Abstract: Cybersecurity threats are growing, making network intrusion detection essential. Traditional machine learning models remain effective in resource-limited environments due to their efficiency, requiring fewer parameters and less computational time. However, handling short and highly imbalanced datasets remains challenging. In this study, we propose the fusion of a Contrastive Attentive Graph Network and Graph Attention Network (CAGN-GAT Fusion) and benchmark it against 15 other models, including both Graph Neural Networks (GNNs) and traditional ML models. Our evaluation is conducted on four benchmark datasets (KDD-CUP-1999, NSL-KDD, UNSW-NB15, and CICIDS2017) using a short and proportionally imbalanced dataset with a constant size of 5000 samples to ensure fairness in comparison. Results show that CAGN-GAT Fusion demonstrates stable and competitive accuracy, recall, and F1-score, even though it does not achieve the highest performance in every dataset. Our analysis also highlights the impact of adaptive graph construction techniques, including small changes in connections (edge perturbation) and selective hiding of features (feature masking), improving detection performance. The findings confirm that GNNs, particularly CAGN-GAT Fusion, are robust and computationally efficient, making them well-suited for resource-constrained environments. Future work will explore GraphSAGE layers and multiview graph construction techniques to further enhance adaptability and detection accuracy.
Authors: Guanlue Li, Chenran Jiang, Ziqi Gao, Yu Liu, Chenyang Liu, Jiean Chen, Yong Huang, Jia Li
Abstract: Effective generation of molecular structures, or new chemical entities, that bind to target proteins is crucial for lead identification and optimization in drug discovery. Despite advancements in atom- and motif-wise deep learning models for 3D molecular generation, current methods often struggle with validity and reliability. To address these issues, we develop the Atom-Motif Consistency Diffusion Model (AMDiff), utilizing a joint-training paradigm for multi-view learning. This model features a hierarchical diffusion architecture that integrates both atom- and motif-level views of molecules, allowing for comprehensive exploration of complementary information. By leveraging classifier-free guidance and incorporating binding site features as conditional inputs, AMDiff ensures robust molecule generation across diverse targets. Compared to existing approaches, AMDiff exhibits superior validity and novelty in generating molecules tailored to fit various protein pockets. Case studies targeting protein kinases, including Anaplastic Lymphoma Kinase (ALK) and Cyclin-dependent kinase 4 (CDK4), demonstrate the model's capability in structure-based de novo drug design. Overall, AMDiff bridges the gap between atom-view and motif-view drug discovery and speeds up the process of target-aware molecular generation.
Authors: Stefan Hegselmann, Helen Zhou, Elizabeth Healey, Trenton Chang, Caleb Ellington, Vishwali Mhasawade, Sana Tonekaboni, Peniel Argaw, Haoran Zhang
Abstract: A collection of the accepted Findings papers that were presented at the 4th Machine Learning for Health symposium (ML4H 2024), which was held on December 15-16, 2024, in Vancouver, BC, Canada. ML4H 2024 invited high-quality submissions describing innovative research in a variety of health-related disciplines including healthcare, biomedicine, and public health. Works could be submitted to either the archival Proceedings track, or the non-archival Findings track. The Proceedings track targeted mature, cohesive works with technical sophistication and high-impact relevance to health. The Findings track promoted works that would spark new insights, collaborations, and discussions at ML4H. Both tracks were given the opportunity to share their work through the in-person poster session. All the manuscripts submitted to ML4H Symposium underwent a double-blind peer-review process.
Authors: Denis Blessing, Julius Berner, Lorenz Richter, Gerhard Neumann
Abstract: We provide a general framework for learning diffusion bridges that transport prior to target distributions. It includes existing diffusion models for generative modeling, but also underdamped versions with degenerate diffusion matrices, where the noise only acts in certain dimensions. Extending previous findings, our framework allows to rigorously show that score matching in the underdamped case is indeed equivalent to maximizing a lower bound on the likelihood. Motivated by superior convergence properties and compatibility with sophisticated numerical integration schemes of underdamped stochastic processes, we propose \emph{underdamped diffusion bridges}, where a general density evolution is learned rather than prescribed by a fixed noising process. We apply our method to the challenging task of sampling from unnormalized densities without access to samples from the target distribution. Across a diverse range of sampling problems, our approach demonstrates state-of-the-art performance, notably outperforming alternative methods, while requiring significantly fewer discretization steps and no hyperparameter tuning.
Authors: Yushan Jiang, Wenchao Yu, Geon Lee, Dongjin Song, Kijung Shin, Wei Cheng, Yanchi Liu, Haifeng Chen
Abstract: Time series analysis provides essential insights for real-world system dynamics and informs downstream decision-making, yet most existing methods often overlook the rich contextual signals present in auxiliary modalities. To bridge this gap, we introduce TimeXL, a multi-modal prediction framework that integrates a prototype-based time series encoder with three collaborating Large Language Models (LLMs) to deliver more accurate predictions and interpretable explanations. First, a multi-modal prototype-based encoder processes both time series and textual inputs to generate preliminary forecasts alongside case-based rationales. These outputs then feed into a prediction LLM, which refines the forecasts by reasoning over the encoder's predictions and explanations. Next, a reflection LLM compares the predicted values against the ground truth, identifying textual inconsistencies or noise. Guided by this feedback, a refinement LLM iteratively enhances text quality and triggers encoder retraining. This closed-loop workflow -- prediction, critique (reflect), and refinement -- continuously boosts the framework's performance and interpretability. Empirical evaluations on four real-world datasets demonstrate that TimeXL achieves up to 8.9\% improvement in AUC and produces human-centric, multi-modal explanations, highlighting the power of LLM-driven reasoning for time series prediction.
Authors: Silas Alberti, Kenan Hasanaliyev, Manav Shah, Stefano Ermon
Abstract: Recent work has shown that diffusion models memorize and reproduce training data examples. At the same time, large copyright lawsuits and legislation such as GDPR have highlighted the need for erasing datapoints from diffusion models. However, retraining from scratch is often too expensive. This motivates the setting of data unlearning, i.e., the study of efficient techniques for unlearning specific datapoints from the training set. Existing concept unlearning techniques require an anchor prompt/class/distribution to guide unlearning, which is not available in the data unlearning setting. General-purpose machine unlearning techniques were found to be either unstable or failed to unlearn data. We therefore propose a family of new loss functions called Subtracted Importance Sampled Scores (SISS) that utilize importance sampling and are the first method to unlearn data with theoretical guarantees. SISS is constructed as a weighted combination between simpler objectives that are responsible for preserving model quality and unlearning the targeted datapoints. When evaluated on CelebA-HQ and MNIST, SISS achieved Pareto optimality along the quality and unlearning strength dimensions. On Stable Diffusion, SISS successfully mitigated memorization on nearly 90% of the prompts we tested.
Authors: Yijing Zhang, Dyah Adila, Changho Shin, Frederic Sala
Abstract: Personalizing large language models (LLMs) is essential for delivering tailored interactions that improve user experience. Many existing personalization methods require fine-tuning LLMs for each user, rendering them prohibitively expensive for widespread adoption. Although retrieval-based approaches offer a more compute-efficient alternative, they still depend on large, high-quality datasets that are not consistently available for all users. To address this challenge, we propose CHAMELEON, a scalable and efficient personalization approach that uses (1) self-generated personal preference data and (2) representation editing to enable quick and cost-effective personalization. Our experiments on various tasks, including those from the LaMP personalization benchmark, show that CHAMELEON efficiently adapts models to personal preferences, improving instruction-tuned models and outperforms two personalization baselines by an average of 40% across two model architectures.
Authors: Yanzhou Pan, Huawei Lin, Yide Ran, Jiamin Chen, Xiaodong Yu, Weijie Zhao, Denghui Zhang, Zhaozhuo Xu
Abstract: Large Language Models (LLMs) heavily rely on high-quality training data, making data valuation crucial for optimizing model performance, especially when working within a limited budget. In this work, we aim to offer a third-party data valuation approach that benefits both data providers and model developers. We introduce a linearized future influence kernel (LinFiK), which assesses the value of individual data samples in improving LLM performance during training. We further propose ALinFiK, a learning strategy to approximate LinFiK, enabling scalable data valuation. Our comprehensive evaluations demonstrate that this approach surpasses existing baselines in effectiveness and efficiency, demonstrating significant scalability advantages as LLM parameters increase.
Authors: Jacob Beck
Abstract: While internet-scale image and textual data have enabled strong generalization in Vision-Language Models (VLMs), the absence of internet-scale control data has impeded the development of similar generalization in standard reinforcement learning (RL) agents. Although VLMs are fundamentally limited in their ability to solve control tasks due to their lack of action-conditioned training data, their capacity for image understanding allows them to provide valuable feedback in RL tasks by recognizing successful outcomes. A key challenge in Reinforcement Learning from AI Feedback (RLAIF) is determining how best to integrate VLM-derived signals into the learning process. We explore this question in the context of offline RL and introduce a class of methods called sub-trajectory filtered optimization. We identify three key insights. First, trajectory length plays a crucial role in offline RL, as full-trajectory preference learning exacerbates the stitching problem, necessitating the use of sub-trajectories. Second, even in Markovian environments, a non-Markovian reward signal from a sequence of images is required to assess trajectory improvement, as VLMs do not interpret control actions and must rely on visual cues over time. Third, a simple yet effective approach--filtered and weighted behavior cloning--consistently outperforms more complex reinforcement learning from human feedback-based methods. We propose sub-trajectory filtered behavior cloning, a method that leverages VLM feedback on sub-trajectories while incorporating a retrospective filtering mechanism that removes sub-trajectories preceding failures to improve robustness and prevent turbulence. This study is preliminary; we provide initial evidence through evaluations on a toy control domain. Please enjoy our airport puns.
Authors: Yuyang Huang, Yuhan Liu, Haryadi S. Gunawi, Beibin Li, Changho Hwang
Abstract: Continual learning has emerged as a promising solution to refine models incrementally by leveraging user feedback, thereby enhancing model performance in applications like code completion, personal assistants, and chat interfaces. In particular, online continual learning - iteratively training the model with small batches of user feedback - has demonstrated notable performance improvements. However, the existing practice of segregating training and serving processes forces the online trainer to recompute the intermediate results already done during serving. Such redundant computations can account for 30%-42% of total training time. In this paper, we propose Alchemist, to the best of our knowledge, the first online continual learning system that efficiently reuses intermediate results computed during serving to reduce redundant computation with minimal impact on the serving latency or capacity. Alchemist introduces two key techniques: (1) minimal activations recording and saving during serving, where activations are recorded and saved only during the prefill phase to minimize overhead; and (2) offloading of serving activations, which dynamically manages GPU memory by freeing activations in the forward order, while reloading them in the backward order during the backward pass. Evaluations with the ShareGPT dataset show that compared with a separate training cluster, Alchemist significantly increases training throughput by up to 1.72x, reduces up to 47% memory usage during training, and supports up to 2x more training tokens - all while maintaining negligible impact on serving latency.
Authors: Gokul Swamy, Sanjiban Choudhury, Wen Sun, Zhiwei Steven Wu, J. Andrew Bagnell
Abstract: From a first-principles perspective, it may seem odd that the strongest results in foundation model fine-tuning (FT) are achieved via a relatively complex, two-stage training procedure. Specifically, one first trains a reward model (RM) on some dataset (e.g. human preferences) before using it to provide online feedback as part of a downstream reinforcement learning (RL) procedure, rather than directly optimizing the policy parameters on the dataset via offline maximum likelihood estimation. In fact, from an information-theoretic perspective, we can only lose information via passing through a reward model and cannot create any new information via on-policy sampling. To explain this discrepancy, we scrutinize several hypotheses on the value of RL in FT through both theoretical and empirical lenses. Of the hypotheses considered, we find the most support for the explanation that on problems with a generation-verification gap, the combination of the ease of learning the relatively simple RM (verifier) from the preference data, coupled with the ability of the downstream RL procedure to then filter its search space to the subset of policies (generators) that are optimal for relatively simple verifiers is what leads to the superior performance of online FT.
Authors: Branislav Kveton, Xintong Li, Julian McAuley, Ryan Rossi, Jingbo Shang, Junda Wu, Tong Yu
Abstract: Direct preference optimization (DPO) is a form of reinforcement learning from human feedback (RLHF) where the policy is learned directly from preferential feedback. Although many models of human preferences exist, the critical task of selecting the most informative feedback for training them is under-explored. We propose an active learning framework for DPO, which can be applied to collect human feedback online or to choose the most informative subset of already collected feedback offline. We propose efficient algorithms for both settings. The key idea is to linearize the DPO objective at the last layer of the neural network representation of the optimized policy and then compute the D-optimal design to collect preferential feedback. We prove that the errors in our DPO logit estimates diminish with more feedback. We show the effectiveness of our algorithms empirically in the setting that matches our theory and also on large language models.
Authors: Asela Hevapathige, Ahad N. Zehmakan, Qing Wang
Abstract: Graph Neural Networks (GNNs) have demonstrated strong representation learning capabilities for graph-based tasks. Recent advances on GNNs leverage geometric properties, such as curvature, to enhance its representation capabilities by modeling complex connectivity patterns and information flow within graphs. However, most existing approaches focus solely on discrete graph topology, overlooking diffusion dynamics and task-specific dependencies essential for effective learning. To address this, we propose integrating Bakry-\'Emery curvature, which captures both structural and task-driven aspects of information propagation. We develop an efficient, learnable approximation strategy, making curvature computation scalable for large graphs. Furthermore, we introduce an adaptive depth mechanism that dynamically adjusts message-passing layers per vertex based on its curvature, ensuring efficient propagation. Our theoretical analysis establishes a link between curvature and feature distinctiveness, showing that high-curvature vertices require fewer layers, while low-curvature ones benefit from deeper propagation. Extensive experiments on benchmark datasets validate the effectiveness of our approach, showing consistent performance improvements across diverse graph learning tasks.
Authors: Hyeon Jeon, Micha\"el Aupetit, DongHwa Shin, Aeri Cho, Seokhyeon Park, Jinwook Seo
Abstract: Clustering techniques are often validated using benchmark datasets where class labels are used as ground-truth clusters. However, depending on the datasets, class labels may not align with the actual data clusters, and such misalignment hampers accurate validation. Therefore, it is essential to evaluate and compare datasets regarding their cluster-label matching (CLM), i.e., how well their class labels match actual clusters. Internal validation measures (IVMs), like Silhouette, can compare CLM over different labeling of the same dataset, but are not designed to do so across different datasets. We thus introduce Adjusted IVMs as fast and reliable methods to evaluate and compare CLM across datasets. We establish four axioms that require validation measures to be independent of data properties not related to cluster structure (e.g., dimensionality, dataset size). Then, we develop standardized protocols to convert any IVM to satisfy these axioms, and use these protocols to adjust six widely used IVMs. Quantitative experiments (1) verify the necessity and effectiveness of our protocols and (2) show that adjusted IVMs outperform the competitors, including standard IVMs, in accurately evaluating CLM both within and across datasets. We also show that the datasets can be filtered or improved using our method to form more reliable benchmarks for clustering validation.
Authors: Haoyang Liu, Jie Wang, Zijie Geng, Xijun Li, Yuxuan Zong, Fangzhou Zhu, Jianye Hao, Feng Wu
Abstract: Leveraging machine learning (ML) to predict an initial solution for mixed-integer linear programming (MILP) has gained considerable popularity in recent years. These methods predict a solution and fix a subset of variables to reduce the problem dimension. Then, they solve the reduced problem to obtain the final solutions. However, directly fixing variable values can lead to low-quality solutions or even infeasible reduced problems if the predicted solution is not accurate enough. To address this challenge, we propose an Alternating prediction-correction neural solving framework (Apollo-MILP) that can identify and select accurate and reliable predicted values to fix. In each iteration, Apollo-MILP conducts a prediction step for the unfixed variables, followed by a correction step to obtain an improved solution (called reference solution) through a trust-region search. By incorporating the predicted and reference solutions, we introduce a novel Uncertainty-based Error upper BOund (UEBO) to evaluate the uncertainty of the predicted values and fix those with high confidence. A notable feature of Apollo-MILP is the superior ability for problem reduction while preserving optimality, leading to high-quality final solutions. Experiments on commonly used benchmarks demonstrate that our proposed Apollo-MILP significantly outperforms other ML-based approaches in terms of solution quality, achieving over a 50% reduction in the solution gap.
Authors: Yuheng Zhang, Nan Jiang
Abstract: We investigate off-policy evaluation (OPE), a central and fundamental problem in reinforcement learning (RL), in the challenging setting of Partially Observable Markov Decision Processes (POMDPs) with large observation spaces. Recent works of Uehara et al. (2023a); Zhang & Jiang (2024) developed a model-free framework and identified important coverage assumptions (called belief and outcome coverage) that enable accurate OPE of memoryless policies with polynomial sample complexities, but handling more general target policies that depend on the entire observable history remained an open problem. In this work, we prove information-theoretic hardness for model-free OPE of history-dependent policies in several settings, characterized by additional assumptions imposed on the behavior policy (memoryless vs. history-dependent) and/or the state-revealing property of the POMDP (single-step vs. multi-step revealing). We further show that some hardness can be circumvented by a natural model-based algorithm -- whose analysis has surprisingly eluded the literature despite the algorithm's simplicity -- demonstrating provable separation between model-free and model-based OPE in POMDPs.
Authors: Jonathan Geuter, Cl\'ement Bonet, Anna Korba, David Alvarez-Melis
Abstract: Deep Equilibrium Models (DEQs) are a class of implicit neural networks that solve for a fixed point of a neural network in their forward pass. Traditionally, DEQs take sequences as inputs, but have since been applied to a variety of data. In this work, we present Distributional Deep Equilibrium Models (DDEQs), extending DEQs to discrete measure inputs, such as sets or point clouds. We provide a theoretically grounded framework for DDEQs. Leveraging Wasserstein gradient flows, we show how the forward pass of the DEQ can be adapted to find fixed points of discrete measures under permutation-invariance, and derive adequate network architectures for DDEQs. In experiments, we show that they can compete with state-of-the-art models in tasks such as point cloud classification and point cloud completion, while being significantly more parameter-efficient.
Authors: Teng Pang, Bingzheng Wang, Guoqiang Wu, Yilong Yin
Abstract: Offline preference-based reinforcement learning (PbRL) mitigates the need for reward definition, aligning with human preferences via preference-driven reward feedback without interacting with the environment. However, the effectiveness of preference-driven reward functions depends on the modeling ability of the learning model, which current MLP-based and Transformer-based methods may fail to adequately provide. To alleviate the failure of the reward function caused by insufficient modeling, we propose a novel preference-based reward acquisition method: Diffusion Preference-based Reward (DPR). Unlike previous methods using Bradley-Terry models for trajectory preferences, we use diffusion models to directly model preference distributions for state-action pairs, allowing rewards to be discriminatively obtained from these distributions. In addition, considering the particularity of preference data that only know the internal relationships of paired trajectories, we further propose Conditional Diffusion Preference-based Reward (C-DPR), which leverages relative preference information to enhance the construction of the diffusion model. We apply the above methods to existing offline reinforcement learning algorithms and a series of experiment results demonstrate that the diffusion-based reward acquisition approach outperforms previous MLP-based and Transformer-based methods.
Authors: Sachit Gaudi, Gautam Sreekumar, Vishnu Boddeti
Abstract: How can we learn generative models to sample data with arbitrary logical compositions of statistically independent attributes? The prevailing solution is to sample from distributions expressed as a composition of attributes' conditional marginal distributions under the assumption that they are statistically independent. This paper shows that standard conditional diffusion models violate this assumption, even when all attribute compositions are observed during training. And, this violation is significantly more severe when only a subset of the compositions is observed. We propose CoInD to address this problem. It explicitly enforces statistical independence between the conditional marginal distributions by minimizing Fisher's divergence between the joint and marginal distributions. The theoretical advantages of CoInD are reflected in both qualitative and quantitative experiments, demonstrating a significantly more faithful and controlled generation of samples for arbitrary logical compositions of attributes. The benefit is more pronounced for scenarios that current solutions relying on the assumption of conditionally independent marginals struggle with, namely, logical compositions involving the NOT operation and when only a subset of compositions are observed during training.
Authors: Shilin Tong, Difei Wu, Xiaona Liu, Le Zheng, Yuchuan Du, Difan Zou
Abstract: Pavement distress significantly compromises road integrity and poses risks to drivers. Accurate prediction of pavement distress deterioration is essential for effective road management, cost reduction in maintenance, and improvement of traffic safety. However, real-world data on pavement distress is usually collected irregularly, resulting in uneven, asynchronous, and sparse spatial-temporal datasets. This hinders the application of existing spatial-temporal models, such as DCRNN, since they are only applicable to regularly and synchronously collected data. To overcome these challenges, we propose the Spatial-Temporal Graph Autoregression Network (STGAN), a novel graph neural network model designed for accurately predicting irregular pavement distress deterioration using complex spatial-temporal data. Specifically, STGAN integrates the temporal domain into the spatial domain, creating a larger graph where nodes are represented by spatial-temporal tuples and edges are formed based on a similarity-based connection mechanism. Furthermore, based on the constructed spatiotemporal graph, we formulate pavement distress deterioration prediction as a graph autoregression task, i.e., the graph size increases incrementally and the prediction is performed sequentially. This is accomplished by a novel spatial-temporal attention mechanism deployed by STGAN. Utilizing the ConTrack dataset, which contains pavement distress records collected from different locations in Shanghai, we demonstrate the superior performance of STGAN in capturing spatial-temporal correlations and addressing the aforementioned challenges. Experimental results further show that STGAN outperforms baseline models, and ablation studies confirm the effectiveness of its novel modules. Our findings contribute to promoting proactive road maintenance decision-making and ultimately enhancing road safety and resilience.
Authors: Xiaobin Hong, Jiawen Zhang, Wenzhong Li, Sanglu Lu, Jia Li
Abstract: The rise of foundation models has revolutionized natural language processing and computer vision, yet their best practices to time series forecasting remains underexplored. Existing time series foundation models often adopt methodologies from these fields without addressing the unique characteristics of time series data. In this paper, we identify two key challenges in cross-domain time series forecasting: the complexity of temporal patterns and semantic misalignment. To tackle these issues, we propose the ``Unify and Anchor" transfer paradigm, which disentangles frequency components for a unified perspective and incorporates external context as domain anchors for guided adaptation. Based on this framework, we introduce ContexTST, a Transformer-based model that employs a time series coordinator for structured representation and the Transformer blocks with a context-informed mixture-of-experts mechanism for effective cross-domain generalization. Extensive experiments demonstrate that ContexTST advances state-of-the-art forecasting performance while achieving strong zero-shot transferability across diverse domains.
Authors: Wenda Chu, Yang Song, Yisong Yue
Abstract: We study the problem of posterior sampling in discrete-state spaces using discrete diffusion models. While posterior sampling methods for continuous diffusion models have achieved remarkable progress, analogous methods for discrete diffusion models remain challenging. In this work, we introduce a principled plug-and-play discrete diffusion posterior sampling algorithm based on split Gibbs sampling, which we call SG-DPS. Our algorithm enables reward-guided generation and solving inverse problems in discrete-state spaces. We demonstrate that SG-DPS converges to the true posterior distribution on synthetic benchmarks, and enjoys state-of-the-art posterior sampling performance on a range of benchmarks for discrete data, achieving up to 2x improved performance compared to existing baselines.
Authors: Xiaoyuan Zhang, Xinyan Cai, Bo Liu, Weidong Huang, Song-Chun Zhu, Siyuan Qi, Yaodong Yang
Abstract: Differentiable environments have heralded new possibilities for learning control policies by offering rich differentiable information that facilitates gradient-based methods. In comparison to prevailing model-free reinforcement learning approaches, model-based reinforcement learning (MBRL) methods exhibit the potential to effectively harness the power of differentiable information for recovering the underlying physical dynamics. However, this presents two primary challenges: effectively utilizing differentiable information to 1) construct models with more accurate dynamic prediction and 2) enhance the stability of policy training. In this paper, we propose a Differentiable Information Enhanced MBRL method, MB-MIX, to address both challenges. Firstly, we adopt a Sobolev model training approach that penalizes incorrect model gradient outputs, enhancing prediction accuracy and yielding more precise models that faithfully capture system dynamics. Secondly, we introduce mixing lengths of truncated learning windows to reduce the variance in policy gradient estimation, resulting in improved stability during policy learning. To validate the effectiveness of our approach in differentiable environments, we provide theoretical analysis and empirical results. Notably, our approach outperforms previous model-based and model-free methods, in multiple challenging tasks involving controllable rigid robots such as humanoid robots' motion control and deformable object manipulation.
Authors: EungGu Yun, Heonjin Ha, Yeongwoo Nam, Bryan Dongik Lee
Abstract: This paper introduces LAFT, a novel feature transformation method designed to incorporate user knowledge and preferences into anomaly detection using natural language. Accurately modeling the boundary of normality is crucial for distinguishing abnormal data, but this is often challenging due to limited data or the presence of nuisance attributes. While unsupervised methods that rely solely on data without user guidance are common, they may fail to detect anomalies of specific interest. To address this limitation, we propose Language-Assisted Feature Transformation (LAFT), which leverages the shared image-text embedding space of vision-language models to transform visual features according to user-defined requirements. Combined with anomaly detection methods, LAFT effectively aligns visual features with user preferences, allowing anomalies of interest to be detected. Extensive experiments on both toy and real-world datasets validate the effectiveness of our method.
Authors: Wenhao Liang, Wei Emma Zhang, Lin Yue, Miao Xu, Olaf Maennel, Weitong Chen
Abstract: Kolmogorov Arnold Networks (KANs) are neural architectures inspired by the Kolmogorov Arnold representation theorem that leverage B Spline parameterizations for flexible, locally adaptive function approximation. Although KANs can capture complex nonlinearities beyond those modeled by standard MultiLayer Perceptrons (MLPs), they frequently exhibit miscalibrated confidence estimates manifesting as overconfidence in dense data regions and underconfidence in sparse areas. In this work, we systematically examine the impact of four critical hyperparameters including Layer Width, Grid Order, Shortcut Function, and Grid Range on the calibration of KANs. Furthermore, we introduce a novel TemperatureScaled Loss (TSL) that integrates a temperature parameter directly into the training objective, dynamically adjusting the predictive distribution during learning. Both theoretical analysis and extensive empirical evaluations on standard benchmarks demonstrate that TSL significantly reduces calibration errors, thereby improving the reliability of probabilistic predictions. Overall, our study provides actionable insights into the design of spline based neural networks and establishes TSL as a robust loss solution for enhancing calibration.
Authors: Yifan Feng, Shiquan Liu, Xiangmin Han, Shaoyi Du, Zongze Wu, Han Hu, Yue Gao
Abstract: Hypergraph neural networks (HGNNs) effectively model complex high-order relationships in domains like protein interactions and social networks by connecting multiple vertices through hyperedges, enhancing modeling capabilities, and reducing information loss. Developing foundation models for hypergraphs is challenging due to their distinct data, which includes both vertex features and intricate structural information. We present Hyper-FM, a Hypergraph Foundation Model for multi-domain knowledge extraction, featuring Hierarchical High-Order Neighbor Guided Vertex Knowledge Embedding for vertex feature representation and Hierarchical Multi-Hypergraph Guided Structural Knowledge Extraction for structural information. Additionally, we curate 10 text-attributed hypergraph datasets to advance research between HGNNs and LLMs. Experiments on these datasets show that Hyper-FM outperforms baseline methods by approximately 13.3\%, validating our approach. Furthermore, we propose the first scaling law for hypergraph foundation models, demonstrating that increasing domain diversity significantly enhances performance, unlike merely augmenting vertex and hyperedge counts. This underscores the critical role of domain diversity in scaling hypergraph models.
Authors: Daksh Mittal, Ang Li, Tzu-Ching Yen, Daniel Guetta, Hongseok Namkoong
Abstract: Autoregressive models have emerged as a powerful framework for modeling exchangeable sequences - i.i.d. observations when conditioned on some latent factor - enabling direct modeling of uncertainty from missing data (rather than a latent). Motivated by the critical role posterior inference plays as a subroutine in decision-making (e.g., active learning, bandits), we study the inferential and architectural inductive biases that are most effective for exchangeable sequence modeling. For the inference stage, we highlight a fundamental limitation of the prevalent single-step generation approach: inability to distinguish between epistemic and aleatoric uncertainty. Instead, a long line of works in Bayesian statistics advocates for multi-step autoregressive generation; we demonstrate this "correct approach" enables superior uncertainty quantification that translates into better performance on downstream decision-making tasks. This naturally leads to the next question: which architectures are best suited for multi-step inference? We identify a subtle yet important gap between recently proposed Transformer architectures for exchangeable sequences (Muller et al., 2022; Nguyen & Grover, 2022; Ye & Namkoong, 2024), and prove that they in fact cannot guarantee exchangeability despite introducing significant computational overhead. We illustrate our findings using controlled synthetic settings, demonstrating how custom architectures can significantly underperform standard causal masks, underscoring the need for new architectural innovations.
Authors: Bo Yang
Abstract: Large language models (LLMs) inadvertently memorize sensitive data from their massive pretraining corpora \cite{jang2022knowledge}. In this work, we propose CE-U (Cross Entropy Unlearning), a novel loss function designed specifically for unlearning tasks. CE-U addresses fundamental limitations of gradient ascent approaches which suffer from instability due to vanishing gradients when model confidence is high and gradient exploding when confidence is low. We also unify standard cross entropy supervision and cross entropy unlearning into a single framework. Notably, on the TOFU benchmark for unlearning \cite{maini2024tofu}, CE-U achieves state-of-the-art results on LLaMA2-7B with 1\% and 5\% forgetting, even without the use of any extra reference model or additional positive samples. Our theoretical analysis further reveals that the gradient instability issues also exist in popular reinforcement learning algorithms like DPO and GRPO, as they include a gradient ascent component. This suggests that applying CE-U principles to reinforcement learning could be a promising direction for improving stability and convergence.
Authors: Aleksandar Avdalovic, Joseph Khoury, Ahmad Taha, Elias Bou-Harb
Abstract: Water systems are vital components of modern infrastructure, yet they are increasingly susceptible to sophisticated cyber attacks with potentially dire consequences on public health and safety. While state-of-the-art machine learning techniques effectively detect anomalies, contemporary model-agnostic attack attribution methods using LIME, SHAP, and LEMNA are deemed impractical for large-scale, interdependent water systems. This is due to the intricate interconnectivity and dynamic interactions that define these complex environments. Such methods primarily emphasize individual feature importance while falling short of addressing the crucial sensor-actuator interactions in water systems, which limits their effectiveness in identifying root cause attacks. To this end, we propose a novel model-agnostic Factorization Machines (FM)-based approach that capitalizes on water system sensor-actuator interactions to provide granular explanations and attributions for cyber attacks. For instance, an anomaly in an actuator pump activity can be attributed to a top root cause attack candidates, a list of water pressure sensors, which is derived from the underlying linear and quadratic effects captured by our approach. We validate our method using two real-world water system specific datasets, SWaT and WADI, demonstrating its superior performance over traditional attribution methods. In multi-feature cyber attack scenarios involving intricate sensor-actuator interactions, our FM-based attack attribution method effectively ranks attack root causes, achieving approximately 20% average improvement over SHAP and LEMNA.
Authors: Seunghun Baek, Injun Choi, Mustafa Dere, Minjeong Kim, Guorong Wu, Won Hwa Kim
Abstract: Stacking excessive layers in DNN results in highly underdetermined system when training samples are limited, which is very common in medical applications. In this regard, we present a framework capable of deriving an efficient high-dimensional space with reasonable increase in model size. This is done by utilizing a transform (i.e., convolution) that leverages scale-space theory with covariance structure. The overall model trains on this transform together with a downstream classifier (i.e., Fully Connected layer) to capture the optimal multi-scale representation of the original data which corresponds to task-specific components in a dual space. Experiments on neuroimaging measures from Alzheimer's Disease Neuroimaging Initiative (ADNI) study show that our model performs better and converges faster than conventional models even when the model size is significantly reduced. The trained model is made interpretable using gradient information over the multi-scale transform to delineate personalized AD-specific regions in the brain.
Authors: Vegard Flovik, Sebastian Winter, Christian Agrell
Abstract: Engineering disciplines often rely on extensive simulations to ensure that structures are designed to withstand harsh conditions while avoiding over-engineering for unlikely scenarios. Assessments such as Serviceability Limit State (SLS) involve evaluating weather events, including estimating loads not expected to be exceeded more than a specified number of times (e.g., 100) throughout the structure's design lifetime. Although physics-based simulations provide robust and detailed insights, they are computationally expensive, making it challenging to generate statistically valid representations of a wide range of weather conditions. To address these challenges, we propose an approach using Gaussian Process (GP) surrogate models trained on a limited set of simulation outputs to directly generate the structural response distribution. We apply this method to an SLS assessment for estimating the order statistics \(Y_{100}\), representing the 100th highest response, of a structure exposed to 25 years of historical weather observations. Our results indicate that the GP surrogate models provide comparable results to full simulations but at a fraction of the computational cost.
Authors: Yuxin Wang, Botian Jiang, Yiran Guo, Quan Gan, David Wipf, Xuanjing Huang, Xipeng Qiu
Abstract: Prior-Fitted Networks (PFNs) have recently been proposed to efficiently perform tabular classification tasks. Although they achieve good performance on small datasets, they encounter limitations with larger datasets. These limitations include significant memory consumption and increased computational complexity, primarily due to the impracticality of incorporating all training samples as inputs within these networks. To address these challenges, we investigate the fitting assumption for PFNs and input samples. Building on this understanding, we propose \textit{BoostPFN} designed to enhance the performance of these networks, especially for large-scale datasets. We also theoretically validate the convergence of BoostPFN and our empirical results demonstrate that the BoostPFN method can outperform standard PFNs with the same size of training samples in large datasets and achieve a significant acceleration in training times compared to other established baselines in the field, including widely-used Gradient Boosting Decision Trees (GBDTs), deep learning methods and AutoML systems. High performance is maintained for up to 50x of the pre-training size of PFNs, substantially extending the limit of training samples. Through this work, we address the challenges of efficiently handling large datasets via PFN-based models, paving the way for faster and more effective tabular data classification training and prediction process. Code is available at Github.
Authors: Yuhan Jing, Jingyu Wang, Lei Zhang, Haifeng Sun, Bo He, Zirui Zhuang, Chengsen Wang, Qi Qi, Jianxin Liao
Abstract: With the growing adoption of time-series anomaly detection (TAD) technology, numerous studies have employed deep learning-based detectors for analyzing time-series data in the fields of Internet services, industrial systems, and sensors. The selection and optimization of anomaly detectors strongly rely on the availability of an effective performance evaluation method for TAD. Since anomalies in time-series data often manifest as a sequence of points, conventional metrics that solely consider the detection of individual point are inadequate. Existing evaluation methods for TAD typically employ point-based or event-based metrics to capture the temporal context. However, point-based metrics tend to overestimate detectors that excel only in detecting long anomalies, while event-based metrics are susceptible to being misled by fragmented detection results. To address these limitations, we propose OIPR, a novel set of TAD evaluation metrics. It models the process of operators receiving detector alarms and handling faults, utilizing area under the operator interest curve to evaluate the performance of TAD algorithms. Furthermore, we build a special scenario dataset to compare the characteristics of different evaluation methods. Through experiments conducted on the special scenario dataset and five real-world datasets, we demonstrate the remarkable performance of OIPR in extreme and complex scenarios. It achieves a balance between point and event perspectives, overcoming their primary limitations and offering applicability to broader situations.
Authors: Qi Li, Runpeng Yu, Xinchao Wang
Abstract: Parameter-level model merging is an emerging paradigm in multi-task learning with significant promise. Previous research has explored its connections with prediction-level model ensembling-commonly viewed as the upper bound for merging-to reveal the potential of achieving performance consistency between the two. However, this observation relies on certain preconditions, such as being limited to two models, using ViT-based models, and all models are fine-tuned from the same pre-trained checkpoint. To further understand the intrinsic connections between model merging and model ensembling, this paper explores an interesting possibility: If these restrictions are removed, can performance consistency still be achieved between merging and ensembling? To answer this question, we first theoretically establish a performance correlation between merging and ensembling. We find that even when previous restrictions are not met, there is still a way for model merging to attain a near-identical and superior performance similar to that of ensembling. To verify whether our findings are practical, we introduce a validation framework termed Neural Ligand (NeuLig). The learning process of NeuLig is meticulously designed with a specialized loss function supported by theoretical foundations. Experimental results demonstrate the robust resilience of NeuLig in terms of both model scale and the number of collaborating models. For instance, for the case involving 5 CLIP-ViT-B/32 models, parameter-level merging achieves the same performance as prediction-level ensembling (merging: 95.44% vs. ensembling: 95.46%).
Authors: Yogesh Verma, Ayush Bharti, Vikas Garg
Abstract: Simulation-based inference (SBI) methods typically require fully observed data to infer parameters of models with intractable likelihood functions. However, datasets often contain missing values due to incomplete observations, data corruptions (common in astrophysics), or instrument limitations (e.g., in high-energy physics applications). In such scenarios, missing data must be imputed before applying any SBI method. We formalize the problem of missing data in SBI and demonstrate that naive imputation methods can introduce bias in the estimation of SBI posterior. We also introduce a novel amortized method that addresses this issue by jointly learning the imputation model and the inference network within a neural posterior estimation (NPE) framework. Extensive empirical results on SBI benchmarks show that our approach provides robust inference outcomes compared to standard baselines for varying levels of missing data. Moreover, we demonstrate the merits of our imputation model on two real-world bioactivity datasets (Adrenergic and Kinase assays). Code is available at https://github.com/Aalto-QuML/RISE.
Authors: Andreas Sauter, Saber Salehkaleybar, Aske Plaat, Erman Acar
Abstract: Predicting the distribution of outcomes under hypothetical interventions is crucial in domains like healthcare, economics, and policy-making. Current methods often rely on strong assumptions, such as known causal graphs or parametric models, and lack amortization across problem instances, limiting their practicality. We propose a novel transformer-based conditional variational autoencoder architecture, named ACTIVA, that extends causal transformer encoders to predict causal effects as mixtures of Gaussians. Our method requires no causal graph and predicts interventional distributions given only observational data and a queried intervention. By amortizing over many simulated instances, it enables zero-shot generalization to novel datasets without retraining. Experiments demonstrate accurate predictions for synthetic and semi-synthetic data, showcasing the effectiveness of our graph-free, amortized causal inference approach.
Authors: Anmol Biswas, Raghav Singhal, Sivakumar Elangovan, Shreyas Sabnis, Udayan Ganguly
Abstract: Efficient inference is critical for deploying deep learning models on edge AI devices. Low-bit quantization (e.g., 3- and 4-bit) with fixed-point arithmetic improves efficiency, while low-power memory technologies like analog nonvolatile memory enable further gains. However, these methods introduce non-ideal hardware behavior, including bit faults and device-to-device variability. We propose a regularization-based quantization-aware training (QAT) framework that supports fixed, learnable step-size, and learnable non-uniform quantization, achieving competitive results on CIFAR-10 and ImageNet. Our method also extends to Spiking Neural Networks (SNNs), demonstrating strong performance on 4-bit networks on CIFAR10-DVS and N-Caltech 101. Beyond quantization, our framework enables fault and variability-aware fine-tuning, mitigating stuck-at faults (fixed weight bits) and device resistance variability. Compared to prior fault-aware training, our approach significantly improves performance recovery under upto 20% bit-fault rate and 40% device-to-device variability. Our results establish a generalizable framework for quantization and robustness-aware training, enhancing efficiency and reliability in low-power, non-ideal hardware.
Authors: Yifang Chen, Xuyang Guo, Xiaoyu Li, Yingyu Liang, Zhenmei Shi, Zhao Song
Abstract: Recently, Large Language Models (LLMs) have achieved remarkable success. A key factor behind this success is the scaling law observed by OpenAI. Specifically, for models with Transformer architecture, the test loss exhibits a power-law relationship with model size, dataset size, and the amount of computation used in training, demonstrating trends that span more than seven orders of magnitude. This scaling law challenges traditional machine learning wisdom, notably the Oscar Scissors principle, which suggests that an overparametrized algorithm will overfit the training datasets, resulting in poor test performance. Recent research has also identified the scaling law in simpler machine learning contexts, such as linear regression. However, fully explaining the scaling law in large practical models remains an elusive goal. In this work, we advance our understanding by demonstrating that the scaling law phenomenon extends to multiple regression and kernel regression settings, which are significantly more expressive and powerful than linear methods. Our analysis provides deeper insights into the scaling law, potentially enhancing our understanding of LLMs.
Authors: Zhiyin Li, Yubo Yang, Tao Yang, Xiaofeng Wu, Ziyu Guo, Bo Hu
Abstract: Federated learning enables distributed model training across clients under central coordination without raw data exchange. However, in wireless implementations, frequent parameter updates between the server and clients create significant communication overhead. While existing research assumes known channel state information (CSI) or stationary distributions, practical wireless channels exhibit non-stationary characteristics due to channel fading, user mobility, and hostile attacks. The unavailability of CSI and time-varying statistics can cause unpredictable transmission failures, exacerbating client staleness and affecting model convergence. To address these challenges, we propose an asynchronous federated learning scheduling framework for non-stationary channel environments to reduce staleness while promoting fair and efficient communication and aggregation.We focus on two channel scenarios: extremely non-stationary and piecewise stationary. Age of Information (AoI) quantifies client staleness under non-stationary conditions. Through a rigorous convergence analysis, we explore how AoI and per-round client participation affect learning performance. The scheduling problem is modeled within a multi-armed bandit (MAB) framework, and we derive the theoretical lower bounds on AoI regret. Based on these findings, we develop scheduling strategies for both scenarios using the GLR-CUCB and M-exp3 algorithms, also deriving their respective upper bounds on AoI regret. To address imbalanced client updates, we introduce an adaptive allocation strategy that incorporates marginal utility and fairness. Simulations demonstrate that our algorithm reduces AoI regret growth, accelerates federated learning convergence, and promotes fairer aggregation.
Authors: Xinyi Wan, Penghui Qi, Guangxing Huang, Jialin Li, Min Lin
Abstract: Pipeline parallelism (PP) is widely used for training large language models (LLMs), yet its scalability is often constrained by high activation memory consumption as the number of in-flight microbatches grows with the degree of PP. In this paper, we focus on addressing this challenge by leveraging the under-explored memory offload strategy in PP. With empirical study, we discover that in the majority of standard configurations, at least half, and potentially all, of the activations can be offloaded with negligible overhead. In the cases where full overload is not possible, we introduce a novel selective offload strategy that decreases peak activation memory in a better-than-linear manner. Furthermore, we integrate memory offload with other techniques to jointly consider overall throughput and memory limitation. Our experiments proves that the per-device activation memory effectively reduces with the total number of stages, making PP a stronger alternative than TP, offering up to a 19\% acceleration with even lower memory consumption. The implementation is open-sourced at \href{https://github.com/sail-sg/zero-bubble-pipeline-parallelism}{this url}.
URLs: https://github.com/sail-sg/zero-bubble-pipeline-parallelism
Authors: Anh Tong, Thanh Nguyen-Tang, Dongeun Lee, Duc Nguyen, Toan Tran, David Hall, Cheongwoong Kang, Jaesik Choi
Abstract: Recent advancements in large language models (LLMs) based on transformer architectures have sparked significant interest in understanding their inner workings. In this paper, we introduce a novel approach to modeling transformer architectures using highly flexible non-autonomous neural ordinary differential equations (ODEs). Our proposed model parameterizes all weights of attention and feed-forward blocks through neural networks, expressing these weights as functions of a continuous layer index. Through spectral analysis of the model's dynamics, we uncover an increase in eigenvalue magnitude that challenges the weight-sharing assumption prevalent in existing theoretical studies. We also leverage the Lyapunov exponent to examine token-level sensitivity, enhancing model interpretability. Our neural ODE transformer demonstrates performance comparable to or better than vanilla transformers across various configurations and datasets, while offering flexible fine-tuning capabilities that can adapt to different architectural constraints.
Authors: Hazem Hesham Yousef Shalby, Manuel Roveri
Abstract: Human activity recognition (HAR) is a research field that employs Machine Learning (ML) techniques to identify user activities. Recent studies have prioritized the development of HAR solutions directly executed on wearable devices, enabling the on-device activity recognition. This approach is supported by the Tiny Machine Learning (TinyML) paradigm, which integrates ML within embedded devices with limited resources. However, existing approaches in the field lack in the capability for on-device learning of new HAR tasks, particularly when supervised data are scarce. To address this limitation, our paper introduces Dendron, a novel TinyML methodology designed to facilitate the on-device learning of new tasks for HAR, even in conditions of limited supervised data. Experimental results on two public-available datasets and an off-the-shelf device (STM32-NUCLEO-F401RE) show the effectiveness and efficiency of the proposed solution.
Authors: Yongqi Huang, Peng Ye, Chenyu Huang, Jianjian Cao, Lin Zhang, Baopu Li, Gang Yu, Tao Chen
Abstract: Upcycled Mixture-of-Experts (MoE) models have shown great potential in various tasks by converting the original Feed-Forward Network (FFN) layers in pre-trained dense models into MoE layers. However, these models still suffer from significant parameter inefficiency due to the introduction of multiple experts. In this work, we propose a novel DeRS (Decompose, Replace, and Synthesis) paradigm to overcome this shortcoming, which is motivated by our observations about the unique redundancy mechanisms of upcycled MoE experts. Specifically, DeRS decomposes the experts into one expert-shared base weight and multiple expert-specific delta weights, and subsequently represents these delta weights in lightweight forms. Our proposed DeRS paradigm can be applied to enhance parameter efficiency in two different scenarios, including: 1) DeRS Compression for inference stage, using sparsification or quantization to compress vanilla upcycled MoE models; and 2) DeRS Upcycling for training stage, employing lightweight sparse or low-rank matrixes to efficiently upcycle dense models into MoE models. Extensive experiments across three different tasks show that the proposed methods can achieve extreme parameter efficiency while maintaining the performance for both training and compression of upcycled MoE models.
Authors: Daniil Sherki, Ivan Oseledets, Ekaterina Muravleva
Abstract: Solving Bayesian inverse problems efficiently remains a significant challenge due to the complexity of posterior distributions and the computational cost of traditional sampling methods. Given a series of observations and the forward model, we want to recover the distribution of the parameters, conditioned on observed experimental data. We show, that combining Conditional Flow Mathching (CFM) with transformer-based architecture, we can efficiently sample from such kind of distribution, conditioned on variable number of observations.
Authors: Peng Yan, Ahmed Abdulkadir, Gerrit A. Schatte, Giulia Anguzzi, Joonsu Gha, Nikola Pascher, Matthias Rosenthal, Yunlong Gao, Benjamin F. Grewe, Thilo Stadelmann
Abstract: To go from (passive) process monitoring to active process control, an effective AI system must learn about the behavior of the complex system from very limited training data, forming an ad-hoc digital twin with respect to process in- and outputs that captures the consequences of actions on the process's world. We propose a novel methodology based on learning world models that disentangles process parameters in the learned latent representation, allowing for fine-grained control. Representation learning is driven by the latent factors that influence the processes through contrastive learning within a joint embedding predictive architecture. This makes changes in representations predictable from changes in inputs and vice versa, facilitating interpretability of key factors responsible for process variations, paving the way for effective control actions to keep the process within operational bounds. The effectiveness of our method is validated on the example of plastic injection molding, demonstrating practical relevance in proposing specific control actions for a notoriously unstable process.
Authors: Max Eissler, Tim Korjakow, Stefan Ganscha, Oliver T. Unke, Klaus-Robert M\"uller, Stefan Gugler
Abstract: Most current neural networks for molecular dynamics (MD) include physical inductive biases, resulting in specialized and complex architectures. This is in contrast to most other machine learning domains, where specialist approaches are increasingly replaced by general-purpose architectures trained on vast datasets. In line with this trend, several recent studies have questioned the necessity of architectural features commonly found in MD models, such as built-in rotational equivariance or energy conservation. In this work, we contribute to the ongoing discussion by evaluating the performance of an MD model with as few specialized architectural features as possible. We present a recipe for MD using an Edge Transformer, an ``off-the-shelf'' transformer architecture that has been minimally modified for the MD domain, termed MD-ET. Our model implements neither built-in equivariance nor energy conservation. We use a simple supervised pre-training scheme on $\sim$30 million molecular structures from the QCML database. Using this ``off-the-shelf'' approach, we show state-of-the-art results on several benchmarks after fine-tuning for a small number of steps. Additionally, we examine the effects of being only approximately equivariant and energy conserving for MD simulations, proposing a novel method for distinguishing the errors resulting from non-equivariance from other sources of inaccuracies like numerical rounding errors. While our model exhibits runaway energy increases on larger structures, we show approximately energy-conserving NVE simulations for a range of small structures.
Authors: Th\'eo Vincent, Tim Faust, Yogesh Tripathi, Jan Peters, Carlo D'Eramo
Abstract: Recent works have successfully demonstrated that sparse deep reinforcement learning agents can be competitive against their dense counterparts. This opens up opportunities for reinforcement learning applications in fields where inference time and memory requirements are cost-sensitive or limited by hardware. Until now, dense-to-sparse methods have relied on hand-designed sparsity schedules that are not synchronized with the agent's learning pace. Crucially, the final sparsity level is chosen as a hyperparameter, which requires careful tuning as setting it too high might lead to poor performances. In this work, we address these shortcomings by crafting a dense-to-sparse algorithm that we name Eau De $Q$-Network (EauDeQN). To increase sparsity at the agent's learning pace, we consider multiple online networks with different sparsity levels, where each online network is trained from a shared target network. At each target update, the online network with the smallest loss is chosen as the next target network, while the other networks are replaced by a pruned version of the chosen network. We evaluate the proposed approach on the Atari $2600$ benchmark and the MuJoCo physics simulator, showing that EauDeQN reaches high sparsity levels while keeping performances high.
Authors: Hyungho Na, Kwanghyeon Lee, Sumin Lee, Il-Chul Moon
Abstract: In the context of multi-agent reinforcement learning, generalization is a challenge to solve various tasks that may require different joint policies or coordination without relying on policies specialized for each task. We refer to this type of problem as a multi-task, and we train agents to be versatile in this multi-task setting through a single training process. To address this challenge, we introduce TRajectory-class-Aware Multi-Agent reinforcement learning (TRAMA). In TRAMA, agents recognize a task type by identifying the class of trajectories they are experiencing through partial observations, and the agents use this trajectory awareness or prediction as additional information for action policy. To this end, we introduce three primary objectives in TRAMA: (a) constructing a quantized latent space to generate trajectory embeddings that reflect key similarities among them; (b) conducting trajectory clustering using these trajectory embeddings; and (c) building a trajectory-class-aware policy. Specifically for (c), we introduce a trajectory-class predictor that performs agent-wise predictions on the trajectory class; and we design a trajectory-class representation model for each trajectory class. Each agent takes actions based on this trajectory-class representation along with its partial observation for task-aware execution. The proposed method is evaluated on various tasks, including multi-task problems built upon StarCraft II. Empirical results show further performance improvements over state-of-the-art baselines.
Authors: Zekang Wang, Zhe He, Edan Toledo, Steven Morad
Abstract: We introduce POPGym Arcade, a benchmark consisting of 7 pixel-based environments each with three difficulties, utilizing a single observation and action space. Each environment offers both fully observable and partially observable variants, enabling counterfactual studies on partial observability. POPGym Arcade utilizes JIT compilation on hardware accelerators to achieve substantial speedups over CPU-bound environments. Moreover, this enables Podracer-style architectures to further increase hardware utilization and training speed. We evaluate memory models on our environments using a Podracer variant of Q learning, and examine the results. Finally, we generate memory saliency maps, uncovering how memories propagate through policies. Our library is available at https://github.com/bolt-research/ popgym_arcade.
Authors: Xi He, Max A. Little
Abstract: In this paper, we introduce a generic data structure called decision trees, which integrates several well-known data structures, including binary search trees, K-D trees, binary space partition trees, and decision tree models from machine learning. We provide the first axiomatic definition of decision trees. These axioms establish a firm mathematical foundation for studying decision tree problems. We refer to decision trees that satisfy the axioms as proper decision trees. We prove that only proper decision trees can be uniquely characterized as K-permutations. Since permutations are among the most well-studied combinatorial structures, this characterization provides a fundamental basis for analyzing the combinatorial and algorithmic properties of decision trees. As a result of this advancement, we develop the first provably correct polynomial-time algorithm for solving the optimal decision tree problem. Our algorithm is derived using a formal program derivation framework, which enables step-by-step equational reasoning to construct side-effect-free programs with guaranteed correctness. The derived algorithm is correct by construction and is applicable to decision tree problems defined by any splitting rules that adhere to the axioms and any objective functions that can be specified in a given form. Examples include the decision tree problems where splitting rules are defined by axis-parallel hyperplanes, arbitrary hyperplanes, and hypersurfaces. By extending the axioms, we can potentially address a broader range of problems. Moreover, the derived algorithm can easily accommodate various constraints, such as tree depth and leaf size, and is amenable to acceleration techniques such as thinning method.
Authors: Huifeng Yin, Yu Zhao, Minghao Wu, Xuanfan Ni, Bo Zeng, Hao Wang, Tianqi Shi, Liangying Shao, Chenyang Lyu, Longyue Wang, Weihua Luo, Kaifu Zhang
Abstract: Large Reasoning Models(LRMs) such as OpenAI o1 and DeepSeek-R1 have shown remarkable reasoning capabilities by scaling test-time compute and generating long Chain-of-Thought(CoT). Distillation--post-training on LRMs-generated data--is a straightforward yet effective method to enhance the reasoning abilities of smaller models, but faces a critical bottleneck: we found that distilled long CoT data poses learning difficulty for small models and leads to the inheritance of biases (i.e. over-thinking) when using Supervised Fine-tuning(SFT) and Reinforcement Learning(RL) methods. To alleviate this bottleneck, we propose constructing tree-based CoT data from scratch via Monte Carlo Tree Search(MCTS). We then exploit a set of CoT-aware approaches, including Thoughts Length Balance, Fine-grained DPO, and Joint Post-training Objective, to enhance SFT and RL on the construted data.
Authors: Abdullah Akg\"ul, Gulcin Baykal, Manuel Hau{\ss}mann, Melih Kandemir
Abstract: On-policy reinforcement learning algorithms use the most recently learned policy to interact with the environment and update it using the latest gathered trajectories, making them well-suited for adapting to non-stationary environments where dynamics change over time. However, previous studies show that they struggle to maintain plasticity$\unicode{x2013}$the ability of neural networks to adjust their synaptic connections$\unicode{x2013}$with overfitting identified as the primary cause. To address this, we present the first application of evidential learning in an on-policy reinforcement learning setting: $\textit{Evidential Proximal Policy Optimization (EPPO)}$. EPPO incorporates all sources of error in the critic network's approximation$\unicode{x2013}$i.e., the baseline function in advantage calculation$\unicode{x2013}$by modeling the epistemic and aleatoric uncertainty contributions to the approximation's total variance. We achieve this by using an evidential neural network, which serves as a regularizer to prevent overfitting. The resulting probabilistic interpretation of the advantage function enables optimistic exploration, thus maintaining the plasticity. Through experiments on non-stationary continuous control tasks, where the environment dynamics change at regular intervals, we demonstrate that EPPO outperforms state-of-the-art on-policy reinforcement learning variants in both task-specific and overall return.
Authors: Mohammad Sadegh Akhondzadeh, Aleksandar Bojchevski, Evangelos Eleftheriou, Martino Dazzi
Abstract: One of the challenges of quantizing a large language model (LLM) is the presence of outliers. Outliers often make uniform quantization schemes less effective, particularly in extreme cases such as 4-bit quantization. We introduce KurTail, a new post-training quantization (PTQ) scheme that leverages Kurtosis-based rotation to mitigate outliers in the activations of LLMs. Our method optimizes Kurtosis as a measure of tailedness. This approach enables the quantization of weights, activations, and the KV cache in 4 bits. We utilize layer-wise optimization, ensuring memory efficiency. KurTail outperforms existing quantization methods, offering a 13.3\% boost in MMLU accuracy and a 15.5\% drop in Wiki perplexity compared to QuaRot. It also outperforms SpinQuant with a 2.6\% MMLU gain and reduces perplexity by 2.9\%, all while reducing the training cost. For comparison, learning the rotation using SpinQuant for Llama3-70B requires at least four NVIDIA H100 80GB GPUs, whereas our method requires only a single GPU, making it a more accessible solution for consumer GPU.
Authors: Yifan Niu, Ziqi Gao, Tingyang Xu, Yang Liu, Yatao Bian, Yu Rong, Junzhou Huang, Jia Li
Abstract: Exploring chemical space to find novel molecules that simultaneously satisfy multiple properties is crucial in drug discovery. However, existing methods often struggle with trading off multiple properties due to the conflicting or correlated nature of chemical properties. To tackle this issue, we introduce InversionGNN framework, an effective yet sample-efficient dual-path graph neural network (GNN) for multi-objective drug discovery. In the direct prediction path of InversionGNN, we train the model for multi-property prediction to acquire knowledge of the optimal combination of functional groups. Then the learned chemical knowledge helps the inversion generation path to generate molecules with required properties. In order to decode the complex knowledge of multiple properties in the inversion path, we propose a gradient-based Pareto search method to balance conflicting properties and generate Pareto optimal molecules. Additionally, InversionGNN is able to search the full Pareto front approximately in discrete chemical space. Comprehensive experimental evaluations show that InversionGNN is both effective and sample-efficient in various discrete multi-objective settings including drug discovery.
Authors: Yufeng Yuan, Yu Yue, Ruofei Zhu, Tiantian Fan, Lin Yan
Abstract: Reinforcement learning (RL) is pivotal for enabling large language models (LLMs) to generate long chains of thought (CoT) for complex tasks like math and reasoning. However, Proximal Policy Optimization (PPO), effective in many RL scenarios, fails in long CoT tasks. This paper identifies that value initialization bias and reward signal decay are the root causes of PPO's failure. We propose Value-Calibrated PPO (VC-PPO) to address these issues. In VC-PPO, the value model is pretrained to tackle initialization bias, and the Generalized Advantage Estimation (GAE) computation is decoupled between the actor and critic to mitigate reward signal decay. Experiments on the American Invitational Mathematics Examination (AIME) show that VC-PPO significantly boosts PPO performance. Ablation studies show that techniques in VC-PPO are essential in enhancing PPO for long CoT tasks.
Authors: Ramkrishna Acharya
Abstract: This study reviews some of the popular stochastic gradient-based schemes based on large least-square problems. These schemes, often called optimizers in machine learning play a crucial role in finding better parameters of a model. Hence this study focuses on viewing such optimizers with different hyper-parameters and analyzing them based on least square problems. Codes that produced results in this work are available on https://github.com/q-viper/gradients-based-methods-on-large-least-square.
URLs: https://github.com/q-viper/gradients-based-methods-on-large-least-square.
Authors: Yuval Ben Dror
Abstract: Over recent decades, extensive research has aimed to overcome the restrictive underlying assumptions required for a Generalized Linear Model to generate accurate and meaningful predictions. These efforts include regularizing coefficients, selecting features, and clustering ordinal categories, among other approaches. Despite these advances, efficiently clustering nominal categories in GLMs without incurring high computational costs remains a challenge. This paper introduces Ranking to Variable Fusion (R2VF), a two-step method designed to efficiently fuse nominal and ordinal categories in GLMs. By first transforming nominal features into an ordinal framework via regularized regression and then applying variable fusion, R2VF strikes a balance between model complexity and interpretability. We demonstrate the effectiveness of R2VF through comparisons with other methods, highlighting its performance in addressing overfitting and finding a proper set of covariates.
Authors: Liang Wu, Ruixi Hu, Yunwen Lei
Abstract: Pairwise learning includes various machine learning tasks, with ranking and metric learning serving as the primary representatives. While randomized coordinate descent (RCD) is popular in various learning problems, there is much less theoretical analysis on the generalization behavior of models trained by RCD, especially under the pairwise learning framework. In this paper, we consider the generalization of RCD for pairwise learning. We measure the on-average argument stability for both convex and strongly convex objective functions, based on which we develop generalization bounds in expectation. The early-stopping strategy is adopted to quantify the balance between estimation and optimization. Our analysis further incorporates the low-noise setting into the excess risk bound to achieve the optimistic bound as $O(1/n)$, where $n$ is the sample size.
Authors: Gilad Yehudai, Noah Amsel, Joan Bruna
Abstract: We study and compare the expressive power of transformers, RNNs, and transformers with chain of thought tokens on a simple and natural class of problems we term Compositional Reasoning Questions (CRQ). This family captures problems like evaluating Boolean formulas and multi-step word problems. Assuming standard hardness assumptions from circuit complexity and communication complexity, we prove that none of these three architectures is capable of solving CRQs unless some hyperparameter (depth, embedding dimension, and number of chain of thought tokens, respectively) grows with the size of the input. We also provide a construction for each architecture that solves CRQs. For transformers, our construction uses depth that is logarithmic in the problem size. For RNNs, logarithmic embedding dimension is necessary and sufficient, so long as the inputs are provided in a certain order. (Otherwise, a linear dimension is necessary). For transformers with chain of thought, our construction uses $n$ CoT tokens. These results show that, while CRQs are inherently hard, there are several different ways for language models to overcome this hardness. Even for a single class of problems, each architecture has strengths and weaknesses, and none is strictly better than the others.
Authors: Yao Zou, Dawei Cheng
Abstract: Credit card fraud imposes significant costs on both cardholders and issuing banks. Fraudsters often disguise their crimes, such as using legitimate transactions through several benign users to bypass anti-fraud detection. Existing graph neural network (GNN) models struggle with learning features of camouflaged, indirect multi-hop transactions due to their inherent over-smoothing issues in deep multi-layer aggregation, presenting a major challenge in detecting disguised relationships. Therefore, in this paper, we propose a novel High-order Graph Representation Learning model (HOGRL) to avoid incorporating excessive noise during the multi-layer aggregation process. In particular, HOGRL learns different orders of \emph{pure} representations directly from high-order transaction graphs. We realize this goal by effectively constructing high-order transaction graphs first and then learning the \emph{pure} representations of each order so that the model could identify fraudsters' multi-hop indirect transactions via multi-layer \emph{pure} feature learning. In addition, we introduce a mixture-of-expert attention mechanism to automatically determine the importance of different orders for jointly optimizing fraud detection performance. We conduct extensive experiments in both the open source and real-world datasets, the result demonstrates the significant improvements of our proposed HOGRL compared with state-of-the-art fraud detection baselines. HOGRL's superior performance also proves its effectiveness in addressing high-order fraud camouflage criminals.
Authors: Kai Fang, Jiangtao Deng, Chengzu Dong, Usman Naseem, Tongcun Liu, Hailin Feng, Wei Wang
Abstract: Frequent fluctuations of client nodes in highly dynamic mobile clusters can lead to significant changes in feature space distribution and data drift, posing substantial challenges to the robustness of existing federated learning (FL) strategies. To address these issues, we proposed a mobile cluster federated learning framework (MoCFL). MoCFL enhances feature aggregation by introducing an affinity matrix that quantifies the similarity between local feature extractors from different clients, addressing dynamic data distribution changes caused by frequent client churn and topology changes. Additionally, MoCFL integrates historical and current feature information when training the global classifier, effectively mitigating the catastrophic forgetting problem frequently encountered in mobile scenarios. This synergistic combination ensures that MoCFL maintains high performance and stability in dynamically changing mobile environments. Experimental results on the UNSW-NB15 dataset show that MoCFL excels in dynamic environments, demonstrating superior robustness and accuracy while maintaining reasonable training costs.
Authors: Hanmo Liu, Shimin Di, Haoyang Li, Xun Jian, Yue Wang, Lei Chen
Abstract: Node classification is a key task in temporal graph learning (TGL). Real-life temporal graphs often introduce new node classes over time, but existing TGL methods assume a fixed set of classes. This assumption brings limitations, as updating models with full data is costly, while focusing only on new classes results in forgetting old ones. Graph continual learning (GCL) methods mitigate forgetting using old-class subsets but fail to account for their evolution. We define this novel problem as temporal graph continual learning (TGCL), which focuses on efficiently maintaining up-to-date knowledge of old classes. To tackle TGCL, we propose a selective learning framework that substitutes the old-class data with its subsets, Learning Towards the Future (LTF). We derive an upper bound on the error caused by such replacement and transform it into objectives for selecting and learning subsets that minimize classification error while preserving the distribution of the full old-class data. Experiments on three real-world datasets validate the effectiveness of LTF on TGCL.
Authors: Yuhao Zhou, Sirui Song, Boyang Liu, Zhiheng Xi, Senjie Jin, Xiaoran Fan, Zhihao Zhang, Wei Li, Xuanjing Huang
Abstract: Rotary Position Embedding (RoPE) enables each attention head to capture multi-frequency information along the sequence dimension and is widely applied in foundation models. However, the nonlinearity introduced by RoPE complicates optimization of the key state in the Key-Value (KV) cache for RoPE-based attention. Existing KV cache compression methods typically store key state before rotation and apply the transformation during decoding, introducing additional computational overhead. This paper introduces EliteKV, a flexible modification framework for RoPE-based models supporting variable KV cache compression ratios. EliteKV first identifies the intrinsic frequency preference of each head using RoPElite, selectively restoring linearity to certain dimensions of key within attention computation. Building on this, joint low-rank compression of key and value enables partial cache sharing. Experimental results show that with minimal uptraining on only $0.6\%$ of the original training data, RoPE-based models achieve a $75\%$ reduction in KV cache size while preserving performance within a negligible margin. Furthermore, EliteKV consistently performs well across models of different scales within the same family.
Authors: Masih Eskandar, Tooba Imtiaz, Davin Hill, Zifeng Wang, Jennifer Dy
Abstract: Humans can naturally learn new and varying tasks in a sequential manner. Continual learning is a class of learning algorithms that updates its learned model as it sees new data (on potentially new tasks) in a sequence. A key challenge in continual learning is that as the model is updated to learn new tasks, it becomes susceptible to catastrophic forgetting, where knowledge of previously learned tasks is lost. A popular approach to mitigate forgetting during continual learning is to maintain a small buffer of previously-seen samples and to replay them during training. However, this approach is limited by the small buffer size, and while forgetting is reduced, it is still present. In this paper, we propose a novel loss function, STAR, that exploits the worst-case parameter perturbation that reduces the KL-divergence of model predictions with that of its local parameter neighborhood to promote stability and alleviate forgetting. STAR can be combined with almost any existing rehearsal-based method as a plug-and-play component. We empirically show that STAR consistently improves the performance of existing methods by up to 15% across varying baselines and achieves superior or competitive accuracy to that of state-of-the-art methods aimed at improving rehearsal-based continual learning.
Authors: Masoud Kavian, Milad Sefidgaran, Abdellatif Zaidi, Romain Chor
Abstract: In this paper, we investigate the effect of data heterogeneity across clients on the performance of distributed learning systems, i.e., one-round Federated Learning, as measured by the associated generalization error. Specifically, \(K\) clients have each \(n\) training samples generated independently according to a possibly different data distribution and their individually chosen models are aggregated by a central server. We study the effect of the discrepancy between the clients' data distributions on the generalization error of the aggregated model. First, we establish in-expectation and tail upper bounds on the generalization error in terms of the distributions. In part, the bounds extend the popular Conditional Mutual Information (CMI) bound which was developed for the centralized learning setting, i.e., \(K=1\), to the distributed learning setting with arbitrary number of clients $K \geq 1$. Then, we use a connection with information theoretic rate-distortion theory to derive possibly tighter \textit{lossy} versions of these bounds. Next, we apply our lossy bounds to study the effect of data heterogeneity across clients on the generalization error for distributed classification problem in which each client uses Support Vector Machines (D-SVM). In this case, we establish explicit generalization error bounds which depend explicitly on the data heterogeneity degree. It is shown that the bound gets smaller as the degree of data heterogeneity across clients gets higher, thereby suggesting that D-SVM generalizes better when the dissimilarity between the clients' training samples is bigger. This finding, which goes beyond D-SVM, is validated experimentally through a number of experiments.
Authors: Henrik Nolte, Mich\`ele Finck, Kristof Meding
Abstract: Does GPT know you? The answer depends on your level of public recognition; however, if your information was available on a website, the answer is probably yes. All Large Language Models (LLMs) memorize training data to some extent. If an LLM training corpus includes personal data, it also memorizes personal data. Developing an LLM typically involves processing personal data, which falls directly within the scope of data protection laws. If a person is identified or identifiable, the implications are far-reaching: the AI system is subject to EU General Data Protection Regulation requirements even after the training phase is concluded. To back our arguments: (1.) We reiterate that LLMs output training data at inference time, be it verbatim or in generalized form. (2.) We show that some LLMs can thus be considered personal data on their own. This triggers a cascade of data protection implications such as data subject rights, including rights to access, rectification, or erasure. These rights extend to the information embedded with-in the AI model. (3.) This paper argues that machine learning researchers must acknowledge the legal implications of LLMs as personal data throughout the full ML development lifecycle, from data collection and curation to model provision on, e.g., GitHub or Hugging Face. (4.) We propose different ways for the ML research community to deal with these legal implications. Our paper serves as a starting point for improving the alignment between data protection law and the technical capabilities of LLMs. Our findings underscore the need for more interaction between the legal domain and the ML community.
Authors: Hamidreza Mirkhani, Behzad Khamidehi, Ehsan Ahmadi, Fazel Arasteh, Mohammed Elmahgiubi, Weize Zhang, Umar Rajguru, Kasra Rezaee
Abstract: In this paper, we introduce CAPS (Context-Aware Priority Sampling), a novel method designed to enhance data efficiency in learning-based autonomous driving systems. CAPS addresses the challenge of imbalanced training datasets in imitation learning by leveraging Vector Quantized Variational Autoencoders (VQ-VAEs). The use of VQ-VAE provides a structured and interpretable data representation, which helps reveal meaningful patterns in the data. These patterns are used to group the data into clusters, with each sample being assigned a cluster ID. The cluster IDs are then used to re-balance the dataset, ensuring that rare yet valuable samples receive higher priority during training. By ensuring a more diverse and informative training set, CAPS improves the generalization of the trained planner across a wide range of driving scenarios. We evaluate our method through closed-loop simulations in the CARLA environment. The results on Bench2Drive scenarios demonstrate that our framework outperforms state-of-the-art methods, leading to notable improvements in model performance.
Authors: Yingxue Xu, Fengtao Zhou, Chenyu Zhao, Yihui Wang, Can Yang, Hao Chen
Abstract: The integration of multimodal data including pathology images and gene profiles is widely applied in precise survival prediction. Despite recent advances in multimodal survival models, collecting complete modalities for multimodal fusion still poses a significant challenge, hindering their application in clinical settings. Current approaches tackling incomplete modalities often fall short, as they typically compensate for only a limited part of the knowledge of missing modalities. To address this issue, we propose a Distilled Prompt Learning framework (DisPro) to utilize the strong robustness of Large Language Models (LLMs) to missing modalities, which employs two-stage prompting for compensation of comprehensive information for missing modalities. In the first stage, Unimodal Prompting (UniPro) distills the knowledge distribution of each modality, preparing for supplementing modality-specific knowledge of the missing modality in the subsequent stage. In the second stage, Multimodal Prompting (MultiPro) leverages available modalities as prompts for LLMs to infer the missing modality, which provides modality-common information. Simultaneously, the unimodal knowledge acquired in the first stage is injected into multimodal inference to compensate for the modality-specific knowledge of the missing modality. Extensive experiments covering various missing scenarios demonstrated the superiority of the proposed method. The code is available at https://github.com/Innse/DisPro.
Authors: Youngbin Choi, Seunghyuk Cho, Minjong Lee, MoonJeong Park, Yesong Ko, Jungseul Ok, Dongwoo Kim
Abstract: Personalizing large language models (LLMs) is important for aligning outputs with diverse user preferences, yet existing methods struggle with flexibility and generalization. We propose CoPL (Collaborative Preference Learning), a graph-based collaborative filtering framework that models user-response relationships to enhance preference estimation, particularly in sparse annotation settings. By integrating a mixture of LoRA experts, CoPL efficiently fine-tunes LLMs while dynamically balancing shared and user-specific preferences. Additionally, an optimization-free adaptation strategy enables generalization to unseen users without fine-tuning. Experiments on UltraFeedback-P demonstrate that CoPL outperforms existing personalized reward models, effectively capturing both common and controversial preferences, making it a scalable solution for personalized LLM alignment.
Authors: Thang Do, Arnulf Jentzen, Adrian Riekert
Abstract: Despite the omnipresent use of stochastic gradient descent (SGD) optimization methods in the training of deep neural networks (DNNs), it remains, in basically all practically relevant scenarios, a fundamental open problem to provide a rigorous theoretical explanation for the success (and the limitations) of SGD optimization methods in deep learning. In particular, it remains an open question to prove or disprove convergence of the true risk of SGD optimization methods to the optimal true risk value in the training of DNNs. In one of the main results of this work we reveal for a general class of activations, loss functions, random initializations, and SGD optimization methods (including, for example, standard SGD, momentum SGD, Nesterov accelerated SGD, Adagrad, RMSprop, Adadelta, Adam, Adamax, Nadam, Nadamax, and AMSGrad) that in the training of any arbitrary fully-connected feedforward DNN it does not hold that the true risk of the considered optimizer converges in probability to the optimal true risk value. Nonetheless, the true risk of the considered SGD optimization method may very well converge to a strictly suboptimal true risk value.
Authors: Lukas Silvester Barth, Hannaneh Fahimi, Parvaneh Joharinad, J\"urgen Jost, Janis Keck
Abstract: Many machine learning algorithms try to visualize high dimensional metric data in 2D in such a way that the essential geometric and topological features of the data are highlighted. In this paper, we introduce a framework for aggregating dissimilarity functions that arise from locally adjusting a metric through density-aware normalization, as employed in the IsUMap method. We formalize these approaches as m-schemes, a class of methods closely related to t-norms and t-conorms in probabilistic metrics, as well as to composition laws in information theory. These m-schemes provide a flexible and theoretically grounded approach to refining distance-based embeddings.
Authors: Arvin Hosseinzadeh, Ladan Khoshnevisan, Mohammad Pirani, Shojaeddin Chenouri, Amir Khajepour
Abstract: In continual time series analysis using neural networks, catastrophic forgetting (CF) of previously learned models when training on new data domains has always been a significant challenge. This problem is especially challenging in vehicle estimation and control, where new information is sequentially introduced to the model. Unfortunately, existing work on continual learning has not sufficiently addressed the adverse effects of catastrophic forgetting in time series analysis, particularly in multivariate output environments. In this paper, we present EM-ReSeleCT (Efficient Multivariate Representative Selection for Continual Learning in Time Series Tasks), an enhanced approach designed to handle continual learning in multivariate environments. Our approach strategically selects representative subsets from old and historical data and incorporates memory-based continual learning techniques with an improved optimization algorithm to adapt the pre-trained model on new information while preserving previously acquired information. Additionally, we develop a sequence-to-sequence transformer model (autoregressive model) specifically designed for vehicle state estimation. Moreover, we propose an uncertainty quantification framework using conformal prediction to assess the sensitivity of the memory size and to showcase the robustness of the proposed method. Experimental results from tests on an electric Equinox vehicle highlight the superiority of our method in continually learning new information while retaining prior knowledge, outperforming state-of-the-art continual learning methods. Furthermore, EM-ReSeleCT significantly reduces training time, a critical advantage in continual learning applications.
Authors: Justin N. Kreikemeyer, Mi{\l}osz Jankowski, Pia Wilsdorf, Adelinde M. Uhrmacher
Abstract: Formal languages are an integral part of modeling and simulation. They allow the distillation of knowledge into concise simulation models amenable to automatic execution, interpretation, and analysis. However, the arguably most humanly accessible means of expressing models is through natural language, which is not easily interpretable by computers. Here, we evaluate how a Large Language Model (LLM) might be used for formalizing natural language into simulation models. Existing studies only explored using very large LLMs, like the commercial GPT models, without fine-tuning model weights. To close this gap, we show how an open-weights, 7B-parameter Mistral model can be fine-tuned to translate natural language descriptions to reaction network models in a domain-specific language, offering a self-hostable, compute-, and memory efficient alternative. To this end, we develop a synthetic data generator to serve as the basis for fine-tuning and evaluation. Our quantitative evaluation shows that our fine-tuned Mistral model can recover the ground truth simulation model in up to 84.5% of cases. In addition, our small-scale user study demonstrates the model's practical potential for one-time generation as well as interactive modeling in various domains. While promising, in its current form, the fine-tuned small LLM cannot catch up with large LLMs. We conclude that higher-quality training data are required, and expect future small and open-source LLMs to offer new opportunities.
Authors: Mufan Qiu, Xinyu Hu, Fengwei Zhan, Sukwon Yun, Jie Peng, Ruichen Zhang, Bhavya Kailkhura, Jiekun Yang, Tianlong Chen
Abstract: Foundation models for single-cell RNA sequencing (scRNA-seq) have shown promising capabilities in capturing gene expression patterns. However, current approaches face critical limitations: they ignore biological prior knowledge encoded in gene regulatory relationships and fail to leverage multi-omics signals that could provide complementary regulatory insights. In this paper, we propose GRNFormer, a new framework that systematically integrates multi-scale Gene Regulatory Networks (GRNs) inferred from multi-omics data into RNA foundation model training. Our framework introduces two key innovations. First, we introduce a pipeline for constructing hierarchical GRNs that capture regulatory relationships at both cell-type-specific and cell-specific resolutions. Second, we design a structure-aware integration framework that addresses the information asymmetry in GRNs through two technical advances: (1) A graph topological adapter using multi-head cross-attention to weight regulatory relationships dynamically, and (2) a novel edge perturbation strategy that perturb GRNs with biologically-informed co-expression links to augment graph neural network training. Comprehensive experiments have been conducted on three representative downstream tasks across multiple model architectures to demonstrate the effectiveness of GRNFormer. It achieves consistent improvements over state-of-the-art (SoTA) baselines: $3.6\%$ increase in drug response prediction correlation, $9.6\%$ improvement in single-cell drug classification AUC, and $1.1\%$ average gain in gene perturbation prediction accuracy.
Authors: Nandi Schoots, Mattia Jacopo Villani, Niels uit de Bos
Abstract: Kolmogorov-Arnold Networks are a new family of neural network architectures which holds promise for overcoming the curse of dimensionality and has interpretability benefits (arXiv:2404.19756). In this paper, we explore the connection between Kolmogorov Arnold Networks (KANs) with piecewise linear (univariate real) functions and ReLU networks. We provide completely explicit constructions to convert a piecewise linear KAN into a ReLU network and vice versa.
Authors: Vatsal Srivastava
Abstract: The aim of this paper is to investigate an attempt to build a binary classification algorithm using principles of geometry such as vectors, planes, and vector algebra. The basic idea behind the proposed algorithm is that a hyperplane can be used to completely separate a given set of data points mapped to n dimensional space, if the given data points are linearly separable in the n dimensions. Since points are the foundational elements of any geometrical construct, by manipulating the position of points used for the construction of a given hyperplane, the position of the hyperplane itself can be manipulated. The paper includes testing data against other classifiers on a variety of standard machine learning datasets. With a focus on support vector machines, since they and our proposed classifier use the same geometrical construct of hyperplane, and the versatility of SVMs make them a good bench mark for comparison. Since the algorithm focuses on moving the points through the hyperspace to which the dataset has been mapped, it has been dubbed as moving points algorithm.
Authors: Minoo Hosseinzadeh, Hana Khamfroush
Abstract: With a recent trend of using Large Language Models (LLMs) for different applications within smart cities, there is a need for pushing these models toward the edge of network while still preserving their performance. Edge Computing (EC) as a physically closer computing resource to the end users can help to reduce the communication delay for serving end users' tasks for LLM-dependent services. However, EC servers have limited capacity in terms of communication, computation, and storage capacity. This paper introduces DILEMMA, a novel framework addressing the challenges of deploying LLMs in EC systems by jointly optimizing layer placement and layer quantization in EC systems. DILEMMA formulates an Integer Linear Programming problem to minimize total inference delay while ensuring acceptable LLM performance levels, leveraging layer-wise quantization and knowledge distillation for LLM performance control. Experimental evaluations on OPT-350 model using the SQuAD dataset demonstrate that DILEMMA achieves a quantization ratio of up to 12.75% while preserving model loss, highlighting its effectiveness in resource-constrained environments.
Authors: Jintao Zhang, Guoliang Li, Jinyang Su
Abstract: Retrieval-augmented generation (RAG) has demonstrated significant proficiency in conducting question-answering (QA) tasks within a specified corpus. Nonetheless, numerous failure instances of RAG in QA still exist. These failures are not solely attributable to the limitations of Large Language Models (LLMs); instead, they predominantly arise from the retrieval of inaccurate information for LLMs due to two limitations: (1) Current RAG methods segment the corpus without considering semantics, making it difficult to find relevant context due to impaired correlation between questions and the segments. (2) There is a trade-off between missing essential context with fewer context retrieved and getting irrelevant context with more context retrieved. In this paper, we introduce a RAG framework (SAGE), to overcome these limitations. First, to address the segmentation issue without considering semantics, we propose to train a semantic segmentation model. This model is trained to segment the corpus into semantically complete chunks. Second, to ensure that only the most relevant chunks are retrieved while the irrelevant ones are ignored, we design a chunk selection algorithm to dynamically select chunks based on the decreasing speed of the relevance score, leading to a more relevant selection. Third, to further ensure the precision of the retrieved chunks, we propose letting LLMs assess whether retrieved chunks are excessive or lacking and then adjust the amount of context accordingly. Experiments show that SAGE outperforms baselines by 61.25% in the quality of QA on average. Moreover, by avoiding retrieving noisy context, SAGE lowers the cost of the tokens consumed in LLM inference and achieves a 49.41% enhancement in cost efficiency on average. Additionally, our work offers valuable insights for boosting RAG.
Authors: Kevin Burrage, Pamela Burrage, Justin N. Kreikemeyer, Adelinde M. Uhrmacher, Hasitha N. Weerasinghe
Abstract: In this paper, we adapt a two species agent-based cancer model that describes the interaction between cancer cells and healthy cells on a uniform grid to include the interaction with a third species -- namely immune cells. We run six different scenarios to explore the competition between cancer and immune cells and the initial concentration of the immune cells on cancer dynamics. We then use coupled equation learning to construct a population-based reaction model for each scenario. We show how they can be unified into a single surrogate population-based reaction model, whose underlying three coupled ordinary differential equations are much easier to analyse than the original agent-based model. As an example, by finding the single steady state of the cancer concentration, we are able to find a linear relationship between this concentration and the initial concentration of the immune cells. This then enables us to estimate suitable values for the competition and initial concentration to reduce the cancer substantially without performing additional complex and expensive simulations from an agent-based stochastic model. The work shows the importance of performing equation learning from agent-based stochastic data for gaining key insights about the behaviour of complex cellular dynamics.
Authors: Ryien Hosseini, Filippo Simini, Venkatram Vishwanath, Rebecca Willett, Henry Hoffmann
Abstract: Deep generative models have recently achieved significant success in modeling graph data, including dynamic graphs, where topology and features evolve over time. However, unlike in vision and natural language domains, evaluating generative models for dynamic graphs is challenging due to the difficulty of visualizing their output, making quantitative metrics essential. In this work, we develop a new quality metric for evaluating generative models of dynamic graphs. Current metrics for dynamic graphs typically involve discretizing the continuous-evolution of graphs into static snapshots and then applying conventional graph similarity measures. This approach has several limitations: (a) it models temporally related events as i.i.d. samples, failing to capture the non-uniform evolution of dynamic graphs; (b) it lacks a unified measure that is sensitive to both features and topology; (c) it fails to provide a scalar metric, requiring multiple metrics without clear superiority; and (d) it requires explicitly instantiating each static snapshot, leading to impractical runtime demands that hinder evaluation at scale. We propose a novel metric based on the \textit{Johnson-Lindenstrauss} lemma, applying random projections directly to dynamic graph data. This results in an expressive, scalar, and application-agnostic measure of dynamic graph similarity that overcomes the limitations of traditional methods. We also provide a comprehensive empirical evaluation of metrics for continuous-time dynamic graphs, demonstrating the effectiveness of our approach compared to existing methods. Our implementation is available at https://github.com/ryienh/jl-metric.
Authors: Nikolaos Nakis, Niels Raunkj{\ae}r Holm, Andreas Lyhne Fiehn, Morten M{\o}rup
Abstract: Low-dimensional embeddings are essential for machine learning tasks involving graphs, such as node classification, link prediction, community detection, network visualization, and network compression. Although recent studies have identified exact low-dimensional embeddings, the limits of the required embedding dimensions remain unclear. We presently prove that lower dimensional embeddings are possible when using Euclidean metric embeddings as opposed to vector-based Logistic PCA (LPCA) embeddings. In particular, we provide an efficient logarithmic search procedure for identifying the exact embedding dimension and demonstrate how metric embeddings enable inference of the exact embedding dimensions of large-scale networks by exploiting that the metric properties can be used to provide linearithmic scaling. Empirically, we show that our approach extracts substantially lower dimensional representations of networks than previously reported for small-sized networks. For the first time, we demonstrate that even large-scale networks can be effectively embedded in very low-dimensional spaces, and provide examples of scalable, exact reconstruction for graphs with up to a million nodes. Our approach highlights that the intrinsic dimensionality of networks is substantially lower than previously reported and provides a computationally efficient assessment of the exact embedding dimension also of large-scale networks. The surprisingly low dimensional representations achieved demonstrate that networks in general can be losslessly represented using very low dimensional feature spaces, which can be used to guide existing network analysis tasks from community detection and node classification to structure revealing exact network visualizations.
Authors: Jos\'e Medina, Amnir Hadachi, Paul Honeine, Abdelaziz Bensrhair
Abstract: Deep neural networks (DNNs) have remarkably succeeded in various image processing tasks. However, their large size and computational complexity present significant challenges for deploying them in resource-constrained environments. This paper presents an innovative approach for integrating Mamba Architecture within a Progressive Knowledge Distillation (PKD) process to address the challenge of reducing model complexity while maintaining accuracy in image classification tasks. The proposed framework distills a large teacher model into progressively smaller student models, designed using Mamba blocks. Each student model is trained using Selective-State-Space Models (S-SSM) within the Mamba blocks, focusing on important input aspects while reducing computational complexity. The work's preliminary experiments use MNIST and CIFAR-10 as datasets to demonstrate the effectiveness of this approach. For MNIST, the teacher model achieves 98% accuracy. A set of seven student models as a group retained 63% of the teacher's FLOPs, approximating the teacher's performance with 98% accuracy. The weak student used only 1% of the teacher's FLOPs and maintained 72% accuracy. Similarly, for CIFAR-10, the students achieved 1% less accuracy compared to the teacher, with the small student retaining 5% of the teacher's FLOPs to achieve 50% accuracy. These results confirm the flexibility and scalability of Mamba Architecture, which can be integrated into PKD, succeeding in the process of finding students as weak learners. The framework provides a solution for deploying complex neural networks in real-time applications with a reduction in computational cost.
Authors: Zhe Gao, Jian Huang, Ting Li, Xueqin Wang
Abstract: Multimodal learning has become a pivotal approach in developing robust learning models with applications spanning multimedia, robotics, large language models, and healthcare. The efficiency of multimodal systems is a critical concern, given the varying costs and resource demands of different modalities. This underscores the necessity for effective modality selection to balance performance gains against resource expenditures. In this study, we propose a novel framework for modality selection that independently learns the representation of each modality. This approach allows for the assessment of each modality's significance within its unique representation space, enabling the development of tailored encoders and facilitating the joint analysis of modalities with distinct characteristics. Our framework aims to enhance the efficiency and effectiveness of multimodal learning by optimizing modality integration and selection.
Authors: Mohammad Rafid Ul Islam, Prasad Tadepalli, Alan Fern
Abstract: Missing values in multivariate time series data can harm machine learning performance and introduce bias. These gaps arise from sensor malfunctions, blackouts, and human error and are typically addressed by data imputation. Previous work has tackled the imputation of missing data in random, complete blackouts and forecasting scenarios. The current paper addresses a more general missing pattern, which we call "partial blackout," where a subset of features is missing for consecutive time steps. We introduce a two-stage imputation process using self-attention and diffusion processes to model feature and temporal correlations. Notably, our model effectively handles missing data during training, enhancing adaptability and ensuring reliable imputation and performance, even with incomplete datasets. Our experiments on benchmark and two real-world time series datasets demonstrate that our model outperforms the state-of-the-art in partial blackout scenarios and shows better scalability.
Authors: Nastaran Mansourian, Arash Mohammadi, M. Omair Ahmad, M. N. S. Swamy
Abstract: Driver emotion recognition plays a crucial role in driver monitoring systems, enhancing human-autonomy interactions and the trustworthiness of Autonomous Driving (AD). Various physiological and behavioural modalities have been explored for this purpose, with Electrocardiogram (ECG) emerging as a standout choice for real-time emotion monitoring, particularly in dynamic and unpredictable driving conditions. Existing methods, however, often rely on multi-channel ECG signals recorded under static conditions, limiting their applicability in real-world dynamic driving scenarios. To address this limitation, the paper introduces ECG-EmotionNet, a novel architecture designed specifically for emotion recognition in dynamic driving environments. ECG-EmotionNet is constructed by adapting a recently introduced ECG Foundation Model (FM) and uniquely employs single-channel ECG signals, ensuring both robust generalizability and computational efficiency. Unlike conventional adaptation methods such as full fine-tuning, linear probing, or low-rank adaptation, we propose an intuitively pleasing alternative, referred to as the nested Mixture of Experts (MoE) adaptation. More precisely, each transformer layer of the underlying FM is treated as a separate expert, with embeddings extracted from these experts fused using trainable weights within a gating mechanism. This approach enhances the representation of both global and local ECG features, leading to a 6% improvement in accuracy and a 7% increase in the F1 score, all while maintaining computational efficiency. The effectiveness of the proposed ECG-EmotionNet architecture is evaluated using a recently introduced and challenging driver emotion monitoring dataset.
Authors: Heming Fu, Hongkai Chen, Shan Lin, Guoliang Xing
Abstract: Alzheimer's Disease (AD) has become an increasingly critical global health concern, which necessitates effective monitoring solutions in smart health applications. However, the development of such solutions is significantly hindered by the scarcity of AD-specific activity datasets. To address this challenge, we propose SHADE-AD, a Large Language Model (LLM) framework for Synthesizing Human Activity Datasets Embedded with AD features. Leveraging both public datasets and our own collected data from 99 AD patients, SHADE-AD synthesizes human activity videos that specifically represent AD-related behaviors. By employing a three-stage training mechanism, it broadens the range of activities beyond those collected from limited deployment settings. We conducted comprehensive evaluations of the generated dataset, demonstrating significant improvements in downstream tasks such as Human Activity Recognition (HAR) detection, with enhancements of up to 79.69%. Detailed motion metrics between real and synthetic data show strong alignment, validating the realism and utility of the synthesized dataset. These results underscore SHADE-AD's potential to advance smart health applications by providing a cost-effective, privacy-preserving solution for AD monitoring.
Authors: Tiansheng Wen, Yifei Wang, Zequn Zeng, Zhong Peng, Yudi Su, Xinyang Liu, Bo Chen, Hongwei Liu, Stefanie Jegelka, Chenyu You
Abstract: Many large-scale systems rely on high-quality deep representations (embeddings) to facilitate tasks like retrieval, search, and generative modeling. Matryoshka Representation Learning (MRL) recently emerged as a solution for adaptive embedding lengths, but it requires full model retraining and suffers from noticeable performance degradations at short lengths. In this paper, we show that sparse coding offers a compelling alternative for achieving adaptive representation with minimal overhead and higher fidelity. We propose Contrastive Sparse Representation (CSR), a method that sparsifies pre-trained embeddings into a high-dimensional but selectively activated feature space. By leveraging lightweight autoencoding and task-aware contrastive objectives, CSR preserves semantic quality while allowing flexible, cost-effective inference at different sparsity levels. Extensive experiments on image, text, and multimodal benchmarks demonstrate that CSR consistently outperforms MRL in terms of both accuracy and retrieval speed-often by large margins-while also cutting training time to a fraction of that required by MRL. Our results establish sparse coding as a powerful paradigm for adaptive representation learning in real-world applications where efficiency and fidelity are both paramount. Code is available at https://github.com/neilwen987/CSR_Adaptive_Rep
Authors: Peijun Hou, Nan Cen
Abstract: Hybrid light fidelity (LiFi) and wireless fidelity (WiFi) indoor networks has been envisioned as a promising technology to alleviate radio frequency spectrum crunch to accommodate the ever-increasing data rate demand in indoor scenarios. The hybrid LiFi/WiFi indoor networks can leverage the advantages of fast data transmission from LiFi and wider coverage of WiFi, thus complementing well with each other and further improving the network performance compared with the standalone networks. However, to leverage the co-existence, several challenges should be addressed, including but not limited to user association, mobility support, and efficient resource allocation. Therefore, the objective of the paper is to design a new user-access point association algorithm to maximize the sum throughput of the hybrid networks. We first mathematically formulate the sum data rate maximization problem by determining the AP selection for each user in indoor networks with consideration of user mobility and practical capacity limitations, which is a nonconvex binary integer programming problem. To solve this problem, we then propose a sequential-proximal policy optimization (S-PPO) based deep reinforcement learning method. Extensive simulations are conducted to evaluate the proposed method by comparing it with exhaustive search (ES), signal strength strategy (SSS), and trust region policy optimization (TRPO) methods. Comprehensive simulation results demonstrate that our solution algorithm can outperform SSS by about 32.25% of the sum throughput and 19.09% of the fairness on average, and outperform TRPO by about 10.34% and 10.23%, respectively.
Authors: Gilad Yehudai, Clayton Sanford, Maya Bechler-Speicher, Orr Fischer, Ran Gilad-Bachrach, Amir Globerson
Abstract: Transformers have revolutionized the field of machine learning. In particular, they can be used to solve complex algorithmic problems, including graph-based tasks. In such algorithmic tasks a key question is what is the minimal size of a transformer that can implement a task. Recent work has begun to explore this problem for graph-based tasks, showing that for sub-linear embedding dimension (i.e., model width) logarithmic depth suffices. However, an open question, which we address here, is what happens if width is allowed to grow linearly. Here we analyze this setting, and provide the surprising result that with linear width, constant depth suffices for solving a host of graph-based problems. This suggests that a moderate increase in width can allow much shallower models, which are advantageous in terms of inference time. For other problems, we show that quadratic width is required. Our results demonstrate the complex and intriguing landscape of transformer implementations of graph-based algorithms. We support our theoretical results with empirical evaluations.
Authors: Alessandro Daniele, Emile van Krieken
Abstract: Deep learning has achieved remarkable success across various domains, largely thanks to the efficiency of backpropagation (BP). However, BP's reliance on differentiability poses challenges in neurosymbolic learning, where discrete computation is combined with neural models. We show that applying BP to Godel logic, which represents conjunction and disjunction as min and max, is equivalent to a local search algorithm for SAT solving, enabling the optimisation of discrete Boolean formulas without sacrificing differentiability. However, deterministic local search algorithms get stuck in local optima. Therefore, we propose the Godel Trick, which adds noise to the model's logits to escape local optima. We evaluate the Godel Trick on SATLIB, and demonstrate its ability to solve a broad range of SAT problems. Additionally, we apply it to neurosymbolic models and achieve state-of-the-art performance on Visual Sudoku, all while avoiding expensive probabilistic reasoning. These results highlight the Godel Trick's potential as a robust, scalable approach for integrating symbolic reasoning with neural architectures.
Authors: Yi-Lin Sung, Prateek Yadav, Jialu Li, Jaehong Yoon, Mohit Bansal
Abstract: Layer-wise quantization is a key technique for efficiently compressing large models without expensive retraining. Previous methods typically quantize the weights of each layer by "uniformly" optimizing the layer reconstruction loss across all output tokens. However, in this paper, we demonstrate that better-quantized models can be obtained by prioritizing learning from important tokens (e.g. which have large attention scores). Building on this finding, we propose RSQ (Rotate, Scale, then Quantize), which (1) applies rotations (orthogonal transformation) to the model to mitigate outliers (those with exceptionally large magnitude), (2) scales the token feature based on its importance, and (3) quantizes the model using the GPTQ framework with the second-order statistics computed by scaled tokens. To compute token importance, we explore both heuristic and dynamic strategies. Based on a thorough analysis of all approaches, we adopt attention concentration, which uses attention scores of each token as its importance, as the best approach. We demonstrate that RSQ consistently outperforms baseline methods across multiple downstream tasks and three model families: LLaMA3, Mistral, and Qwen2.5. Additionally, models quantized with RSQ achieve superior performance on long-context tasks, further highlighting its effectiveness. Lastly, RSQ demonstrates generalizability across various setups, including different model sizes, calibration datasets, bit precisions, and quantization methods.
Authors: Xingyu Zhu, Abhishek Panigrahi, Sanjeev Arora
Abstract: We formalize a new concept for LLMs, context-enhanced learning. It involves standard gradient-based learning on text except that the context is enhanced with additional data on which no auto-regressive gradients are computed. This setting is a gradient-based analog of usual in-context learning (ICL) and appears in some recent works. Using a multi-step reasoning task, we prove in a simplified setting that context-enhanced learning can be exponentially more sample-efficient than standard learning when the model is capable of ICL. At a mechanistic level, we find that the benefit of context-enhancement arises from a more accurate gradient learning signal. We also experimentally demonstrate that it appears hard to detect or recover learning materials that were used in the context during training. This may have implications for data security as well as copyright.
Authors: Sai Sumedh R. Hindupur, Ekdeep Singh Lubana, Thomas Fel, Demba Ba
Abstract: Sparse Autoencoders (SAEs) are widely used to interpret neural networks by identifying meaningful concepts from their representations. However, do SAEs truly uncover all concepts a model relies on, or are they inherently biased toward certain kinds of concepts? We introduce a unified framework that recasts SAEs as solutions to a bilevel optimization problem, revealing a fundamental challenge: each SAE imposes structural assumptions about how concepts are encoded in model representations, which in turn shapes what it can and cannot detect. This means different SAEs are not interchangeable -- switching architectures can expose entirely new concepts or obscure existing ones. To systematically probe this effect, we evaluate SAEs across a spectrum of settings: from controlled toy models that isolate key variables, to semi-synthetic experiments on real model activations and finally to large-scale, naturalistic datasets. Across this progression, we examine two fundamental properties that real-world concepts often exhibit: heterogeneity in intrinsic dimensionality (some concepts are inherently low-dimensional, others are not) and nonlinear separability. We show that SAEs fail to recover concepts when these properties are ignored, and we design a new SAE that explicitly incorporates both, enabling the discovery of previously hidden concepts and reinforcing our theoretical insights. Our findings challenge the idea of a universal SAE and underscores the need for architecture-specific choices in model interpretability. Overall, we argue an SAE does not just reveal concepts -- it determines what can be seen at all.
Authors: David Klindt, Charles O'Neill, Patrik Reizinger, Harald Maurer, Nina Miolane
Abstract: Understanding how information is represented in neural networks is a fundamental challenge in both neuroscience and artificial intelligence. Despite their nonlinear architectures, recent evidence suggests that neural networks encode features in superposition, meaning that input concepts are linearly overlaid within the network's representations. We present a perspective that explains this phenomenon and provides a foundation for extracting interpretable representations from neural activations. Our theoretical framework consists of three steps: (1) Identifiability theory shows that neural networks trained for classification recover latent features up to a linear transformation. (2) Sparse coding methods can extract disentangled features from these representations by leveraging principles from compressed sensing. (3) Quantitative interpretability metrics provide a means to assess the success of these methods, ensuring that extracted features align with human-interpretable concepts. By bridging insights from theoretical neuroscience, representation learning, and interpretability research, we propose an emerging perspective on understanding neural representations in both artificial and biological systems. Our arguments have implications for neural coding theories, AI transparency, and the broader goal of making deep learning models more interpretable.
Authors: Anders Sildnes, Nikita Shvetsov, Masoud Tafavvoghi, Vi Ngoc-Nha Tran, Kajsa M{\o}llersen, Lill-Tove Rasmussen Busund, Thomas K. Kilv{\ae}r, Lars Ailo Bongo
Abstract: Even foundational models that are trained on datasets with billions of data samples may develop shortcuts that lead to overfitting and bias. Shortcuts are non-relevant patterns in data, such as the background color or color intensity. So, to ensure the robustness of deep learning applications, there is a need for methods to detect and remove such shortcuts. Today's model debugging methods are time consuming since they often require customization to fit for a given model architecture in a specific domain. We propose a generalized, model-agnostic framework to debug deep learning models. We focus on the domain of histopathology, which has very large images that require large models - and therefore large computation resources. It can be run on a workstation with a commodity GPU. We demonstrate that our framework can replicate non-image shortcuts that have been found in previous work for self-supervised learning models, and we also identify possible shortcuts in a foundation model. Our easy to use tests contribute to the development of more reliable, accurate, and generalizable models for WSI analysis. Our framework is available as an open-source tool available on github.
Authors: Adri\`a L\'opez Escoriza, Nicklas Hansen, Stone Tao, Tongzhou Mu, Hao Su
Abstract: Long-horizon tasks in robotic manipulation present significant challenges in reinforcement learning (RL) due to the difficulty of designing dense reward functions and effectively exploring the expansive state-action space. However, despite a lack of dense rewards, these tasks often have a multi-stage structure, which can be leveraged to decompose the overall objective into manageable subgoals. In this work, we propose DEMO3, a framework that exploits this structure for efficient learning from visual inputs. Specifically, our approach incorporates multi-stage dense reward learning, a bi-phasic training scheme, and world model learning into a carefully designed demonstration-augmented RL framework that strongly mitigates the challenge of exploration in long-horizon tasks. Our evaluations demonstrate that our method improves data-efficiency by an average of 40% and by 70% on particularly difficult tasks compared to state-of-the-art approaches. We validate this across 16 sparse-reward tasks spanning four domains, including challenging humanoid visual control tasks using as few as five demonstrations.
Authors: Maria Drencheva, Ivo Petrov, Maximilian Baader, Dimitar I. Dimitrov, Martin Vechev
Abstract: Federated learning claims to enable collaborative model training among multiple clients with data privacy by transmitting gradient updates instead of the actual client data. However, recent studies have shown the client privacy is still at risk due to the, so called, gradient inversion attacks which can precisely reconstruct clients' text and image data from the shared gradient updates. While these attacks demonstrate severe privacy risks for certain domains and architectures, the vulnerability of other commonly-used data types, such as graph-structured data, remain under-explored. To bridge this gap, we present GRAIN, the first exact gradient inversion attack on graph data in the honest-but-curious setting that recovers both the structure of the graph and the associated node features. Concretely, we focus on Graph Convolutional Networks (GCN) and Graph Attention Networks (GAT) -- two of the most widely used frameworks for learning on graphs. Our method first utilizes the low-rank structure of GNN gradients to efficiently reconstruct and filter the client subgraphs which are then joined to complete the input graph. We evaluate our approach on molecular, citation, and social network datasets using our novel metric. We show that GRAIN reconstructs up to 80% of all graphs exactly, significantly outperforming the baseline, which achieves up to 20% correctly positioned nodes.
Authors: Dayal Singh Kalra, John Kirchenbauer, Maissam Barkeshli, Tom Goldstein
Abstract: Adam is the go-to optimizer for training modern machine learning models, but it requires additional memory to maintain the moving averages of the gradients and their squares. While various low-memory optimizers have been proposed that sometimes match the performance of Adam, their lack of reliability has left Adam as the default choice. In this work, we apply a simple layer-wise Signal-to-Noise Ratio (SNR) analysis to quantify when second-moment tensors can be effectively replaced by their means across different dimensions. Our SNR analysis reveals how architecture, training hyperparameters, and dataset properties impact compressibility along Adam's trajectory, naturally leading to $\textit{SlimAdam}$, a memory-efficient Adam variant. $\textit{SlimAdam}$ compresses the second moments along dimensions with high SNR when feasible, and leaves when compression would be detrimental. Through experiments across a diverse set of architectures and training scenarios, we show that $\textit{SlimAdam}$ matches Adam's performance and stability while saving up to $98\%$ of total second moments. Code for $\textit{SlimAdam}$ is available at https://github.com/dayal-kalra/low-memory-adam.
Authors: Gopi Krishna Jha, Sameh Gobriel, Liubov Talamanova, Alexander Kozlov, Nilesh Jain
Abstract: Key-value (KV) caching has emerged as a crucial optimization technique for accelerating inference in large language models (LLMs). By allowing the attention operation to scale linearly rather than quadratically with the total sequence length, KV caching significantly enhances generation throughput. However, due to large context lengths in the modern LLMs, the memory footprint of the KV is a huge bottleneck for model deployment directly impacting the model's batch size, hindering its ability to deliver high-throughput. Existing research addresses this challenge using several techniques, such as discarding low-attention tokens, quantization, and matrix approximation which typically lead to a negative impact on the model accuracy. In this paper, We propose KVCrush technology which can be combined with many KV compression technologies to improve the model accuracy at a much smaller memory. KVCrush provides an alternate representation scheme for key-value states, along with a low-overhead token pruning algorithm that accounts for the token distribution in the KV cache, which in turn allows for a a smaller footprint while maintaining the accuracy of the model. Based on our results, KVCrush reduces LongBench KV Cache size by 4x with less than 1% accuracy drop and achieves state-of-the-art average accuracy with minimal overhead, incurring less than 0.5% total inference latency. KVCrush not only outperforms the accuracy of state-of-the-art importance-based token retention schemes but is also compatible with typical practical LLM deployments using KV cache paging schemes such as vLLM and mixed precision quantization.
Authors: Yan Su, Qiulin Wu, Weizhen Li, Chengchang Pan, Honggang Qi
Abstract: The training of deep learning models relies on a large amount of labeled data. However, the high cost of medical labeling seriously hinders the development of deep learning in the medical field. Our study proposes a general disease diagnosis approach based on Zero-Shot Learning. The Siamese neural network is used to find similar diseases for the target diseases, and the U-Net segmentation model is used to accurately segment the key lesions of the disease. Finally, based on the ResNet-Agglomerative clustering algorithm, a clustering model is trained on a large number of sample data of similar diseases to obtain a approximate diagnosis of the target disease. Zero-Shot Learning of the target disease is then successfully achieved. To evaluate the validity of the model, we validated our method on a dataset of ophthalmic diseases in CFP modality. The external dataset was used to test its performance, and the accuracy=0.8395, precision=0.8094, recall=0.8463, F1 Score=0.8274, AUC=0.9226, which exceeded the indexes of most Few-Shot Learning and One-Shot Learning models. It proves that our method has great potential and reference value in the medical field, where annotation data is usually scarce and expensive to obtain.
Authors: Sarmistha Das, Basha Mujavarsheik, R E Zera Lyngkhoi, Sriparna Saha, Alka Maurya
Abstract: In today's competitive marketing landscape, effective complaint management is crucial for customer service and business success. Video complaints, integrating text and image content, offer invaluable insights by addressing customer grievances and delineating product benefits and drawbacks. However, comprehending nuanced complaint aspects within vast daily multimodal financial data remains a formidable challenge. Addressing this gap, we have curated a proprietary multimodal video complaint dataset comprising 433 publicly accessible instances. Each instance is meticulously annotated at the utterance level, encompassing five distinct categories of financial aspects and their associated complaint labels. To support this endeavour, we introduce Solution 3.0, a model designed for multimodal aspect-based complaint identification task. Solution 3.0 is tailored to perform three key tasks: 1) handling multimodal features ( audio and video), 2) facilitating multilabel aspect classification, and 3) conducting multitasking for aspect classifications and complaint identification parallelly. Solution 3.0 utilizes a CLIP-based dual frozen encoder with an integrated image segment encoder for global feature fusion, enhanced by contextual attention (ISEC) to improve accuracy and efficiency. Our proposed framework surpasses current multimodal baselines, exhibiting superior performance across nearly all metrics by opening new ways to strengthen appropriate customer care initiatives and effectively assisting individuals in resolving their problems.
Authors: Qiusi Zhan, Richard Fang, Henil Shalin Panchal, Daniel Kang
Abstract: Large Language Model (LLM) agents exhibit remarkable performance across diverse applications by using external tools to interact with environments. However, integrating external tools introduces security risks, such as indirect prompt injection (IPI) attacks. Despite defenses designed for IPI attacks, their robustness remains questionable due to insufficient testing against adaptive attacks. In this paper, we evaluate eight different defenses and bypass all of them using adaptive attacks, consistently achieving an attack success rate of over 50%. This reveals critical vulnerabilities in current defenses. Our research underscores the need for adaptive attack evaluation when designing defenses to ensure robustness and reliability. The code is available at https://github.com/uiuc-kang-lab/AdaptiveAttackAgent.
Authors: Jing Xu, Franziska Boenisch, Adam Dziedzic
Abstract: Graph Neural Networks (GNNs) achieve high performance in various real-world applications, such as drug discovery, traffic states prediction, and recommendation systems. The fact that building powerful GNNs requires a large amount of training data, powerful computing resources, and human expertise turns the models into lucrative targets for model stealing attacks. Prior work has revealed that the threat vector of stealing attacks against GNNs is large and diverse, as an attacker can leverage various heterogeneous signals ranging from node labels to high-dimensional node embeddings to create a local copy of the target GNN at a fraction of the original training costs. This diversity in the threat vector renders the design of effective and general defenses challenging and existing defenses usually focus on one particular stealing setup. Additionally, they solely provide means to identify stolen model copies rather than preventing the attack. To close this gap, we propose the first and general Active Defense Against GNN Extraction (ADAGE). By analyzing the queries to the GNN, tracking their diversity in terms of proximity to different communities identified in the underlying graph, and increasing the defense strength with the growing fraction of communities that have been queried, ADAGE can prevent stealing in all common attack setups. Our extensive experimental evaluation using six benchmark datasets, four GNN models, and three types of adaptive attackers shows that ADAGE penalizes attackers to the degree of rendering stealing impossible, whilst not harming predictive performance for legitimate users. ADAGE, thereby, contributes towards securely sharing valuable GNNs in the future.
Authors: Alireza Gharahighehi, Achilleas Ghinis, Michela Venturini, Frederik Cornillie, Celine Vens
Abstract: Massive Open Online Courses (MOOCs) are emerging as a popular alternative to traditional education, offering learners the flexibility to access a wide range of courses from various disciplines, anytime and anywhere. Despite this accessibility, a significant number of enrollments in MOOCs result in dropouts. To enhance learner engagement, it is crucial to recommend courses that align with their preferences and needs. Course Recommender Systems (RSs) can play an important role in this by modeling learners' preferences based on their previous interactions within the MOOC platform. Time-to-dropout and time-to-completion in MOOCs, like other time-to-event prediction tasks, can be effectively modeled using survival analysis (SA) methods. In this study, we apply SA methods to improve collaborative filtering recommendation performance by considering time-to-event in the context of MOOCs. Our proposed approach demonstrates superior performance compared to collaborative filtering methods trained based on learners' interactions with MOOCs, as evidenced by two performance measures on three publicly available datasets. The findings underscore the potential of integrating SA methods with RSs to enhance personalization in MOOCs.
Authors: Alexander Immer, Jan-Matthis Lueckmann, Alex Bo-Yuan Chen, Peter H. Li, Mariela D. Petkova, Nirmala A. Iyer, Aparna Dev, Gudrun Ihrke, Woohyun Park, Alyson Petruncio, Aubrey Weigel, Wyatt Korff, Florian Engert, Jeff W. Lichtman, Misha B. Ahrens, Viren Jain, Micha{\l} Januszewski
Abstract: Large-scale neuronal activity recordings with fluorescent calcium indicators are increasingly common, yielding high-resolution 2D or 3D videos. Traditional analysis pipelines reduce this data to 1D traces by segmenting regions of interest, leading to inevitable information loss. Inspired by the success of deep learning on minimally processed data in other domains, we investigate the potential of forecasting neuronal activity directly from volumetric videos. To capture long-range dependencies in high-resolution volumetric whole-brain recordings, we design a model with large receptive fields, which allow it to integrate information from distant regions within the brain. We explore the effects of pre-training and perform extensive model selection, analyzing spatio-temporal trade-offs for generating accurate forecasts. Our model outperforms trace-based forecasting approaches on ZAPBench, a recently proposed benchmark on whole-brain activity prediction in zebrafish, demonstrating the advantages of preserving the spatial structure of neuronal activity.
Authors: Chi-Sheng Chen, Samuel Yen-Chi Chen, Huan-Hsin Tseng
Abstract: Electroencephalography (EEG) is widely used in neuroscience and clinical research for analyzing brain activity. While deep learning models such as EEGNet have shown success in decoding EEG signals, they often struggle with data complexity, inter-subject variability, and noise robustness. Recent advancements in quantum machine learning (QML) offer new opportunities to enhance EEG analysis by leveraging quantum computing's unique properties. In this study, we extend the previously proposed Quantum-EEGNet (QEEGNet), a hybrid neural network incorporating quantum layers into EEGNet, to investigate its generalization ability across multiple EEG datasets. Our evaluation spans a diverse set of cognitive and motor task datasets, assessing QEEGNet's performance in different learning scenarios. Experimental results reveal that while QEEGNet demonstrates competitive performance and maintains robustness in certain datasets, its improvements over traditional deep learning methods remain inconsistent. These findings suggest that hybrid quantum-classical architectures require further optimization to fully leverage quantum advantages in EEG processing. Despite these limitations, our study provides new insights into the applicability of QML in EEG research and highlights challenges that must be addressed for future advancements.
Authors: Zhenxing Cui, Lu Chen, Yunhai Wang, Daniel Haehn, Yong Wang, Hanspeter Pfister
Abstract: This paper presents a systematic study of the generalization of convolutional neural networks (CNNs) and humans on relational reasoning tasks with bar charts. We first revisit previous experiments on graphical perception and update the benchmark performance of CNNs. We then test the generalization performance of CNNs on a classic relational reasoning task: estimating bar length ratios in a bar chart, by progressively perturbing the standard visualizations. We further conduct a user study to compare the performance of CNNs and humans. Our results show that CNNs outperform humans only when the training and test data have the same visual encodings. Otherwise, they may perform worse. We also find that CNNs are sensitive to perturbations in various visual encodings, regardless of their relevance to the target bars. Yet, humans are mainly influenced by bar lengths. Our study suggests that robust relational reasoning with visualizations is challenging for CNNs. Improving CNNs' generalization performance may require training them to better recognize task-related visual properties.
Authors: Xinyu Yuan, Zichen Wang, Marcus Collins, Huzefa Rangwala
Abstract: Recent years have witnessed a surge in the development of protein structural tokenization methods, which chunk protein 3D structures into discrete or continuous representations. Structure tokenization enables the direct application of powerful techniques like language modeling for protein structures, and large multimodal models to integrate structures with protein sequences and functional texts. Despite the progress, the capabilities and limitations of these methods remain poorly understood due to the lack of a unified evaluation framework. We first introduce StructTokenBench, a framework that comprehensively evaluates the quality and efficiency of structure tokenizers, focusing on fine-grained local substructures rather than global structures, as typical in existing benchmarks. Our evaluations reveal that no single model dominates all benchmarking perspectives. Observations of codebook under-utilization led us to develop AminoAseed, a simple yet effective strategy that enhances codebook gradient updates and optimally balances codebook size and dimension for improved tokenizer utilization and quality. Compared to the leading model ESM3, our method achieves an average of 6.31% performance improvement across 24 supervised tasks, with sensitivity and utilization rates increased by 12.83% and 124.03%, respectively.
Authors: Nikita Soni, Pranav Chitale, Khushboo Singh, Niranjan Balasubramanian, H. Andrew Schwartz
Abstract: Like most of NLP, models for human-centered NLP tasks -- tasks attempting to assess author-level information -- predominantly use representations derived from hidden states of Transformer-based LLMs. However, what component of the LM is used for the representation varies widely. Moreover, there is a need for Human Language Models (HuLMs) that implicitly model the author and provide a user-level hidden state. Here, we systematically evaluate different ways of representing documents and users using different LM and HuLM architectures to predict task outcomes as both dynamically changing states and averaged trait-like user-level attributes of valence, arousal, empathy, and distress. We find that representing documents as an average of the token hidden states performs the best generally. Further, while a user-level hidden state itself is rarely the best representation, we find its inclusion in the model strengthens token or document embeddings used to derive document- and user-level representations resulting in best performances.
Authors: Farouk Mokhtar, Joosep Pata, Michael Kagan, Dolores Garcia, Eric Wulff, Mengke Zhang, Javier Duarte
Abstract: We demonstrate transfer learning capabilities in a machine-learned algorithm trained for particle-flow reconstruction in high energy particle colliders. This paper presents a cross-detector fine-tuning study, where we initially pre-train the model on a large full simulation dataset from one detector design, and subsequently fine-tune the model on a sample with a different collider and detector design. Specifically, we use the Compact Linear Collider detector (CLICdet) model for the initial training set, and demonstrate successful knowledge transfer to the CLIC-like detector (CLD) proposed for the Future Circular Collider in electron-positron mode (FCC-ee). We show that with an order of magnitude less samples from the second dataset, we can achieve the same performance as a costly training from scratch, across particle-level and event-level performance metrics; including jet resolution and missing transverse momentum resolution. Furthermore, we find that the fine-tuned model achieves comparable performance to the traditional rule-based particle-flow approach on event-level metrics after training on 100,000 CLD events, whereas a model trained from scratch requires at least 1 million CLD events to achieve similar reconstruction performance. To our knowledge, this represents the first full-simulation cross-detector transfer learning study for particle-flow. These findings offer valuable insights towards building large physics models that can be fine-tuned across different detector designs and geometries, helping accelerate the development cycle for new detectors, and opening the door to rapid detector design and optimization using machine learning.
Authors: Abdessamad El-Kabid, El-Mahdi El-Mhamdi
Abstract: Availability attacks are one of the strongest forms of training-phase attacks in machine learning, making the model unusable. While prior work in distributed ML has demonstrated such effect via gradient attacks and, more recently, data poisoning, we ask: can similar damage be inflicted solely by flipping training labels, without altering features? In this work, we introduce a novel formalization of label flipping attacks and derive an attacker-optimized loss function that better illustrates label flipping capabilities. To compare the damaging effect of label flipping with that of gradient attacks, we use a setting that allows us to compare their \emph{writing power} on the ML model. Our contribution is threefold, (1) we provide the first evidence for an availability attack through label flipping alone, (2) we shed light on an interesting interplay between what the attacker gains from more \emph{write access} versus what they gain from more \emph{flipping budget} and (3) we compare the power of targeted label flipping attack to that of an untargeted label flipping attack.
Authors: Tom Pan, Evan Dramko, Mitchell D. Miller, George N. Phillips Jr., Anastasios Kyrillidis
Abstract: Determining protein structures at an atomic level remains a significant challenge in structural biology. We introduce $\texttt{RecCrysFormer}$, a hybrid model that exploits the strengths of transformers with the aim of integrating experimental and ML approaches to protein structure determination from crystallographic data. $\texttt{RecCrysFormer}$ leverages Patterson maps and incorporates known standardized partial structures of amino acid residues to directly predict electron density maps, which are essential for constructing detailed atomic models through crystallographic refinement processes. $\texttt{RecCrysFormer}$ benefits from a ``recycling'' training regimen that iteratively incorporates results from crystallographic refinements and previous training runs as additional inputs in the form of template maps. Using a preliminary dataset of synthetic peptide fragments based on Protein Data Bank, $\texttt{RecCrysFormer}$ achieves good accuracy in structural predictions and shows robustness against variations in crystal parameters, such as unit cell dimensions and angles.
Authors: Engin Zeydan, Cristian J. Vaca-Rubio, Luis Blanco, Roberto Pereira, Marius Caus, Kapal Dev
Abstract: Non-Terrestrial Networks (NTNs) are becoming a critical component of modern communication infrastructures, especially with the advent of Low Earth Orbit (LEO) satellite systems. Traditional centralized learning approaches face major challenges in such networks due to high latency, intermittent connectivity and limited bandwidth. Federated Learning (FL) is a promising alternative as it enables decentralized training while maintaining data privacy. However, existing FL models, such as Federated Learning with Multi-Layer Perceptrons (Fed-MLP), can struggle with high computational complexity and poor adaptability to dynamic NTN environments. This paper provides a detailed analysis for Federated Learning with Kolmogorov-Arnold Networks (Fed-KAN), its implementation and performance improvements over traditional FL models in NTN environments for traffic forecasting. The proposed Fed-KAN is a novel approach that utilises the functional approximation capabilities of KANs in a FL framework. We evaluate Fed-KAN compared to Fed-MLP on a traffic dataset of real satellite operator and show a significant reduction in training and test loss. Our results show that Fed-KAN can achieve a 77.39% reduction in average test loss compared to Fed-MLP, highlighting its improved performance and better generalization ability. At the end of the paper, we also discuss some potential applications of Fed-KAN within O-RAN and Fed-KAN usage for split functionalities in NTN architecture.
Authors: Aakash Patel, Tianqing Zhang, Camille Avestruz, Jeffrey Regier, the LSST Dark Energy Science Collaboration
Abstract: Neural posterior estimation (NPE), a type of amortized variational inference, is a computationally efficient means of constructing probabilistic catalogs of light sources from astronomical images. To date, NPE has not been used to perform inference in models with spatially varying covariates. However, ground-based astronomical images have spatially varying sky backgrounds and point spread functions (PSFs), and accounting for this variation is essential for constructing accurate catalogs of imaged light sources. In this work, we introduce a method of performing NPE with spatially varying backgrounds and PSFs. In this method, we generate synthetic catalogs and semi-synthetic images for these catalogs using randomly sampled PSF and background estimates from existing surveys. Using this data, we train a neural network, which takes an astronomical image and representations of its background and PSF as input, to output a probabilistic catalog. Our experiments with Sloan Digital Sky Survey data demonstrate the effectiveness of NPE in the presence of spatially varying backgrounds and PSFs for light source detection, star/galaxy separation, and flux measurement.
Authors: Lili Lu, Chuan Meng, Federico Ravenda, Mohammad Aliannejadi, Fabio Crestani
Abstract: Clarification need prediction (CNP) is a key task in conversational search, aiming to predict whether to ask a clarifying question or give an answer to the current user query. However, current research on CNP suffers from the issues of limited CNP training data and low efficiency. In this paper, we propose a zero-shot and efficient CNP framework (Zef-CNP), in which we first prompt large language models (LLMs) in a zero-shot manner to generate two sets of synthetic queries: ambiguous and specific (unambiguous) queries. We then use the generated queries to train efficient CNP models. Zef-CNP eliminates the need for human-annotated clarification-need labels during training and avoids the use of LLMs with high query latency at query time. To further improve the generation quality of synthetic queries, we devise a topic-, information-need-, and query-aware chain-of-thought (CoT) prompting strategy (TIQ-CoT). Moreover, we enhance TIQ-CoT with counterfactual query generation (CoQu), which guides LLMs first to generate a specific/ambiguous query and then sequentially generate its corresponding ambiguous/specific query. Experimental results show that Zef-CNP achieves superior CNP effectiveness and efficiency compared with zero- and few-shot LLM-based CNP predictors.
Authors: Federico Pizarro Bejarano, Bryson Jones, Daniel Pastor Moreno, Joseph Bowkett, Paul G. Backes, Angela P. Schoellig
Abstract: Diffusion models have revolutionized imitation learning, allowing robots to replicate complex behaviours. However, diffusion often relies on cameras and other exteroceptive sensors to observe the environment and lacks long-term memory. In space, military, and underwater applications, robots must be highly robust to failures in exteroceptive sensors, operating using only proprioceptive information. In this paper, we propose ProDapt, a method of incorporating long-term memory of previous contacts between the robot and the environment in the diffusion process, allowing it to complete tasks using only proprioceptive data. This is achieved by identifying "keypoints", essential past observations maintained as inputs to the policy. We test our approach using a UR10e robotic arm in both simulation and real experiments and demonstrate the necessity of this long-term memory for task completion.
Authors: Jiawei Zhang, Xuan Yang, Taiqi Wang, Yu Yao, Aleksandr Petiushko, Bo Li
Abstract: Traditional autonomous driving systems often struggle to integrate high-level reasoning with low-level control, resulting in suboptimal and sometimes unsafe driving behaviors. The emergence of Multimodal Large Language Models (MLLMs), which can process both visual and textual data, presents an opportunity to unify perception and reasoning tasks within a single framework. However, effectively embedding precise safety knowledge into MLLMs for autonomous driving remains a significant challenge. To address this, we propose SafeAuto, a novel framework that enhances MLLM-based autonomous driving systems by incorporating both unstructured and structured knowledge. Specifically, we first introduce the Position-Dependent Cross-Entropy (PDCE) loss function, designed to improve the accuracy of low-level control signal predictions when numerical values are represented as text. Second, to ensure safe autonomous driving by explicitly integrating precise safety knowledge into the MLLM, we develop a reasoning component for SafeAuto. This component translates driving safety regulations into first-order logic rules (e.g., "red light => stop") and incorporates these rules into a probabilistic graphical model, such as a Markov Logic Network (MLN). The MLN is trained to verify the predicted next actions using environmental attributes identified by attribute recognition models (e.g., detecting a red light) to form the predicates. Additionally, we construct a Multimodal RAG model that leverages video data, control signals, and environmental attributes to learn more effectively from past similar driving experiences. By integrating PDCE, MLN, and Multimodal RAG, SafeAuto significantly outperforms existing baselines across multiple datasets. This advancement enables more accurate, reliable, and safer autonomous driving systems that learn from experience, obey traffic laws, and perform precise control actions.
Authors: Alex Alberts, Ilias Bilionis
Abstract: Physics-informed machine learning is one of the most commonly used methods for fusing physical knowledge in the form of partial differential equations with experimental data. The idea is to construct a loss function where the physical laws take the place of a regularizer and minimize it to reconstruct the underlying physical fields and any missing parameters. However, there is a noticeable lack of a direct connection between physics-informed loss functions and an overarching Bayesian framework. In this work, we demonstrate that Brownian bridge Gaussian processes can be viewed as a softly-enforced physics-constrained prior for the Poisson equation. We first show equivalence between the variational form of the physics-informed loss function for the Poisson equation and a kernel ridge regression objective. Then, through the connection between Gaussian process regression and kernel methods, we identify a Gaussian process for which the posterior mean function and physics-informed loss function minimizer agree. This connection allows us to probe different theoretical questions, such as convergence and behavior of inverse problems. We also connect the method to the important problem of identifying model-form error in applications.
Authors: Sheng Long, Angelos Chatzimparmpas, Emma Alexander, Matthew Kay, Jessica Hullman
Abstract: Judging the similarity of visualizations is crucial to various applications, such as visualization-based search and visualization recommendation systems. Recent studies show deep-feature-based similarity metrics correlate well with perceptual judgments of image similarity and serve as effective loss functions for tasks like image super-resolution and style transfer. We explore the application of such metrics to judgments of visualization similarity. We extend a similarity metric using five ML architectures and three pre-trained weight sets. We replicate results from previous crowd-sourced studies on scatterplot and visual channel similarity perception. Notably, our metric using pre-trained ImageNet weights outperformed gradient-descent tuned MS-SSIM, a multi-scale similarity metric based on luminance, contrast, and structure. Our work contributes to understanding how deep-feature-based metrics can enhance similarity assessments in visualization, potentially improving visual analysis tools and techniques. Supplementary materials are available at https://osf.io/dj2ms.
URLs: https://osf.io/dj2ms.
Authors: Kaleab A. Kinfu, Ren\'e Vidal
Abstract: Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) have led to significant progress in 2D body pose estimation. However, achieving a good balance between accuracy, efficiency, and robustness remains a challenge. For instance, CNNs are computationally efficient but struggle with long-range dependencies, while ViTs excel in capturing such dependencies but suffer from quadratic computational complexity. This paper proposes two ViT-based models for accurate, efficient, and robust 2D pose estimation. The first one, EViTPose, operates in a computationally efficient manner without sacrificing accuracy by utilizing learnable joint tokens to select and process a subset of the most important body patches, enabling us to control the trade-off between accuracy and efficiency by changing the number of patches to be processed. The second one, UniTransPose, while not allowing for the same level of direct control over the trade-off, efficiently handles multiple scales by combining (1) an efficient multi-scale transformer encoder that uses both local and global attention with (2) an efficient sub-pixel CNN decoder for better speed and accuracy. Moreover, by incorporating all joints from different benchmarks into a unified skeletal representation, we train robust methods that learn from multiple datasets simultaneously and perform well across a range of scenarios -- including pose variations, lighting conditions, and occlusions. Experiments on six benchmarks demonstrate that the proposed methods significantly outperform state-of-the-art methods while improving computational efficiency. EViTPose exhibits a significant decrease in computational complexity (30% to 44% less in GFLOPs) with a minimal drop of accuracy (0% to 3.5% less), and UniTransPose achieves accuracy improvements ranging from 0.9% to 43.8% across these benchmarks.
Authors: Yuzhou Gu, Yanjun Han, Jian Qian
Abstract: We study the evolution of information in interactive decision making through the lens of a stochastic multi-armed bandit problem. Focusing on a fundamental example where a unique optimal arm outperforms the rest by a fixed margin, we characterize the optimal success probability and mutual information over time. Our findings reveal distinct growth phases in mutual information -- initially linear, transitioning to quadratic, and finally returning to linear -- highlighting curious behavioral differences between interactive and non-interactive environments. In particular, we show that optimal success probability and mutual information can be decoupled, where achieving optimal learning does not necessarily require maximizing information gain. These findings shed new light on the intricate interplay between information and learning in interactive decision making.
Authors: Marin Mlinarevi\'c (University College London), George K. Holt (STFC Hartree Centre), Adriano Agnello (STFC Hartree Centre)
Abstract: A surrogate model for particle-in-cell plasma simulations based on a graph neural network is presented. The graph is constructed in such a way as to enable the representation of electromagnetic fields on a fixed spatial grid. The model is applied to simulate beams of electrons in one dimension over a wide range of temperatures, drift momenta and densities, and is shown to reproduce two-stream instabilities - a common and fundamental plasma instability. Qualitatively, the characteristic phase-space mixing of counterpropagating electron beams is observed. Quantitatively, the model's performance is evaluated in terms of the accuracy of its predictions of number density distributions, the electric field, and their Fourier decompositions, particularly the growth rate of the fastest-growing unstable mode, as well as particle position, momentum distributions, energy conservation and run time. The model achieves high accuracy with a time step longer than conventional simulation by two orders of magnitude. This work demonstrates that complex plasma dynamics can be learned and shows promise for the development of fast differentiable simulators suitable for solving forward and inverse problems in plasma physics.
Authors: Hui Li, Zhiguo Wang, Bohui Chen, Li Sheng
Abstract: Equivariant neural networks play a pivotal role in analyzing datasets with symmetry properties, particularly in complex data structures. However, integrating equivariance with Markov properties presents notable challenges due to the inherent dependencies within such data. Previous research has primarily concentrated on establishing generalization bounds under the assumption of independently and identically distributed data, frequently neglecting the influence of Markov dependencies. In this study, we investigate the impact of Markov properties on generalization performance alongside the role of equivariance within this context. We begin by applying a new McDiarmid's inequality to derive a generalization bound for neural networks trained on Markov datasets, using Rademacher complexity as a central measure of model capacity. Subsequently, we utilize group theory to compute the covering number under equivariant constraints, enabling us to obtain an upper bound on the Rademacher complexity based on this covering number. This bound provides practical insights into selecting low-dimensional irreducible representations, enhancing generalization performance for fixed-width equivariant neural networks.
Authors: Raghav Gupta, Ryan Sullivan, Yunxuan Li, Samrat Phatale, Abhinav Rastogi
Abstract: Multi-objective preference alignment of large language models (LLMs) is critical for developing AI systems that are more configurable, personalizable, helpful, and safe. However, optimizing model outputs to satisfy diverse objectives with variable weights at inference time for truly personalized models presents a significant challenge. Existing approaches are either computationally expensive to train or do not sufficiently steer model behaviors. This paper introduces the Multi-Objective Online DPO (MO-ODPO) algorithm, designed to robustly and efficiently align model behaviors with multiple, potentially conflicting human preferences. Our approach incorporates a prompt conditioning mechanism, allowing us to train a single preference-conditional policy, that can adapt to new preference combinations at inference. Experiments on two popular benchmarks show that MO-ODPO Pareto-dominates existing baselines while providing excellent inference-time steerability between diverse objectives.
Authors: Benjamin Hoffman, David Robinson, Marius Miron, Vittorio Baglione, Daniela Canestrari, Damian Elias, Eva Trapote, Olivier Pietquin
Abstract: We propose a methodology for training foundation models that enhances their in-context learning capabilities within the domain of bioacoustic signal processing. We use synthetically generated training data, introducing a domain-randomization-based pipeline that constructs diverse acoustic scenes with temporally strong labels. We generate over 8.8 thousand hours of strongly-labeled audio and train a query-by-example, transformer-based model to perform few-shot bioacoustic sound event detection. Our second contribution is a public benchmark of 13 diverse few-shot bioacoustics tasks. Our model outperforms previously published methods by 49%, and we demonstrate that this is due to both model design and data scale. We make our trained model available via an API, to provide ecologists and ethologists with a training-free tool for bioacoustic sound event detection.
Authors: Tianci Liu, Ruirui Li, Yunzhe Qi, Hui Liu, Xianfeng Tang, Tianqi Zheng, Qingyu Yin, Monica Xiao Cheng, Jun Huan, Haoyu Wang, Jing Gao
Abstract: Large language models (LLMs) have achieved remarkable performance on various natural language tasks. However, they are trained on static corpora and their knowledge can become outdated quickly in the fast-changing world. This motivates the development of knowledge editing methods designed to update certain knowledge in LLMs without changing unrelated others. To make selective edits, previous efforts often sought to update a small amount of parameters in some specific layer(s) of a LLM. Nonetheless, in challenging scenarios, they still fall short in making successful edits while preserving knowledge irrelevant to the updates simultaneously, resulting in a notable editing-locality trade-off. In this work, we question if the trade-offs are caused by the fact that parameter-based updates have a global effect, i.e., edited parameters affect all inputs indiscriminately. In light of this, we explore the feasibility of representation fine-tuning, which applied some linear update to a few representations in a learned subspace, for knowledge editing. While being effective to enhance an LLM's general ability as demonstrated in the previous work, we theoretically show that this linear update imposes a tension in editing-locality trade-off. Subsequently, BaFT is proposed to break the linearity. BaFT computes a weight for each basis that spans a dimension of the subspace based on the input representation. This input-dependent weighting mechanism allows BaFT to manage different types of knowledge in an adaptive way, thereby achieving a better editing-locality trade-off. Experiments on three LLMs with five editing benchmarks in diverse scenarios show the superiority of our method.
Authors: Benjamin Schneider, Florian Kerschbaum, Wenhu Chen
Abstract: Visual embedding models excel at zero-shot tasks like visual retrieval and classification. However, these models cannot be used for tasks that contain ambiguity or require user instruction. These tasks necessitate a multimodal embedding model, which outputs embeddings that combine visual and natural language input. Existing CLIP-based approaches embed images and text independently, and fuse the result. We find that this results in weak interactions between modalities, and poor user control over the representation. We introduce ABC, an open-source multimodal embedding model that uses a vision-language model backbone to deeply integrate image features with natural language instructions. ABC achieves bestfor-size performance on MSCOCO image-to-text retrieval and is the top performing model on classification and VQA tasks in the Massive Multimodal Embedding Benchmark. With a strongly unified vision-language representation, ABC can use natural language to solve subtle and potentially ambiguous visual retrieval problems. To evaluate this capability, we design CtrlBench, a benchmark that requires interleaving textual instructions with image content for correct retrieval. ABC advances the state of multimodal embeddings by offering high-quality representations and flexible natural language control. Our model and datasets are available at our project page.
Authors: Anne-Men Huijzer, Thomas Chaffey, Bart Besselink, Henk J. van Waarde
Abstract: Energy-based learning algorithms are alternatives to backpropagation and are well-suited to distributed implementations in analog electronic devices. However, a rigorous theory of convergence is lacking. We make a first step in this direction by analysing a particular energy-based learning algorithm, Contrastive Learning, applied to a network of linear adjustable resistors. It is shown that, in this setup, Contrastive Learning is equivalent to projected gradient descent on a convex function, for any step size, giving a guarantee of convergence for the algorithm.
Authors: Smruti P. Dash, Kedar V. Khandeparkar, Nipun Agrawal
Abstract: The modern power grids are integrated with digital technologies and automation systems. The inclusion of digital technologies has made the smart grids vulnerable to cyber-attacks. Cyberattacks on smart grids can compromise data integrity and jeopardize the reliability of the power supply. Traditional intrusion detection systems often need help to effectively detect novel and sophisticated attacks due to their reliance on labeled training data, which may only encompass part of the spectrum of potential threats. This work proposes a semi-supervised method for cyber-attack detection in smart grids by leveraging the labeled and unlabeled measurement data. We implement consistency regularization and pseudo-labeling to identify deviations from expected behavior and predict the attack classes. We use a curriculum learning approach to improve pseudo-labeling performance, capturing the model uncertainty. We demonstrate the efficiency of the proposed method in detecting different types of cyberattacks, minimizing the false positives by implementing them on publicly available datasets. The method proposes a promising solution by improving the detection accuracy to 99% in the presence of unknown samples and significantly reducing false positives.
Authors: Hamed Khosravi, Mohammad Reza Shafie, Ahmed Shoyeb Raihan, Srinjoy Das, Imtiaz Ahmed
Abstract: Existing contextual multi-armed bandit (MAB) algorithms fail to effectively capture both long-term trends and local patterns across all arms, leading to suboptimal performance in environments with rapidly changing reward structures. They also rely on static exploration rates, which do not dynamically adjust to changing conditions. To overcome these limitations, we propose LNUCB-TA, a hybrid bandit model integrating a novel nonlinear component (adaptive k-Nearest Neighbors (k-NN)) for reducing time complexity, alongside a global-and-local attention-based exploration mechanism. Our approach uniquely combines linear and nonlinear estimation techniques, with the nonlinear module dynamically adjusting k based on reward variance to enhance spatiotemporal pattern recognition. This reduces the likelihood of selecting suboptimal arms while improving reward estimation accuracy and computational efficiency. The attention-based mechanism ranks arms by past performance and selection frequency, dynamically adjusting exploration and exploitation in real time without requiring manual tuning of exploration rates. By integrating global attention (assessing all arms collectively) and local attention (focusing on individual arms), LNUCB-TA efficiently adapts to temporal and spatial complexities. Empirical results show LNUCB-TA significantly outperforms state-of-the-art linear, nonlinear, and hybrid bandits in cumulative and mean reward, convergence, and robustness across different exploration rates. Theoretical analysis further confirms its reliability with a sub-linear regret bound.
Authors: Tianyu Huai, Jie Zhou, Xingjiao Wu, Qin Chen, Qingchun Bai, Ze Zhou, Liang He
Abstract: Multimodal large language models (MLLMs) have garnered widespread attention from researchers due to their remarkable understanding and generation capabilities in visual language tasks (e.g., visual question answering). However, the rapid pace of knowledge updates in the real world makes offline training of MLLMs costly, and when faced with non-stationary data streams, MLLMs suffer from catastrophic forgetting during learning. In this paper, we propose an MLLMs-based dual momentum Mixture-of-Experts (CL-MoE) framework for continual visual question answering (VQA). We integrate MLLMs with continual learning to utilize the rich commonsense knowledge in LLMs. We introduce a Dual-Router MoE (RMoE) strategy to select the global and local experts using task-level and instance-level routers, to robustly assign weights to the experts most appropriate for the task. Then, we design a dynamic Momentum MoE (MMoE) to update the parameters of experts dynamically based on the relationships between the experts and tasks/instances, so that the model can absorb new knowledge while maintaining existing knowledge. The extensive experimental results indicate that our method achieves state-of-the-art performance on 10 VQA tasks, proving the effectiveness of our approach.
Authors: Pedram Asef, Christopher Vagg
Abstract: Advanced slot and winding designs are imperative to create future high performance electrical machines (EM). As a result, the development of methods to design and improve slot filling factor (SFF) has attracted considerable research. Recent developments in manufacturing processes, such as additive manufacturing and alternative materials, has also highlighted a need for novel high-fidelity design techniques to develop high performance complex geometries and topologies. This study therefore introduces a novel physics-informed machine learning (PIML) design optimization process for improving SFF in traction electrical machines used in electric vehicles. A maximum entropy sampling algorithm (MESA) is used to seed a physics-informed Bayesian optimization (PIBO) algorithm, where the target function and its approximations are produced by Gaussian processes (GP)s. The proposed PIBO-MESA is coupled with a 2D finite element model (FEM) to perform a GP-based surrogate and provide the first demonstration of the optimal combination of complex design variables for an electrical machine. Significant computational gains were achieved using the new PIBO-MESA approach, which is 45% faster than existing stochastic methods, such as the non-dominated sorting genetic algorithm II (NSGA-II). The FEM results confirm that the new design optimization process and keystone shaped wires lead to a higher SFF (i.e. by 20%) and electromagnetic improvements (e.g. maximum torque by 12%) with similar resistivity. The newly developed PIBO-MESA design optimization process therefore presents significant benefits in the design of high-performance electric machines, with reduced development time and costs.
Authors: Lixu Wang, Bingqi Shang, Yi Li, Payal Mohapatra, Wei Dong, Xiao Wang, Qi Zhu
Abstract: Vision Transformers (ViTs), extensively pre-trained on large-scale datasets, have become essential to foundation models, allowing excellent performance on diverse downstream tasks with minimal adaptation. Consequently, there is growing interest in adapting pre-trained ViTs across various fields, including privacy-sensitive domains where clients are often reluctant to share their data. Existing adaptation methods typically require direct data access, rendering them infeasible under these constraints. A straightforward solution may be sending the pre-trained ViT to clients for local adaptation, which poses issues of model intellectual property protection and incurs heavy client computation overhead. To address these issues, we propose a novel split adaptation (SA) method that enables effective downstream adaptation while protecting data and models. SA, inspired by split learning (SL), segments the pre-trained ViT into a frontend and a backend, with only the frontend shared with the client for data representation extraction. But unlike regular SL, SA replaces frontend parameters with low-bit quantized values, preventing direct exposure of the model. SA allows the client to add bi-level noise to the frontend and the extracted data representations, ensuring data protection. Accordingly, SA incorporates data-level and model-level out-of-distribution enhancements to mitigate noise injection's impact on adaptation performance. Our SA focuses on the challenging few-shot adaptation and adopts patch retrieval augmentation for overfitting alleviation. Extensive experiments on multiple datasets validate SA's superiority over state-of-the-art methods and demonstrate its defense against advanced data reconstruction attacks while preventing model leakage with minimal computation cost on the client side. The source codes can be found at https://github.com/conditionWang/Split_Adaptation.
Authors: Roman Remme, Tobias Kaczun, Tim Ebert, Christof A. Gehrig, Dominik Geng, Gerrit Gerhartz, Marc K. Ickler, Manuel V. Klockow, Peter Lippmann, Johannes S. Schmidt, Simon Wagner, Andreas Dreuw, Fred A. Hamprecht
Abstract: Hohenberg and Kohn have proven that the electronic energy and the one-particle electron density can, in principle, be obtained by minimizing an energy functional with respect to the density. Given that decades of theoretical work have so far failed to produce this elusive exact energy functional promising great computational savings, it is reasonable to try and learn it empirically. Using rotationally equivariant atomistic machine learning, we obtain for the first time a density functional that, when applied to the organic molecules in QM9, yields energies with chemical accuracy while also converging to meaningful electron densities. Augmenting the training data with densities obtained from perturbed potentials proved key to these advances. Altogether, we are now closer than ever to fulfilling Hohenberg and Kohn's promise, paving the way for more efficient calculations in large molecular systems.
Authors: Earnest Paul Ijjina, Aniruddha Srinivas Joshi, Goutham Kanahasabai, Keerthi Priyanka P
Abstract: The analysis of sales information, is a vital step in designing an effective marketing strategy. This work proposes a novel approach to analyse the shopping behaviour of customers to identify their purchase patterns. An extended version of the Multi-Cluster Overlapping k-Means Extension (MCOKE) algorithm with weighted k-Means algorithm is utilized to map customers to the garments of interest. The age & gender traits of the customer; the time spent and the expressions exhibited while selecting garments for purchase, are utilized to associate a customer or a group of customers to a garments they are interested in. Such study on the customer base of a retail business, may help in inferring the products of interest of their consumers, and enable them in developing effective business strategies, thus ensuring customer satisfaction, loyalty, increased sales and profits.
Authors: Earnest Paul Ijjina, Goutham Kanahasabai, Aniruddha Srinivas Joshi
Abstract: In the current information era, customer analytics play a key role in the success of any business. Since customer demographics primarily dictate their preferences, identification and utilization of age & gender information of customers in sales forecasting, may maximize retail sales. In this work, we propose a computer vision based approach to age and gender prediction in surveillance video. The proposed approach leverage the effectiveness of Wide Residual Networks and Xception deep learning models to predict age and gender demographics of the consumers. The proposed approach is designed to work with raw video captured in a typical CCTV video surveillance system. The effectiveness of the proposed approach is evaluated on real-life garment store surveillance video, which is captured by low resolution camera, under non-uniform illumination, with occlusions due to crowding, and environmental noise. The system can also detect customer facial expressions during purchase in addition to demographics, that can be utilized to devise effective marketing strategies for their customer base, to maximize sales.
Authors: Weihao Lu, Haobo Zhang, Yicheng Li, Qian Lin
Abstract: The saturation effects, which originally refer to the fact that kernel ridge regression (KRR) fails to achieve the information-theoretical lower bound when the regression function is over-smooth, have been observed for almost 20 years and were rigorously proved recently for kernel ridge regression and some other spectral algorithms over a fixed dimensional domain. The main focus of this paper is to explore the saturation effects for a large class of spectral algorithms (including the KRR, gradient descent, etc.) in large dimensional settings where $n \asymp d^{\gamma}$. More precisely, we first propose an improved minimax lower bound for the kernel regression problem in large dimensional settings and show that the gradient flow with early stopping strategy will result in an estimator achieving this lower bound (up to a logarithmic factor). Similar to the results in KRR, we can further determine the exact convergence rates (both upper and lower bounds) of a large class of (optimal tuned) spectral algorithms with different qualification $\tau$'s. In particular, we find that these exact rate curves (varying along $\gamma$) exhibit the periodic plateau behavior and the polynomial approximation barrier. Consequently, we can fully depict the saturation effects of the spectral algorithms and reveal a new phenomenon in large dimensional settings (i.e., the saturation effect occurs in large dimensional setting as long as the source condition $s>\tau$ while it occurs in fixed dimensional setting as long as $s>2\tau$).
Authors: Zhan Qu, Shuzhou Yuan, Michael F\"arber, Marius Brennfleck, Niklas Wartha, Anton Stephan
Abstract: Wake vortices - strong, coherent air turbulences created by aircraft - pose a significant risk to aviation safety and therefore require accurate and reliable detection methods. In this paper, we present an advanced, explainable machine learning method that utilizes Light Detection and Ranging (LiDAR) data for effective wake vortex detection. Our method leverages a dynamic graph CNN (DGCNN) with semantic segmentation to partition a 3D LiDAR point cloud into meaningful segments. Further refinement is achieved through clustering techniques. A novel feature of our research is the use of a perturbation-based explanation technique, which clarifies the model's decision-making processes for air traffic regulators and controllers, increasing transparency and building trust. Our experimental results, based on measured and simulated LiDAR scans compared against four baseline methods, underscore the effectiveness and reliability of our approach. This combination of semantic segmentation and clustering for real-time wake vortex tracking significantly advances aviation safety measures, ensuring that these are both effective and comprehensible.
Authors: Wanli Hong, Yuliang Shi, Jonathan Niles-Weed
Abstract: Motivated by applications in trajectory inference and particle tracking, we introduce Smooth Schr\"odinger Bridges. Our proposal generalizes prior work by allowing the reference process in the Schr\"odinger Bridge problem to be a smooth Gaussian process, leading to more regular and interpretable trajectories in applications. Though na\"ively smoothing the reference process leads to a computationally intractable problem, we identify a class of processes (including the Mat\'ern processes) for which the resulting Smooth Schr\"odinger Bridge problem can be lifted to a simpler problem on phase space, which can be solved in polynomial time. We develop a practical approximation of this algorithm that outperforms existing methods on numerous simulated and real single-cell RNAseq datasets. The code can be found at https://github.com/WanliHongC/Smooth_SB
Authors: Haofei Lu, Zhe Wu, Junliang Xing, Jianshu Li, Ruoyu Li, Zhe Li, Yuanchun Shi
Abstract: Embodiment co-design aims to optimize a robot's morphology and control policy simultaneously. While prior work has demonstrated its potential for generating environment-adaptive robots, this field still faces persistent challenges in optimization efficiency due to the (i) combinatorial nature of morphological search spaces and (ii) intricate dependencies between morphology and control. We prove that the ineffective morphology representation and unbalanced reward signals between the design and control stages are key obstacles to efficiency. To advance towards efficient embodiment co-design, we propose BodyGen, which utilizes (1) topology-aware self-attention for both design and control, enabling efficient morphology representation with lightweight model sizes; (2) a temporal credit assignment mechanism that ensures balanced reward signals for optimization. With our findings, Body achieves an average 60.03% performance improvement against state-of-the-art baselines. We provide codes and more results on the website: https://genesisorigin.github.io.
Authors: Tiansheng Huang, Sihao Hu, Fatih Ilhan, Selim Furkan Tekin, Zachary Yahn, Yichang Xu, Ling Liu
Abstract: Safety alignment is an important procedure before the official deployment of a Large Language Model (LLM). While safety alignment has been extensively studied for LLM, there is still a large research gap for Large Reasoning Models (LRMs) that equip with improved reasoning capability. We in this paper systematically examine a simplified pipeline for producing safety aligned LRMs. With our evaluation of various LRMs, we deliver two main findings: i) Safety alignment can be done upon the LRM to restore its safety capability. ii) Safety alignment leads to a degradation of the reasoning capability of LRMs. The two findings show that there exists a trade-off between reasoning and safety capability with the sequential LRM production pipeline. The discovered trade-off, which we name Safety Tax, should shed light on future endeavors of safety research on LRMs. As a by-product, we curate a dataset called DirectRefusal, which might serve as an alternative dataset for safety alignment. Our source code is available at https://github.com/git-disl/Safety-Tax.
Authors: Sakshi Arya, Hyebin Song
Abstract: The multi-armed bandits (MAB) framework is a widely used approach for sequential decision-making, where a decision-maker selects an arm in each round with the goal of maximizing long-term rewards. Moreover, in many practical applications, such as personalized medicine and recommendation systems, feedback is provided in batches, contextual information is available at the time of decision-making, and rewards from different arms are related rather than independent. We propose a novel semi-parametric framework for batched bandits with covariates and a shared parameter across arms, leveraging the single-index regression (SIR) model to capture relationships between arm rewards while balancing interpretability and flexibility. Our algorithm, Batched single-Index Dynamic binning and Successive arm elimination (BIDS), employs a batched successive arm elimination strategy with a dynamic binning mechanism guided by the single-index direction. We consider two settings: one where a pilot direction is available and another where the direction is estimated from data, deriving theoretical regret bounds for both cases. When a pilot direction is available with sufficient accuracy, our approach achieves minimax-optimal rates (with $d = 1$) for nonparametric batched bandits, circumventing the curse of dimensionality. Extensive experiments on simulated and real-world datasets demonstrate the effectiveness of our algorithm compared to the nonparametric batched bandit method introduced by \cite{jiang2024batched}.
Authors: Adrian Buganza Tepole, Asghar Jadoon, Manuel Rausch, Jan N. Fuhg
Abstract: Accurate constitutive models of soft materials are crucial for understanding their mechanical behavior and ensuring reliable predictions in the design process. To this end, scientific machine learning research has produced flexible and general material model architectures that can capture the behavior of a wide range of materials, reducing the need for expert-constructed closed-form models. The focus has gradually shifted towards embedding physical constraints in the network architecture to regularize these over-parameterized models. Two popular approaches are input convex neural networks (ICNN) and neural ordinary differential equations (NODE). A related alternative has been the generalization of closed-form models, such as sparse regression from a large library. Remarkably, all prior work using ICNN or NODE uses the invariants of the Cauchy-Green tensor and none uses the principal stretches. In this work, we construct general polyconvex functions of the principal stretches in a physics-aware deep-learning framework and offer insights and comparisons to invariant-based formulations. The framework is based on recent developments to characterize polyconvex functions in terms of convex functions of the right stretch tensor $\mathbf{U}$, its cofactor $\text{cof}\mathbf{U}$, and its determinant $J$. Any convex function of a symmetric second-order tensor can be described with a convex and symmetric function of its eigenvalues. Thus, we first describe convex functions of $\mathbf{U}$ and $\text{cof}\mathbf{U}$ in terms of their respective eigenvalues using deep Holder sets composed with ICNN functions. A third ICNN takes as input $J$ and the two convex functions of $\mathbf{U}$ and $\text{cof}\mathbf{U}$, and returns the strain energy as output. The ability of the model to capture arbitrary materials is demonstrated using synthetic and experimental data.
Authors: Atul Anand Gopalakrishnan, Jakir Hossain, Tugrulcan Elmas, Ahmet Erdem Sariyuce
Abstract: Social media users and inauthentic accounts, such as bots, may coordinate in promoting their topics. Such topics may give the impression that they are organically popular among the public, even though they are astroturfing campaigns that are centrally managed. It is challenging to predict if a topic is organic or a coordinated campaign due to the lack of reliable ground truth. In this paper, we create such ground truth by detecting the campaigns promoted by ephemeral astroturfing attacks. These attacks push any topic to Twitter's (X) trends list by employing bots that tweet in a coordinated manner in a short period and then immediately delete their tweets. We manually curate a dataset of organic Twitter trends. We then create engagement networks out of these datasets which can serve as a challenging testbed for graph classification task to distinguish between campaigns and organic trends. Engagement networks consist of users as nodes and engagements as edges (retweets, replies, and quotes) between users. We release the engagement networks for 179 campaigns and 135 non-campaigns, and also provide finer-grain labels to characterize the type of the campaigns and non-campaigns. Our dataset, LEN (Large Engagement Networks), is available in the URL below. In comparison to traditional graph classification datasets, which are small with tens of nodes and hundreds of edges at most, graphs in LEN are larger. The average graph in LEN has ~11K nodes and ~23K edges. We show that state-of-the-art GNN methods give only mediocre results for campaign vs. non-campaign and campaign type classification on LEN. LEN offers a unique and challenging playfield for the graph classification problem. We believe that LEN will help advance the frontiers of graph classification techniques on large networks and also provide an interesting use case in terms of distinguishing coordinated campaigns and organic trends.
Authors: Lin An, Andrew A. Li, Vaisnavi Nemala, Gabriel Visotsky
Abstract: Real-time personalization has advanced significantly in recent years, with platforms utilizing machine learning models to predict user preferences based on rich behavioral data on each individual user. Traditional approaches usually rely on embedding-based machine learning models to capture user preferences, and then reduce the final optimization task to nearest-neighbors, which can be performed extremely fast. However, these models struggle to capture complex user behaviors, which are essential for making accurate recommendations. Transformer-based models, on the other hand, are known for their practical ability to model sequential behaviors, and hence have been intensively used in personalization recently to overcome these limitations. However, optimizing recommendations under transformer-based models is challenging due to their complicated architectures. In this paper, we address this challenge by considering a specific class of transformers, showing its ability to represent complex user preferences, and developing efficient algorithms for real-time personalization. We focus on a particular set of transformers, called simple transformers, which contain a single self-attention layer. We show that simple transformers are capable of capturing complex user preferences. We then develop an algorithm that enables fast optimization of recommendation tasks based on simple transformers. Our algorithm achieves near-optimal performance in sub-linear time. Finally, we demonstrate the effectiveness of our approach through an empirical study on datasets from Spotify and Trivago. Our experiment results show that (1) simple transformers can model/predict user preferences substantially more accurately than non-transformer models and nearly as accurately as more complex transformers, and (2) our algorithm completes simple-transformer-based recommendation tasks quickly and effectively.
Authors: Faye Zhang, Jasmine Wan, Qianyu Cheng, Jinfeng Rao
Abstract: Online platforms like Pinterest hosting vast content collections traditionally rely on manual curation or user-generated search logs to create keyword landing pages (KLPs) -- topic-centered collection pages that serve as entry points for content discovery. While manual curation ensures quality, it doesn't scale to millions of collections, and search log approaches result in limited topic coverage and imprecise content matching. In this paper, we present PinLanding, a novel content-first architecture that transforms the way platforms create topical collections. Instead of deriving topics from user behavior, our system employs a multi-stage pipeline combining vision-language model (VLM) for attribute extraction, large language model (LLM) for topic generation, and a CLIP-based dual-encoder architecture for precise content matching. Our model achieves 99.7% Recall@10 on Fashion200K benchmark, demonstrating strong attribute understanding capabilities. In production deployment for search engine optimization with 4.2 million shopping landing pages, the system achieves a 4X increase in topic coverage and 14.29% improvement in collection attribute precision over the traditional search log-based approach via human evaluation. The architecture can be generalized beyond search traffic to power various user experiences, including content discovery and recommendations, providing a scalable solution to transform unstructured content into curated topical collections across any content domain.
Authors: Jianqing Fan, Yingying Fan, Jinchi Lv, Fan Yang, Diwen Yu
Abstract: Laplacian matrices are commonly employed in many real applications, encoding the underlying latent structural information such as graphs and manifolds. The use of the normalization terms naturally gives rise to random matrices with dependency. It is well-known that dependency is a major bottleneck of new random matrix theory (RMT) developments. To this end, in this paper, we formally introduce a class of generalized (and regularized) Laplacian matrices, which contains the Laplacian matrix and the random adjacency matrix as a specific case, and suggest the new framework of the asymptotic theory of eigenvectors for latent embeddings with generalized Laplacian matrices (ATE-GL). Our new theory is empowered by the tool of generalized quadratic vector equation for dealing with RMT under dependency, and delicate high-order asymptotic expansions of the empirical spiked eigenvectors and eigenvalues based on local laws. The asymptotic normalities established for both spiked eigenvectors and eigenvalues will enable us to conduct precise inference and uncertainty quantification for applications involving the generalized Laplacian matrices with flexibility. We discuss some applications of the suggested ATE-GL framework and showcase its validity through some numerical examples.
Authors: Gian Marco Visani, Michael N. Pun, Anastasia A. Minervina, Philip Bradley, Paul Thomas, Armita Nourmohammad
Abstract: T-cells play a key role in adaptive immunity by mounting specific responses against diverse pathogens. An effective binding between T-cell receptors (TCRs) and pathogen-derived peptides presented on Major Histocompatibility Complexes (MHCs) mediate an immune response. However, predicting these interactions remains challenging due to limited functional data on T-cell reactivities. Here, we introduce a computational approach to predict TCR interactions with peptides presented on MHC class I alleles, and to design novel immunogenic peptides for specified TCR-MHC complexes. Our method leverages HERMES, a structure-based, physics-guided machine learning model trained on the protein universe to predict amino acid preferences based on local structural environments. Despite no direct training on TCR-pMHC data, the implicit physical reasoning in HERMES enables us to make accurate predictions of both TCR-pMHC binding affinities and T-cell activities across diverse viral epitopes and cancer neoantigens, achieving up to 72% correlation with experimental data. Leveraging our TCR recognition model, we develop a computational protocol for de novo design of immunogenic peptides. Through experimental validation in three TCR-MHC systems targeting viral and cancer peptides, we demonstrate that our designs--with up to five substitutions from the native sequence--activate T-cells at success rates of up to 50%. Lastly, we use our generative framework to quantify the diversity of the peptide recognition landscape for various TCR-MHC complexes, offering key insights into T-cell specificity in both humans and mice. Our approach provides a platform for immunogenic peptide and neoantigen design, opening new computational paths for T-cell vaccine development against viruses and cancer.
Authors: Pouya Fathollahzadeh, Mariam El Mezouar, Hao Li, Ying Zou, Ahmed E. Hassan
Abstract: In software engineering chatrooms, communication is often hindered by imprecise questions that cannot be answered. Recognizing key entities can be essential for improving question clarity and facilitating better exchange. However, existing research using natural language processing techniques often overlooks these software-specific nuances. In this paper, we introduce Software-specific Named Entity Recognition, Intent Detection, and Resolution Classification (SENIR), a labeling approach that leverages a Large Language Model to annotate entities, intents, and resolution status in developer chatroom conversations. To offer quantitative guidance for improving question clarity and resolvability, we build a resolution prediction model that leverages SENIR's entity and intent labels along with additional predictive features. We evaluate SENIR on the DISCO dataset using a subset of annotated chatroom dialogues. SENIR achieves an 86% F-score for entity recognition, a 71% F-score for intent detection, and an 89% F-score for resolution status classification. Furthermore, our resolution prediction model, tested with various sampling strategies (random undersampling and oversampling with SMOTE) and evaluation methods (5-fold cross-validation, 10-fold cross-validation, and bootstrapping), demonstrates AUC values ranging from 0.7 to 0.8. Key factors influencing resolution include positive sentiment and entities such as Programming Language and User Variable across multiple intents, while diagnostic entities are more relevant in error-related questions. Moreover, resolution rates vary significantly by intent: questions about API Usage and API Change achieve higher resolution rates, whereas Discrepancy and Review have lower resolution rates. A Chi-Square analysis confirms the statistical significance of these differences.
Authors: Maria Ana Cardei, Afsaneh Doryab
Abstract: Mass casualty incidents (MCIs) are a growing concern, characterized by complexity and uncertainty that demand adaptive decision-making strategies. The victim tagging step in the emergency medical response must be completed quickly and is crucial for providing information to guide subsequent time-constrained response actions. In this paper, we present a mathematical formulation of multi-agent victim tagging to minimize the time it takes for responders to tag all victims. Five distributed heuristics are formulated and evaluated with simulation experiments. The heuristics considered are on-the go, practical solutions that represent varying levels of situational uncertainty in the form of global or local communication capabilities, showcasing practical constraints. We further investigate the performance of a multi-agent reinforcement learning (MARL) strategy, factorized deep Q-network (FDQN), to minimize victim tagging time as compared to baseline heuristics. Extensive simulations demonstrate that between the heuristics, methods with local communication are more efficient for adaptive victim tagging, specifically choosing the nearest victim with the option to replan. Analyzing all experiments, we find that our FDQN approach outperforms heuristics in smaller-scale scenarios, while heuristics excel in more complex scenarios. Our experiments contain diverse complexities that explore the upper limits of MARL capabilities for real-world applications and reveal key insights.
Authors: Haoyu Li, Srikanth Kandula, Maria Angels de Luis Balaguer, Aditya Akella, Venkat Arun
Abstract: Analyzing large datasets requires responsive query execution, but executing SQL queries on massive datasets can be slow. This paper explores whether query execution can begin even before the user has finished typing, allowing results to appear almost instantly. We propose SpeQL, a system that leverages Large Language Models (LLMs) to predict likely queries based on the database schema, the user's past queries, and their incomplete query. Since exact query prediction is infeasible, SpeQL speculates on partial queries in two ways: 1) it predicts the query structure to compile and plan queries in advance, and 2) it precomputes smaller temporary tables that are much smaller than the original database, but are still predicted to contain all information necessary to answer the user's final query. Additionally, SpeQL continuously displays results for speculated queries and subqueries in real time, aiding exploratory analysis. A utility/user study showed that SpeQL improved task completion time, and participants reported that its speculative display of results helped them discover patterns in the data more quickly. In the study, SpeQL improves user's query latency by up to $289\times$ and kept the overhead reasonable, at $\$4$ per hour.
Authors: Iman Modarressi, Jann Spiess, Amar Venugopal
Abstract: We propose a machine-learning tool that yields causal inference on text in randomized trials. Based on a simple econometric framework in which text may capture outcomes of interest, our procedure addresses three questions: First, is the text affected by the treatment? Second, which outcomes is the effect on? And third, how complete is our description of causal effects? To answer all three questions, our approach uses large language models (LLMs) that suggest systematic differences across two groups of text documents and then provides valid inference based on costly validation. Specifically, we highlight the need for sample splitting to allow for statistical validation of LLM outputs, as well as the need for human labeling to validate substantive claims about how documents differ across groups. We illustrate the tool in a proof-of-concept application using abstracts of academic manuscripts.
Authors: Ziwei Huang, Jianan Zhou, Zhiguang Cao, Yixin Xu
Abstract: Light decoder-based solvers have gained popularity for solving vehicle routing problems (VRPs) due to their efficiency and ease of integration with reinforcement learning algorithms. However, they often struggle with generalization to larger problem instances or different VRP variants. This paper revisits light decoder-based approaches, analyzing the implications of their reliance on static embeddings and the inherent challenges that arise. Specifically, we demonstrate that in the light decoder paradigm, the encoder is implicitly tasked with capturing information for all potential decision scenarios during solution construction within a single set of embeddings, resulting in high information density. Furthermore, our empirical analysis reveals that the overly simplistic decoder struggles to effectively utilize this dense information, particularly as task complexity increases, which limits generalization to out-of-distribution (OOD) settings. Building on these insights, we show that enhancing the decoder capacity, with a simple addition of identity mapping and a feed-forward layer, can considerably alleviate the generalization issue. Experimentally, our method significantly enhances the OOD generalization of light decoder-based approaches on large-scale instances and complex VRP variants, narrowing the gap with the heavy decoder paradigm. Our code is available at: https://github.com/ziweileonhuang/reld-nco.
Authors: Elahe Delavari, Aws Khalil, Jaerock Kwon
Abstract: End-to-end vision-based imitation learning has demonstrated promising results in autonomous driving by learning control commands directly from expert demonstrations. However, traditional approaches rely on either regressionbased models, which provide precise control but lack confidence estimation, or classification-based models, which offer confidence scores but suffer from reduced precision due to discretization. This limitation makes it challenging to quantify the reliability of predicted actions and apply corrections when necessary. In this work, we introduce a dual-head neural network architecture that integrates both regression and classification heads to improve decision reliability in imitation learning. The regression head predicts continuous driving actions, while the classification head estimates confidence, enabling a correction mechanism that adjusts actions in low-confidence scenarios, enhancing driving stability. We evaluate our approach in a closed-loop setting within the CARLA simulator, demonstrating its ability to detect uncertain actions, estimate confidence, and apply real-time corrections. Experimental results show that our method reduces lane deviation and improves trajectory accuracy by up to 50%, outperforming conventional regression-only models. These findings highlight the potential of classification-guided confidence estimation in enhancing the robustness of vision-based imitation learning for autonomous driving. The source code is available at https://github.com/ElaheDlv/Confidence_Aware_IL.
Authors: Gal Yefet, Ayellet Tal
Abstract: This study addresses the challenge of self-supervised learning for 3D mesh analysis. It presents an new approach that uses random walks as a form of data augmentation to generate diverse representations of mesh surfaces. Furthermore, it employs a combination of contrastive and clustering losses. The contrastive learning framework maximizes similarity between augmented instances of the same mesh while minimizing similarity between different meshes. We integrate this with a clustering loss, enhancing class distinction across training epochs and mitigating training variance. Our model's effectiveness is evaluated using mean Average Precision (mAP) scores and a supervised SVM linear classifier on extracted features, demonstrating its potential for various downstream tasks such as object classification and shape retrieval.
Authors: Yalun Dai, Lingao Xiao, Ivor W. Tsang, Yang He
Abstract: Existing dataset pruning techniques primarily focus on classification tasks, limiting their applicability to more complex and practical tasks like instance segmentation. Instance segmentation presents three key challenges: pixel-level annotations, instance area variations, and class imbalances, which significantly complicate dataset pruning efforts. Directly adapting existing classification-based pruning methods proves ineffective due to their reliance on time-consuming model training process. To address this, we propose a novel Training-Free Dataset Pruning (TFDP) method for instance segmentation. Specifically, we leverage shape and class information from image annotations to design a Shape Complexity Score (SCS), refining it into a Scale-Invariant (SI-SCS) and Class-Balanced (CB-SCS) versions to address instance area variations and class imbalances, all without requiring model training. We achieve state-of-the-art results on VOC 2012, Cityscapes, and COCO datasets, generalizing well across CNN and Transformer architectures. Remarkably, our approach accelerates the pruning process by an average of 1349$\times$ on COCO compared to the adapted baselines. Source code is available at: https://github.com/he-y/dataset-pruning-for-instance-segmentation
URLs: https://github.com/he-y/dataset-pruning-for-instance-segmentation
Authors: Zirui Zhao, Junchao Xia, Si Wu, Xiaoke Wang, Guanping Xu, Yinghao Zhu, Jing Sun, Hai-Feng Li
Abstract: In recent years, researchers have increasingly sought batteries as an efficient and cost-effective solution for energy storage and supply, owing to their high energy density, low cost, and environmental resilience. However, the issue of dendrite growth has emerged as a significant obstacle in battery development. Excessive dendrite growth during charging and discharging processes can lead to battery short-circuiting, degradation of electrochemical performance, reduced cycle life, and abnormal exothermic events. Consequently, understanding the dendrite growth process has become a key challenge for researchers. In this study, we investigated dendrite growth mechanisms in batteries using a combined machine learning approach, specifically a two-dimensional artificial convolutional neural network (CNN) model, along with computational methods. We developed two distinct computer models to predict dendrite growth in batteries. The CNN-1 model employs standard convolutional neural network techniques for dendritic growth prediction, while CNN-2 integrates additional physical parameters to enhance model robustness. Our results demonstrate that CNN-2 significantly enhances prediction accuracy, offering deeper insights into the impact of physical factors on dendritic growth. This improved model effectively captures the dynamic nature of dendrite formation, exhibiting high accuracy and sensitivity. These findings contribute to the advancement of safer and more reliable energy storage systems.
Authors: Miao Peng, Nuo Chen, Zongrui Suo, Jia Li
Abstract: Despite significant advancements in Large Language Models (LLMs), developing advanced reasoning capabilities in LLMs remains a key challenge. Process Reward Models (PRMs) have demonstrated exceptional promise in enhancing reasoning by providing step-wise feedback, particularly in the context of mathematical reasoning. However, their application to broader reasoning domains remains understudied, largely due to the high costs associated with manually creating step-level supervision. In this work, we explore the potential of PRMs in graph reasoning problems - a domain that demands sophisticated multi-step reasoning and offers opportunities for automated step-level data generation using established graph algorithms. We introduce GraphSILO, the largest dataset for graph reasoning problems with fine-grained step-wise labels, built using automated Task-oriented Trajectories and Monte Carlo Tree Search (MCTS) to generate detailed reasoning steps with step-wise labels. Building upon this dataset, we train GraphPRM, the first PRM designed for graph reasoning problems, and evaluate its effectiveness in two key settings: inference-time scaling and reinforcement learning via Direct Preference Optimization (DPO). Experimental results show that GraphPRM significantly improves LLM performance across 13 graph reasoning tasks, delivering a 9% gain for Qwen2.5-7B and demonstrating transferability to new graph reasoning datasets and new reasoning domains like mathematical problem-solving. Notably, GraphPRM enhances LLM performance on GSM8K and Math500, underscoring the cross-domain applicability of graph-based reasoning rewards. Our findings highlight the potential of PRMs in advancing reasoning across diverse domains, paving the way for more versatile and effective LLMs.
Authors: Samet Demir, Zafer Dogan
Abstract: In this work, we study the training and generalization performance of two-layer neural networks (NNs) after one gradient descent step under structured data modeled by Gaussian mixtures. While previous research has extensively analyzed this model under isotropic data assumption, such simplifications overlook the complexities inherent in real-world datasets. Our work addresses this limitation by analyzing two-layer NNs under Gaussian mixture data assumption in the asymptotically proportional limit, where the input dimension, number of hidden neurons, and sample size grow with finite ratios. We characterize the training and generalization errors by leveraging recent advancements in Gaussian universality. Specifically, we prove that a high-order polynomial model performs equivalent to the nonlinear neural networks under certain conditions. The degree of the equivalent model is intricately linked to both the "data spread" and the learning rate employed during one gradient step. Through extensive simulations, we demonstrate the equivalence between the original model and its polynomial counterpart across various regression and classification tasks. Additionally, we explore how different properties of Gaussian mixtures affect learning outcomes. Finally, we illustrate experimental results on Fashion-MNIST classification, indicating that our findings can translate to realistic data.
Authors: Vijay Kumar Sutrakar, Anjana P K
Abstract: An aircraft consists of various cavities including air intake ducts, cockpit, radome, inlet and exhaust of heat exchangers, passage for engine bay/other bay cooling etc. These cavities are prime radar cross-section (RCS) contributors of aircraft. The major such cavity is air intake duct, and it contributes significantly to frontal sector RCS of an aircraft. The RCS reductions of air intake duct is very important to achieve a low RCS (or stealthy) aircraft configuration. In general, radar absorbing materials (RAM) are getting utilized for RCS reduction of air intake duct. It can also be noticed that a large number of fasteners are used for integration of air intake duct with the aircraft structures. The installation of fasteners on RAS may lead to degradation of RCS performance of air intake. However, no such studies are reported in the literature on the impact of rivets on the RCS performance of RAS air intake duct. In this paper, radar absorbing material of thickness 6.25 mm is designed which givens more than -10 dB reflection loss from 4 to 18GHz of frequencies. Next, the effect of rivet installation on these RAS is carried out using three different rivet configurations. The RCS performance of RAS is evaluated for duct of different lengths from 1 to 18GHz of frequencies. In order to see the RCS performance, five different air intake cases are considered The RCS performance with increase in percentage surface area of rivet heads to RAS is reported in detail. At the last, an open-source aircraft CAD model is considered and the RCS performance of RAS air intake with and without rivets is evaluated.
Authors: Xinyu Zhang, Daolang Huang, Samuel Kaski, Julien Martinelli
Abstract: Preferential Bayesian Optimization (PBO) is a sample-efficient method to learn latent user utilities from preferential feedback over a pair of designs. It relies on a statistical surrogate model for the latent function, usually a Gaussian process, and an acquisition strategy to select the next candidate pair to get user feedback on. Due to the non-conjugacy of the associated likelihood, every PBO step requires a significant amount of computations with various approximate inference techniques. This computational overhead is incompatible with the way humans interact with computers, hindering the use of PBO in real-world cases. Building on the recent advances of amortized BO, we propose to circumvent this issue by fully amortizing PBO, meta-learning both the surrogate and the acquisition function. Our method comprises a novel transformer neural process architecture, trained using reinforcement learning and tailored auxiliary losses. On a benchmark composed of synthetic and real-world datasets, our method is several orders of magnitude faster than the usual Gaussian process-based strategies and often outperforms them in accuracy.
Authors: Daiki Nishiyama, Hiroaki Miyoshi, Noriaki Hashimoto, Koichi Ohshima, Hidekata Hontani, Ichiro Takeuchi, Jun Sakuma
Abstract: Malignant lymphoma subtype classification directly impacts treatment strategies and patient outcomes, necessitating classification models that achieve both high accuracy and sufficient explainability. This study proposes a novel explainable Multi-Instance Learning (MIL) framework that identifies subtype-specific Regions of Interest (ROIs) from Whole Slide Images (WSIs) while integrating cell distribution characteristics and image information. Our framework simultaneously addresses three objectives: (1) indicating appropriate ROIs for each subtype, (2) explaining the frequency and spatial distribution of characteristic cell types, and (3) achieving high-accuracy subtyping by leveraging both image and cell-distribution modalities. The proposed method fuses cell graph and image features extracted from each patch in the WSI using a Mixture-of-Experts (MoE) approach and classifies subtypes within an MIL framework. Experiments on a dataset of 1,233 WSIs demonstrate that our approach achieves state-of-the-art accuracy among ten comparative methods and provides region-level and cell-level explanations that align with a pathologist's perspectives.
Authors: Pu Li, Wenhao Zhang, Jianwei Guo, Jinglu Chen, Dong-Ming Yan
Abstract: The integration of deep generative networks into generating Computer-Aided Design (CAD) models has garnered increasing attention over recent years. Traditional methods often rely on discrete sequences of parametric line/curve segments to represent sketches. Differently, we introduce RECAD, a novel framework that generates Raster sketches and 3D Extrusions for CAD models. Representing sketches as raster images offers several advantages over discrete sequences: 1) it breaks the limitations on the types and numbers of lines/curves, providing enhanced geometric representation capabilities; 2) it enables interpolation within a continuous latent space; and 3) it allows for more intuitive user control over the output. Technically, RECAD employs two diffusion networks: the first network generates extrusion boxes conditioned on the number and types of extrusions, while the second network produces sketch images conditioned on these extrusion boxes. By combining these two networks, RECAD effectively generates sketch-and-extrude CAD models, offering a more robust and intuitive approach to CAD model generation. Experimental results indicate that RECAD achieves strong performance in unconditional generation, while also demonstrating effectiveness in conditional generation and output editing.
Authors: Qing Wan, Shilong Deng, Xun Wang
Abstract: Deep neural networks (DNNs) are highly susceptible to adversarial examples--subtle perturbations applied to inputs that are often imperceptible to humans yet lead to incorrect model predictions. In black-box scenarios, however, existing adversarial examples exhibit limited transferability and struggle to effectively compromise multiple unseen DNN models. Previous strategies enhance the cross-model generalization of adversarial examples by introducing versatility into adversarial perturbations, thereby improving transferability. However, further refining perturbation versatility often demands intricate algorithm development and substantial computation consumption. In this work, we propose an input transpose method that requires almost no additional labor and computation costs but can significantly improve the transferability of existing adversarial strategies. Even without adding adversarial perturbations, our method demonstrates considerable effectiveness in cross-model attacks. Our exploration finds that on specific datasets, a mere $1^\circ$ left or right rotation might be sufficient for most adversarial examples to deceive unseen models. Our further analysis suggests that this transferability improvement triggered by rotating only $1^\circ$ may stem from visible pattern shifts in the DNN's low-level feature maps. Moreover, this transferability exhibits optimal angles that, when identified under unrestricted query conditions, could potentially yield even greater performance.
Authors: Minh H. Vu, Lorenzo Tronchin, Tufve Nyholm, Tommy L\"ofstedt
Abstract: Recent years have witnessed a growing academic and industrial interest in deep learning (DL) for medical imaging. To perform well, DL models require very large labeled datasets. However, most medical imaging datasets are small, with a limited number of annotated samples. The reason they are small is usually because delineating medical images is time-consuming and demanding for oncologists. There are various techniques that can be used to augment a dataset, for example, to apply affine transformations or elastic transformations to available images, or to add synthetic images generated by a Generative Adversarial Network (GAN). In this work, we have developed a novel conditional variant of a current GAN method, the StyleGAN2, to generate multi-modal high-resolution medical images with the purpose to augment small medical imaging datasets with these synthetic images. We use the synthetic and real images from six datasets to train models for the downstream task of semantic segmentation. The quality of the generated medical images and the effect of this augmentation on the segmentation performance were evaluated afterward. Finally, the results indicate that the downstream segmentation models did not benefit from the generated images. Further work and analyses are required to establish how this augmentation affects the segmentation performance.
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: Onur Boyar, Indra Priyadarsini, Seiji Takeda, Lisa Hamada
Abstract: Discovering materials with desirable properties in an efficient way remains a significant problem in materials science. Many studies have tackled this problem by using different sets of information available about the materials. Among them, multimodal approaches have been found to be promising because of their ability to combine different sources of information. However, fusion algorithms to date remain simple, lacking a mechanism to provide a rich representation of multiple modalities. This paper presents LLM-Fusion, a novel multimodal fusion model that leverages large language models (LLMs) to integrate diverse representations, such as SMILES, SELFIES, text descriptions, and molecular fingerprints, for accurate property prediction. Our approach introduces a flexible LLM-based architecture that supports multimodal input processing and enables material property prediction with higher accuracy than traditional methods. We validate our model on two datasets across five prediction tasks and demonstrate its effectiveness compared to unimodal and naive concatenation baselines.
Authors: Yasamin Jalalian, Juan Felipe Osorio Ramirez, Alexander Hsu, Bamdad Hosseini, Houman Owhadi
Abstract: We introduce a novel kernel-based framework for learning differential equations and their solution maps that is efficient in data requirements, in terms of solution examples and amount of measurements from each example, and computational cost, in terms of training procedures. Our approach is mathematically interpretable and backed by rigorous theoretical guarantees in the form of quantitative worst-case error bounds for the learned equation. Numerical benchmarks demonstrate significant improvements in computational complexity and robustness while achieving one to two orders of magnitude improvements in terms of accuracy compared to state-of-the-art algorithms.
Authors: Elham Ghelichkhan, Tolga Tasdizen
Abstract: Chest diseases rank among the most prevalent and dangerous global health issues. Object detection and phrase grounding deep learning models interpret complex radiology data to assist healthcare professionals in diagnosis. Object detection locates abnormalities for classes, while phrase grounding locates abnormalities for textual descriptions. This paper investigates how text enhances abnormality localization in chest X-rays by comparing the performance and explainability of these two tasks. To establish an explainability baseline, we proposed an automatic pipeline to generate image regions for report sentences using radiologists' eye-tracking data. The better performance - mIoU = 0.36 vs. 0.20 - and explainability - Containment ratio 0.48 vs. 0.26 - of the phrase grounding model infers the effectiveness of text in enhancing chest X-ray abnormality localization.
Authors: Anav Sood
Abstract: Upon observing $n$-dimensional multivariate Gaussian data, when can we infer that the largest $K$ observations came from the largest $K$ means? When $K=1$ and the covariance is isotropic, \cite{Gutmann} argue that this inference is justified when the two-sided difference-of-means test comparing the largest and second largest observation rejects. Leveraging tools from selective inference, we provide a generalization of their procedure that applies for both any $K$ and any covariance structure. We show that our procedure draws the desired inference whenever the two-sided difference-of-means test comparing the pair of observations inside and outside the top $K$ with the smallest standardized difference rejects, and sometimes even when this test fails to reject. Using this insight, we argue that our procedure renders existing simultaneous inference approaches inadmissible when $n > 2$. When the observations are independent (with possibly unequal variances) or equicorrelated, our procedure corresponds exactly to running the two-sided difference-of-means test comparing the pair of observations inside and outside the top $K$ with the smallest standardized difference.
Authors: Yu Fu, Michael Stanley Smith, Anastasios Panagiotelis
Abstract: Variational inference (VI) is a popular method to estimate statistical and econometric models. The key to VI is the selection of a tractable density to approximate the Bayesian posterior. For large and complex models a common choice is to assume independence between multivariate blocks in a partition of the parameter space. While this simplifies the problem it can reduce accuracy. This paper proposes using vector copulas to capture dependence between the blocks parsimoniously. Tailored multivariate marginals are constructed using learnable cyclically monotone transformations. We call the resulting joint distribution a ``dependent block posterior'' approximation. Vector copula models are suggested that make tractable and flexible variational approximations. They allow for differing marginals, numbers of blocks, block sizes and forms of between block dependence. They also allow for solution of the variational optimization using fast and efficient stochastic gradient methods. The efficacy and versatility of the approach is demonstrated using four different statistical models and 16 datasets which have posteriors that are challenging to approximate. In all cases, our method produces more accurate posterior approximations than benchmark VI methods that either assume block independence or factor-based dependence, at limited additional computational cost.
Authors: Ziheng Guo, James Greene, Ming Zhong
Abstract: Stochastic differential equations (SDEs) are a ubiquitous modeling framework that finds applications in physics, biology, engineering, social science, and finance. Due to the availability of large-scale data sets, there is growing interest in learning mechanistic models from observations with stochastic noise. In this work, we present a nonparametric framework to learn both the drift and diffusion terms in systems of SDEs where the stochastic noise is singular. Specifically, inspired by second-order equations from classical physics, we consider systems which possess structured noise, i.e. noise with a singular covariance matrix. We provide an algorithm for constructing estimators given trajectory data and demonstrate the effectiveness of our methods via a number of examples from physics and biology. As the developed framework is most naturally applicable to systems possessing a high degree of dimensionality reduction (i.e. symmetry), we also apply it to the high dimensional Cucker-Smale flocking model studied in collective dynamics and show that it is able to accurately infer the low dimensional interaction kernel from particle data.
Authors: Mykola Kozlenko, Volodymyr Sendetskyi, Oleksiy Simkiv, Nazar Savchenko, Andy Bosyi
Abstract: Object recognition and detection are well-studied problems with a developed set of almost standard solutions. Identity documents recognition, classification, detection, and localization are the tasks required in a number of applications, particularly, in physical access control security systems at critical infrastructure premises. In this paper, we propose the new original architecture of a model based on an artificial convolutional neural network and semantic segmentation approach for the recognition and detection of identity documents in images. The challenge with the processing of such images is the limited computational performance and the limited amount of memory when such an application is running on industrial oneboard microcomputer hardware. The aim of this research is to prove the feasibility of the proposed technique and to obtain quality metrics. The methodology of the research is to evaluate the deep learning detection model trained on the mobile identity document video dataset. The dataset contains five hundred video clips for fifty different identity document types. The numerical results from simulations are used to evaluate the quality metrics. We present the results as accuracy versus threshold of the intersection over union value. The paper reports an accuracy above 0.75 for the intersection over union (IoU) threshold value of 0.8. Besides, we assessed the size of the model and proved the feasibility of running the model on an industrial one-board microcomputer or smartphone hardware.
Authors: Jon Donnelly, Zhicheng Guo, Alina Jade Barnett, Hayden McTavish, Chaofan Chen, Cynthia Rudin
Abstract: Interpretability is critical for machine learning models in high-stakes settings because it allows users to verify the model's reasoning. In computer vision, prototypical part models (ProtoPNets) have become the dominant model type to meet this need. Users can easily identify flaws in ProtoPNets, but fixing problems in a ProtoPNet requires slow, difficult retraining that is not guaranteed to resolve the issue. This problem is called the "interaction bottleneck." We solve the interaction bottleneck for ProtoPNets by simultaneously finding many equally good ProtoPNets (i.e., a draw from a "Rashomon set"). We show that our framework - called Proto-RSet - quickly produces many accurate, diverse ProtoPNets, allowing users to correct problems in real time while maintaining performance guarantees with respect to the training set. We demonstrate the utility of this method in two settings: 1) removing synthetic bias introduced to a bird identification model and 2) debugging a skin cancer identification model. This tool empowers non-machine-learning experts, such as clinicians or domain experts, to quickly refine and correct machine learning models without repeated retraining by machine learning experts.
Authors: Kaiwen Zheng, Yongxin Chen, Huayu Chen, Guande He, Ming-Yu Liu, Jun Zhu, Qinsheng Zhang
Abstract: While likelihood-based generative models, particularly diffusion and autoregressive models, have achieved remarkable fidelity in visual generation, the maximum likelihood estimation (MLE) objective inherently suffers from a mode-covering tendency that limits the generation quality under limited model capacity. In this work, we propose Direct Discriminative Optimization (DDO) as a unified framework that bridges likelihood-based generative training and the GAN objective to bypass this fundamental constraint. Our key insight is to parameterize a discriminator implicitly using the likelihood ratio between a learnable target model and a fixed reference model, drawing parallels with the philosophy of Direct Preference Optimization (DPO). Unlike GANs, this parameterization eliminates the need for joint training of generator and discriminator networks, allowing for direct, efficient, and effective finetuning of a well-trained model to its full potential beyond the limits of MLE. DDO can be performed iteratively in a self-play manner for progressive model refinement, with each round requiring less than 1% of pretraining epochs. Our experiments demonstrate the effectiveness of DDO by significantly advancing the previous SOTA diffusion model EDM, reducing FID scores from 1.79/1.58 to new records of 1.30/0.97 on CIFAR-10/ImageNet-64 datasets, and by consistently improving both guidance-free and CFG-enhanced FIDs of visual autoregressive models on ImageNet 256$\times$256.
Authors: Jianzhe Xue, Dongcheng Yuan, Zhanxi Ma, Tiankai Jiang, Yu Sun, Haibo Zhou, Xuemin Shen
Abstract: Channel prediction is crucial for high-mobility vehicular networks, as it enables the anticipation of future channel conditions and the proactive adjustment of communication strategies. However, achieving accurate vehicular channel prediction is challenging due to significant Doppler effects and rapid channel variations resulting from high-speed vehicle movement and complex propagation environments. In this paper, we propose a novel delay-Doppler (DD) domain channel prediction framework tailored for high-mobility vehicular networks. By transforming the channel representation into the DD domain, we obtain an intuitive, sparse, and stable depiction that closely aligns with the underlying physical propagation processes, effectively reducing the complex vehicular channel to a set of time-series parameters with enhanced predictability. Furthermore, we leverage the large artificial intelligence (AI) model to predict these DD-domain time-series parameters, capitalizing on their advanced ability to model temporal correlations. The zero-shot capability of the pre-trained large AI model facilitates accurate channel predictions without requiring task-specific training, while subsequent fine-tuning on specific vehicular channel data further improves prediction accuracy. Extensive simulation results demonstrate the effectiveness of our DD-domain channel prediction framework and the superior accuracy of the large AI model in predicting time-series channel parameters, thereby highlighting the potential of our approach for robust vehicular communication systems.
Authors: Zahra Zanjani Foumani, Ramin Bostanabad
Abstract: Bayesian optimization (BO) is increasingly employed in critical applications to find the optimal design with minimal cost. While BO is known for its sample efficiency, relying solely on costly high-fidelity data can still result in high costs. This is especially the case in constrained search spaces where BO must not only optimize but also ensure feasibility. A related issue in the BO literature is the lack of a systematic stopping criterion. To solve these challenges, we develop a constrained cost-aware multi-fidelity BO (CMFBO) framework whose goal is to minimize overall sampling costs by utilizing inexpensive low-fidelity sources while ensuring feasibility. In our case, the constraints can change across the data sources and may be even black-box functions. We also introduce a systematic stopping criterion that addresses the long-lasting issue associated with BO's convergence assessment. Our framework is publicly available on GitHub through the GP+ Python package and herein we validate it's efficacy on multiple benchmark problems.
Authors: Junyi Li, Yongqiang Chen, Chenxi Liu, Qianyi Cai, Tongliang Liu, Bo Han, Kun Zhang, Hui Xiong
Abstract: Designing proper experiments and selecting optimal intervention targets is a longstanding problem in scientific or causal discovery. Identifying the underlying causal structure from observational data alone is inherently difficult.Obtaining interventional data, on the other hand, is crucial to causal discovery, yet it is usually expensive and time-consuming to gather sufficient interventional data to facilitate causal discovery.Previous approaches commonly utilize uncertainty or gradient signals to determine the intervention targets. However, numerical-based approaches may yield suboptimal results due to the inaccurate estimation of the guiding signals at the beginning when with limited interventional data. In this work, we investigate a different approach, whether we can leverage Large Language Models (LLMs) to assist with the intervention targeting in causal discovery by making use of the rich world knowledge about the experimental design in LLMs.Specifically, we present \oursfull (\ours) -- a robust framework that effectively incorporates LLMs to augment existing numerical approaches for the intervention targeting in causal discovery. Across $4$ realistic benchmark scales, \ours demonstrates significant improvements and robustness over existing methods and even surpasses humans, which demonstrates the usefulness of LLMs in assisting with experimental design for scientific discovery.
Authors: Rin Ashizawa, Yoichi Hirose, Nozomu Yoshinari, Kento Uchida, Shinichi Shirakawa
Abstract: Prompt optimization aims to search for effective prompts that enhance the performance of large language models (LLMs). Although existing prompt optimization methods have discovered effective prompts, they often differ from sophisticated prompts carefully designed by human experts. Prompt design strategies, representing best practices for improving prompt performance, can be key to improving prompt optimization. Recently, a method termed the Autonomous Prompt Engineering Toolbox (APET) has incorporated various prompt design strategies into the prompt optimization process. In APET, the LLM is needed to implicitly select and apply the appropriate strategies because prompt design strategies can have negative effects. This implicit selection may be suboptimal due to the limited optimization capabilities of LLMs. This paper introduces Optimizing Prompts with sTrategy Selection (OPTS), which implements explicit selection mechanisms for prompt design. We propose three mechanisms, including a Thompson sampling-based approach, and integrate them into EvoPrompt, a well-known prompt optimizer. Experiments optimizing prompts for two LLMs, Llama-3-8B-Instruct and GPT-4o mini, were conducted using BIG-Bench Hard. Our results show that the selection of prompt design strategies improves the performance of EvoPrompt, and the Thompson sampling-based mechanism achieves the best overall results. Our experimental code is provided at https://github.com/shiralab/OPTS .
Authors: Deyu Bo, Songhua Liu, Xinchao Wang
Abstract: Dataset distillation (DD) has emerged as a promising approach to compress datasets and speed up model training. However, the underlying connections among various DD methods remain largely unexplored. In this paper, we introduce UniDD, a spectral filtering framework that unifies diverse DD objectives. UniDD interprets each DD objective as a specific filter function that affects the eigenvalues of the feature-feature correlation (FFC) matrix and modulates the frequency components of the feature-label correlation (FLC) matrix. In this way, UniDD reveals that the essence of DD fundamentally lies in matching frequency-specific features. Moreover, according to the filter behaviors, we classify existing methods into low-frequency matching and high-frequency matching, encoding global texture and local details, respectively. However, existing methods rely on fixed filter functions throughout distillation, which cannot capture the low- and high-frequency information simultaneously. To address this limitation, we further propose Curriculum Frequency Matching (CFM), which gradually adjusts the filter parameter to cover both low- and high-frequency information of the FFC and FLC matrices. Extensive experiments on small-scale datasets, such as CIFAR-10/100, and large-scale datasets, including ImageNet-1K, demonstrate the superior performance of CFM over existing baselines and validate the practicality of UniDD.
Authors: Sijin Sun, Ming Deng, Xinrui Yu, Liangbin Zhao
Abstract: Incorrect boundary division, complex semantic representation, and differences in pronunciation and meaning often lead to errors in Chinese Named Entity Recognition(CNER). To address these issues, this paper proposes HREB-CRF framework: Hierarchical Reduced-bias EMA with CRF. The proposed method amplifies word boundaries and pools long text gradients through exponentially fixed-bias weighted average of local and global hierarchical attention. Experimental results on the MSRA, Resume, and Weibo datasets show excellent in F1, outperforming the baseline model by 1.1\%, 1.6\%, and 9.8\%. The significant improvement in F1 shows evidences of strong effectiveness and robustness of approach in CNER tasks.
Authors: Sahar Sinene Mehdoui, Abdelhamid Bouzid, Daniel Sierra-Sosa, Adel Elmaghraby
Abstract: Dementia, a progressive neurodegenerative disorder, affects memory, reasoning, and daily functioning, creating challenges for individuals and healthcare systems. Early detection is crucial for timely interventions that may slow disease progression. Large pre-trained models (LPMs) for text and audio, such as Generative Pre-trained Transformer (GPT), Bidirectional Encoder Representations from Transformers (BERT), and Contrastive Language-Audio Pretraining (CLAP), have shown promise in identifying cognitive impairments. However, existing studies generally rely heavily on expert-annotated datasets and unimodal approaches, limiting robustness and scalability. This study proposes a context-based multimodal method, integrating both text and audio data using the best-performing LPMs in each modality. By incorporating contextual embeddings, our method improves dementia detection performance. Additionally, motivated by the effectiveness of contextual embeddings, we further experimented with a context-based In-Context Learning (ICL) as a complementary technique. Results show that GPT-based embeddings, particularly when fused with CLAP audio features, achieve an F1-score of $83.33\%$, surpassing state-of-the-art dementia detection models. Furthermore, raw text data outperforms expert-annotated datasets, demonstrating that LPMs can extract meaningful linguistic and acoustic patterns without extensive manual labeling. These findings highlight the potential for scalable, non-invasive diagnostic tools that reduce reliance on costly annotations while maintaining high accuracy. By integrating multimodal learning with contextual embeddings, this work lays the foundation for future advancements in personalized dementia detection and cognitive health research.
Authors: Shuyi Jia, Shitij Govil, Manav Ramprasad, Victor Fung
Abstract: In recent years, pre-trained graph neural networks (GNNs) have been developed as general models which can be effectively fine-tuned for various potential downstream tasks in materials science, and have shown significant improvements in accuracy and data efficiency. The most widely used pre-training methods currently involve either supervised training to fit a general force field or self-supervised training by denoising atomic structures equilibrium. Both methods require datasets generated from quantum mechanical calculations, which quickly become intractable when scaling to larger datasets. Here we propose a novel pre-training objective which instead uses cheaply-computed structural fingerprints as targets while maintaining comparable performance across a range of different structural descriptors. Our experiments show this approach can act as a general strategy for pre-training GNNs with application towards large scale foundational models for atomistic data.
Authors: Sijin Sun, Ming Deng, Xingrui Yu, Xinyu Xi, Liangbin Zhao
Abstract: Metal defect detection is critical in industrial quality assurance, yet existing methods struggle with grayscale variations and complex defect states, limiting its robustness. To address these challenges, this paper proposes a Self-Adaptive Gamma Context-Aware SSM-based model(GCM-DET). This advanced detection framework integrating a Dynamic Gamma Correction (GC) module to enhance grayscale representation and optimize feature extraction for precise defect reconstruction. A State-Space Search Management (SSM) architecture captures robust multi-scale features, effectively handling defects of varying shapes and scales. Focal Loss is employed to mitigate class imbalance and refine detection accuracy. Additionally, the CD5-DET dataset is introduced, specifically designed for port container maintenance, featuring significant grayscale variations and intricate defect patterns. Experimental results demonstrate that the proposed model achieves substantial improvements, with mAP@0.5 gains of 27.6\%, 6.6\%, and 2.6\% on the CD5-DET, NEU-DET, and GC10-DET datasets.
Authors: Jensen Gao, Suneel Belkhale, Sudeep Dasari, Ashwin Balakrishna, Dhruv Shah, Dorsa Sadigh
Abstract: Machine learning for robotics promises to unlock generalization to novel tasks and environments. Guided by this promise, many recent works have focused on scaling up robot data collection and developing larger, more expressive policies to achieve this. But how do we measure progress towards this goal of policy generalization in practice? Evaluating and quantifying generalization is the Wild West of modern robotics, with each work proposing and measuring different types of generalization in their own, often difficult to reproduce, settings. In this work, our goal is (1) to outline the forms of generalization we believe are important in robot manipulation in a comprehensive and fine-grained manner, and (2) to provide reproducible guidelines for measuring these notions of generalization. We first propose STAR-Gen, a taxonomy of generalization for robot manipulation structured around visual, semantic, and behavioral generalization. We discuss how our taxonomy encompasses most prior notions of generalization in robotics. Next, we instantiate STAR-Gen with a concrete real-world benchmark based on the widely-used Bridge V2 dataset. We evaluate a variety of state-of-the-art models on this benchmark to demonstrate the utility of our taxonomy in practice. Our taxonomy of generalization can yield many interesting insights into existing models: for example, we observe that current vision-language-action models struggle with various types of semantic generalization, despite the promise of pre-training on internet-scale language datasets. We believe STAR-Gen and our guidelines can improve the dissemination and evaluation of progress towards generalization in robotics, which we hope will guide model design and future data collection efforts. We provide videos and demos at our website stargen-taxonomy.github.io.
Authors: Nam Huynh, Beiyu Lin
Abstract: Large Language Models (LLMs) have demonstrated their remarkable capabilities in numerous fields. This survey focuses on how LLMs empower users, regardless of their technical background, to use human languages to automatically generate executable code. We begin with understanding LLMs' limitations and challenges in automated code generation. Subsequently, we review various fine-tuning techniques designed to enhance both the performance and adaptability of LLMs in code generation tasks. We then review the existing metrics and benchmarks for evaluations to assess model performance based on fine-tuning techniques. Finally, we explore the applications of LLMs (e.g. CodeLlama, GitHub Copilot, ToolGen) in code generation tasks to illustrate their roles and functionalities. This survey provides a comprehensive overview of LLMs for code generation, helps researchers in diverse fields better understand the current state-of-the-art technologies, and offers the potential of effectively leveraging LLMs for code generation tasks.
Authors: S. Chen, D. Ma, M. Raviselvan, S. Sundaramoorthy, K. Popuri, M. J. Ju, M. V. Sarunic, D. Ratra, M. F. Beg
Abstract: This study presents an AI-driven pipeline for automated retinal segmentation and thickness analysis in diabetic retinopathy (DR) using SD-OCT imaging. A deep neural network was trained to segment ten retinal layers, intra-retinal fluid, and hyperreflective foci (HRF), with performance evaluated across multiple architectures. SwinUNETR achieved the highest segmentation accuracy, while VM-Unet excelled in specific layers. Analysis revealed distinct thickness variations between NPDR and PDR, with correlations between layer thickness and visual acuity. The proposed method enhances DR assessment by reducing manual annotation effort and providing clinically relevant thickness maps for disease monitoring and treatment planning.
Authors: Md Abrar Jahin, Soudeep Shahriar, M. F. Mridha, Nilanjan Dey
Abstract: Soybean leaf disease detection is critical for agricultural productivity but faces challenges due to visually similar symptoms and limited interpretability in conventional methods. While Convolutional Neural Networks (CNNs) excel in spatial feature extraction, they often neglect inter-image relational dependencies, leading to misclassifications. This paper proposes an interpretable hybrid Sequential CNN-Graph Neural Network (GNN) framework that synergizes MobileNetV2 for localized feature extraction and GraphSAGE for relational modeling. The framework constructs a graph where nodes represent leaf images, with edges defined by cosine similarity-based adjacency matrices and adaptive neighborhood sampling. This design captures fine-grained lesion features and global symptom patterns, addressing inter-class similarity challenges. Cross-modal interpretability is achieved via Grad-CAM and Eigen-CAM visualizations, generating heatmaps to highlight disease-influential regions. Evaluated on a dataset of ten soybean leaf diseases, the model achieves $97.16\%$ accuracy, surpassing standalone CNNs ($\le95.04\%$) and traditional machine learning models ($\le77.05\%$). Ablation studies validate the sequential architecture's superiority over parallel or single-model configurations. With only 2.3 million parameters, the lightweight MobileNetV2-GraphSAGE combination ensures computational efficiency, enabling real-time deployment in resource-constrained environments. The proposed approach bridges the gap between accurate classification and practical applicability, offering a robust, interpretable tool for agricultural diagnostics while advancing CNN-GNN integration in plant pathology research.
Authors: Jan-Hendrik Ewers, Joe Gibbs, David Anderson
Abstract: Management of sensing resources is a non-trivial problem for future military air assets with future systems deploying heterogeneous sensors to generate information of the battlespace. Machine learning techniques including deep reinforcement learning (DRL) have been identified as promising approaches, but require high-fidelity training environments and feature extractors to generate information for the agent. This paper presents a deep reinforcement learning training approach, utilising the Stone Soup tracking framework as a feature extractor to train an agent for a sensor management task. A general framework for embedding Stone Soup tracker components within a Gymnasium environment is presented, enabling fast and configurable tracker deployments for RL training using Stable Baselines3. The approach is demonstrated in a sensor management task where an agent is trained to search and track a region of airspace utilising track lists generated from Stone Soup trackers. A sample implementation using three neural network architectures in a search-and-track scenario demonstrates the approach and shows that RL agents can outperform simple sensor search and track policies when trained within the Gymnasium and Stone Soup environment.
Authors: Tingchen Fu, Fazl Barez
Abstract: Insensitivity to semantically-preserving variations of prompts (paraphrases) is crucial for reliable behavior and real-world deployment of large language models. However, language models exhibit significant performance degradation when faced with semantically equivalent but differently phrased prompts, and existing solutions either depend on trial-and-error prompt engineering or require computationally expensive inference-time algorithms. In this study, built on the key insight that worst-case prompts exhibit a drift in embedding space, we present Latent Adversarial Paraphrasing (LAP), a dual-loop adversarial framework: the inner loop trains a learnable perturbation to serve as a "latent continuous paraphrase" while preserving semantics through Lagrangian regulation, and the outer loop optimizes the language model parameters on these perturbations. We conduct extensive experiments to demonstrate the effectiveness of LAP across multiple LLM architectures on the RobustAlpaca benchmark with a 0.5%-4% absolution improvement on worst-case win-rate compared with vanilla supervised fine-tuning.
Authors: O. Duranthon, L. Zdeborov\'a
Abstract: Graph neural networks (GNNs) are designed to process data associated with graphs. They are finding an increasing range of applications; however, as with other modern machine learning techniques, their theoretical understanding is limited. GNNs can encounter difficulties in gathering information from nodes that are far apart by iterated aggregation steps. This situation is partly caused by so-called oversmoothing; and overcoming it is one of the practically motivated challenges. We consider the situation where information is aggregated by multiple steps of convolution, leading to graph convolutional networks (GCNs). We analyze the generalization performance of a basic GCN, trained for node classification on data generated by the contextual stochastic block model. We predict its asymptotic performance by deriving the free energy of the problem, using the replica method, in the high-dimensional limit. Calling depth the number of convolutional steps, we show the importance of going to large depth to approach the Bayes-optimality. We detail how the architecture of the GCN has to scale with the depth to avoid oversmoothing. The resulting large depth limit can be close to the Bayes-optimality and leads to a continuous GCN. Technically, we tackle this continuous limit via an approach that resembles dynamical mean-field theory (DMFT) with constraints at the initial and final times. An expansion around large regularization allows us to solve the corresponding equations for the performance of the deep GCN. This promising tool may contribute to the analysis of further deep neural networks.
Authors: Yanghai Zhang, Changyi Liu, Keting Fu, Wenbin Zhou, Qingdu Li, Jianwei Zhang
Abstract: This paper proposes FABG (Facial Affective Behavior Generation), an end-to-end imitation learning system for human-robot interaction, designed to generate natural and fluid facial affective behaviors. In interaction, effectively obtaining high-quality demonstrations remains a challenge. In this work, we develop an immersive virtual reality (VR) demonstration system that allows operators to perceive stereoscopic environments. This system ensures "the operator's visual perception matches the robot's sensory input" and "the operator's actions directly determine the robot's behaviors" - as if the operator replaces the robot in human interaction engagements. We propose a prediction-driven latency compensation strategy to reduce robotic reaction delays and enhance interaction fluency. FABG naturally acquires human interactive behaviors and subconscious motions driven by intuition, eliminating manual behavior scripting. We deploy FABG on a real-world 25-degree-of-freedom (DoF) humanoid robot, validating its effectiveness through four fundamental interaction tasks: expression response, dynamic gaze, foveated attention, and gesture recognition, supported by data collection and policy training. Project website: https://cybergenies.github.io
Authors: Joel Niklaus, Jakob Merane, Luka Nenadic, Sina Ahmadi, Yingqiang Gao, Cyrill A. H. Chevalley, Claude Humbel, Christophe G\"osken, Lorenzo Tanzi, Thomas L\"uthi, Stefan Palombo, Spencer Poff, Boling Yang, Nan Wu, Matthew Guillod, Robin Mami\'e, Daniel Brunner, Julio Pereyra, Niko Grupen
Abstract: In Switzerland legal translation is uniquely important due to the country's four official languages and requirements for multilingual legal documentation. However, this process traditionally relies on professionals who must be both legal experts and skilled translators -- creating bottlenecks and impacting effective access to justice. To address this challenge, we introduce SwiLTra-Bench, a comprehensive multilingual benchmark of over 180K aligned Swiss legal translation pairs comprising laws, headnotes, and press releases across all Swiss languages along with English, designed to evaluate LLM-based translation systems. Our systematic evaluation reveals that frontier models achieve superior translation performance across all document types, while specialized translation systems excel specifically in laws but under-perform in headnotes. Through rigorous testing and human expert validation, we demonstrate that while fine-tuning open SLMs significantly improves their translation quality, they still lag behind the best zero-shot prompted frontier models such as Claude-3.5-Sonnet. Additionally, we present SwiLTra-Judge, a specialized LLM evaluation system that aligns best with human expert assessments.
Authors: Tim Schwabe, Louisa Siebel, Patrik Valach, Maribel Acosta
Abstract: Question answering (QA) requires accurately aligning user questions with structured queries, a process often limited by the scarcity of high-quality query-natural language (Q-NL) pairs. To overcome this, we present Q-NL Verifier, an approach to generating high-quality synthetic pairs of queries and NL translations. Our approach relies on large language models (LLMs) to generate semantically precise natural language paraphrases of structured queries. Building on these synthetic Q-NL pairs, we introduce a learned verifier component that automatically determines whether a generated paraphrase is semantically equivalent to the original query. Our experiments with the well-known LC-QuAD 2.0 benchmark show that Q-NL Verifier generalizes well to paraphrases from other models and even human-authored translations. Our approach strongly aligns with human judgments across varying query complexities and outperforms existing NLP metrics in assessing semantic correctness. We also integrate the verifier into QA pipelines, showing that verifier-filtered synthetic data has significantly higher quality in terms of translation correctness and enhances NL to Q translation accuracy. Lastly, we release an updated version of the LC-QuAD 2.0 benchmark containing our synthetic Q-NL pairs and verifier scores, offering a new resource for robust and scalable QA.
Authors: Leonardo Nizzoli, Marco Avvenuti, Maurizio Tesconi, Stefano Cresci
Abstract: Online social networks convey rich information about geospatial facets of reality. However in most cases, geographic information is not explicit and structured, thus preventing its exploitation in real-time applications. We address this limitation by introducing a novel geoparsing and geotagging technique called Geo-Semantic-Parsing (GSP). GSP identifies location references in free text and extracts the corresponding geographic coordinates. To reach this goal, we employ a semantic annotator to identify relevant portions of the input text and to link them to the corresponding entity in a knowledge graph. Then, we devise and experiment with several efficient strategies for traversing the knowledge graph, thus expanding the available set of information for the geoparsing task. Finally, we exploit all available information for learning a regression model that selects the best entity with which to geotag the input text. We evaluate GSP on a well-known reference dataset including almost 10k event-related tweets, achieving $F1=0.66$. We extensively compare our results with those of 2 baselines and 3 state-of-the-art geoparsing techniques, achieving the best performance. On the same dataset, competitors obtain $F1 \leq 0.55$. We conclude by providing in-depth analyses of our results, showing that the overall superior performance of GSP is mainly due to a large improvement in recall, with respect to existing techniques.
Authors: Thibault Gauthier, Josef Urban
Abstract: We develop a self-learning approach for conjecturing of induction predicates on a dataset of 16197 problems derived from the OEIS. These problems are hard for today's SMT and ATP systems because they require a combination of inductive and arithmetical reasoning. Starting from scratch, our approach consists of a feedback loop that iterates between (i) training a neural translator to learn the correspondence between the problems solved so far and the induction predicates useful for them, (ii) using the trained neural system to generate many new induction predicates for the problems, (iii) fast runs of the z3 prover attempting to prove the problems using the generated predicates, (iv) using heuristics such as predicate size and solution speed on the proved problems to choose the best predicates for the next iteration of training. The algorithm discovers on its own many interesting induction predicates, ultimately solving 5565 problems, compared to 2265 problems solved by CVC5, Vampire or Z3 in 60 seconds.
Authors: Dawid Mazur, Tomasz Rybotycki, Piotr Gawron
Abstract: Training of machine learning models consumes large amounts of energy. Since the energy consumption becomes a major problem in the development and implementation of artificial intelligence systems there exists a need to investigate the ways to reduce use of the resources by these systems. In this work we study how application of quantum annealers could lead to reduction of energy cost in training models aiming at pixel-level segmentation of hyperspectral images. Following the results of QBM4EO team, we propose a classical machine learning model, partially trained using quantum annealer, for hyperspectral image segmentation. We show that the model trained using quantum annealer is better or at least comparable with models trained using alternative algorithms, according to the preselected, common metrics. While direct energy use comparison does not make sense at the current stage of quantum computing technology development, we believe that our work proves that quantum annealing should be considered as a tool for training at least some machine learning models.
Authors: Piotr Kalaczy\'nski (on behalf of the KM3NeT Collaboration)
Abstract: The KM3NeT Collaboration is installing the ARCA and ORCA neutrino detectors at the bottom of the Mediterranean Sea. The focus of ARCA is neutrino astronomy, while ORCA is optimised for neutrino oscillation studies. Both detectors are already operational in their intermediate states and collect valuable data, including the measurements of the muons produced by cosmic ray interactions in the atmosphere. This work explores the potential of machine learning models for the reconstruction of muon bundles, which are multi-muon events. For this, data collected with intermediate detector configurations of ARCA and ORCA was used in addition to simulated data from the envisaged final configurations of those detectors. Prediction of the total number of muons in a bundle as well as their total energy and even the energy of the primary cosmic ray is presented.
Authors: Dan Garber
Abstract: We consider the problem of minimizing a smooth and convex function over the $n$-dimensional spectrahedron -- the set of real symmetric $n\times n$ positive semidefinite matrices with unit trace, which underlies numerous applications in statistics, machine learning and additional domains. Standard first-order methods often require high-rank matrix computations which are prohibitive when the dimension $n$ is large. The well-known Frank-Wolfe method on the other hand, only requires efficient rank-one matrix computations, however suffers from worst-case slow convergence, even under conditions that enable linear convergence rates for standard methods. In this work we present the first Frank-Wolfe-based algorithm that only applies efficient rank-one matrix computations and, assuming quadratic growth and strict complementarity conditions, is guaranteed, after a finite number of iterations, to converges linearly, in expectation, and independently of the ambient dimension.
Authors: Xu Wan, Chao Yang, Cheng Yang, Jie Song, Mingyang Sun
Abstract: Although multi-agent reinforcement learning (MARL) has shown its success across diverse domains, extending its application to large-scale real-world systems still faces significant challenges. Primarily, the high complexity of real-world environments exacerbates the credit assignment problem, substantially reducing training efficiency. Moreover, the variability of agent populations in large-scale scenarios necessitates scalable decision-making mechanisms. To address these challenges, we propose a novel framework: Sequential rollout with Sequential value estimation (SrSv). This framework aims to capture agent interdependence and provide a scalable solution for cooperative MARL. Specifically, SrSv leverages the autoregressive property of the Transformer model to handle varying populations through sequential action rollout. Furthermore, to capture the interdependence of policy distributions and value functions among multiple agents, we introduce an innovative sequential value estimation methodology and integrates the value approximation into an attention-based sequential model. We evaluate SrSv on three benchmarks: Multi-Agent MuJoCo, StarCraft Multi-Agent Challenge, and DubinsCars. Experimental results demonstrate that SrSv significantly outperforms baseline methods in terms of training efficiency without compromising convergence performance. Moreover, when implemented in a large-scale DubinsCar system with 1,024 agents, our framework surpasses existing benchmarks, highlighting the excellent scalability of SrSv.
Authors: Hanyu Wang, Emmanuel K. Tsinda, Anthony J. Dunn, Francis Chikweto, Alain B. Zemkoho
Abstract: Motivation: PCR is more economical and quicker than Next Generation Sequencing for detecting target organisms, with primer design being a critical step. In epidemiology with rapidly mutating viruses, designing effective primers is challenging. Traditional methods require substantial manual intervention and struggle to ensure effective primer design across different strains. For organisms with large, similar genomes like Escherichia coli and Shigella flexneri, differentiating between species is also difficult but crucial. Results: We developed Primer C-VAE, a model based on a Variational Auto-Encoder framework with Convolutional Neural Networks to identify variants and generate specific primers. Using SARS-CoV-2, our model classified variants (alpha, beta, gamma, delta, omicron) with 98% accuracy and generated variant-specific primers. These primers appeared with >95% frequency in target variants and <5% in others, showing good performance in in-silico PCR tests. For Alpha, Delta, and Omicron, our primer pairs produced fragments <200 bp, suitable for qPCR detection. The model also generated effective primers for organisms with longer gene sequences like E. coli and S. flexneri. Conclusion: Primer C-VAE is an interpretable deep learning approach for developing specific primer pairs for target organisms. This flexible, semi-automated and reliable tool works regardless of sequence completeness and length, allowing for qPCR applications and can be applied to organisms with large and highly similar genomes.
Authors: A. Tajja, A. Aghabiglou, E. Tolley, J-P. Kneib, J-P. Thiran, Y. Wiaux
Abstract: Recently, the R2D2 paradigm, standing for ''Residual-to-Residual DNN series for high-Dynamic-range imaging'', was introduced for image formation in Radio Interferometry (RI) as a learned version of the traditional algorithm CLEAN. The first incarnations of R2D2 are limited to planar imaging on small fields of view, failing to meet the spherical-imaging requirement of modern telescopes observing wide fields. To address this limitation, we propose the spherical-imaging extension S-R2D2. Firstly, as R2D2, S-R2D2 encapsulates its minor cycles in existing 2D-Euclidean deep neural network (DNN) architectures, but adapts its iterative scheme to incorporate the wide-field measurement model mapping a spherical image to visibility data. We implemented this model as the composition of an efficient Fourier-based interpolator mapping the spherical image onto the equatorial plane, with the standard RI operator mapping the equatorial-plane image to visibility data. Importantly, the interpolation step must inevitably be performed at a lower-than-optimal resolution on the plane, to meet the high-resolution requirement on the sphere of wide-field imaging while preserving scalability. Therefore, secondly, we design S-R2D2's DNN training loss to jointly learn to correct the interpolation approximations and identify residual image structures on the sphere, ensuring consistency with the spherical ground truth using the adjoint plane-to-sphere interpolator. Finally, we demonstrate through simulations S-R2D2's capability to perform fast and accurate reconstructions of spherical monochromatic intensity images, across high-resolution, high-dynamic-range settings.
Authors: Lu Dai, Yijie Xu, Jinhui Ye, Hao Liu, Hui Xiong
Abstract: Large Language Models (LLMs) have demonstrated improved generation performance by incorporating externally retrieved knowledge, a process known as retrieval-augmented generation (RAG). Despite the potential of this approach, existing studies evaluate RAG effectiveness by 1) assessing retrieval and generation components jointly, which obscures retrieval's distinct contribution, or 2) examining retrievers using traditional metrics such as NDCG, which creates a gap in understanding retrieval's true utility in the overall generation process. To address the above limitations, in this work, we introduce an automatic evaluation method that measures retrieval quality through the lens of information gain within the RAG framework. Specifically, we propose Semantic Perplexity (SePer), a metric that captures the LLM's internal belief about the correctness of the retrieved information. We quantify the utility of retrieval by the extent to which it reduces semantic perplexity post-retrieval. Extensive experiments demonstrate that SePer not only aligns closely with human preferences but also offers a more precise and efficient evaluation of retrieval utility across diverse RAG scenarios.
Authors: Simon Welker, Matthew Le, Ricky T. Q. Chen, Wei-Ning Hsu, Timo Gerkmann, Alexander Richard, Yi-Chiao Wu
Abstract: We propose FlowDec, a neural full-band audio codec for general audio sampled at 48 kHz that combines non-adversarial codec training with a stochastic postfilter based on a novel conditional flow matching method. Compared to the prior work ScoreDec which is based on score matching, we generalize from speech to general audio and move from 24 kbit/s to as low as 4 kbit/s, while improving output quality and reducing the required postfilter DNN evaluations from 60 to 6 without any fine-tuning or distillation techniques. We provide theoretical insights and geometric intuitions for our approach in comparison to ScoreDec as well as another recent work that uses flow matching, and conduct ablation studies on our proposed components. We show that FlowDec is a competitive alternative to the recent GAN-dominated stream of neural codecs, achieving FAD scores better than those of the established GAN-based codec DAC and listening test scores that are on par, and producing qualitatively more natural reconstructions for speech and harmonic structures in music.
Authors: Matteo Gasparin, Aaditya Ramdas
Abstract: Vovk (2015) introduced cross-conformal prediction, a modification of split conformal designed to improve the width of prediction sets. The method, when trained with a miscoverage rate equal to $\alpha$ and $n \gg K$, ensures a marginal coverage of at least $1 - 2\alpha - 2(1-\alpha)(K-1)/(n+K)$, where $n$ is the number of observations and $K$ denotes the number of folds. A simple modification of the method achieves coverage of at least $1-2\alpha$. In this work, we propose new variants of both methods that yield smaller prediction sets without compromising the latter theoretical guarantee. The proposed methods are based on recent results deriving more statistically efficient combination of p-values that leverage exchangeability and randomization. Simulations confirm the theoretical findings and bring out some important tradeoffs.
Authors: Disen Lan, Weigao Sun, Jiaxi Hu, Jusen Du, Yu Cheng
Abstract: Transformers with linear recurrent modeling offer linear-time training and constant-memory inference. Despite their demonstrated efficiency and performance, pretraining such non-standard architectures from scratch remains costly and risky. The linearization of large language models (LLMs) transforms pretrained standard models into linear recurrent structures, enabling more efficient deployment. However, current linearization methods typically introduce additional feature map modules that require extensive fine-tuning and overlook the gating mechanisms used in state-of-the-art linear recurrent models. To address these issues, this paper presents Liger, short for Linearizing LLMs to gated recurrent structures. Liger is a novel approach for converting pretrained LLMs into gated linear recurrent models without adding extra parameters. It repurposes the pretrained key matrix weights to construct diverse gating mechanisms, facilitating the formation of various gated recurrent structures while avoiding the need to train additional components from scratch. Using lightweight fine-tuning with Low-Rank Adaptation (LoRA), Liger restores the performance of the linearized gated recurrent models to match that of the original LLMs. Additionally, we introduce Liger Attention, an intra-layer hybrid attention mechanism, which significantly recovers 93\% of the Transformer-based LLM at 0.02\% pre-training tokens during the linearization process, achieving competitive results across multiple benchmarks, as validated on models ranging from 1B to 8B parameters. Code is available at https://github.com/OpenSparseLLMs/Linearization.
Authors: Ramkrishna Acharya
Abstract: In this paper, we propose a novel approach for air drawing that uses image processing techniques to draw on the screen by moving fingers in the air. This approach benefits a wide range of applications such as sign language, in-air drawing, and 'writing' in the air as a new way of input. The approach starts with preparing ROI (Region of Interest) background images by taking a running average in initial camera frames and later subtracting it from the live camera frames to get a binary mask image. We calculate the pointer's position as the top of the contour on the binary image. When drawing a circle on the canvas in that position, it simulates the drawing. Furthermore, we combine the pre-trained Tesseract model for OCR purposes. To address the false contours, we perform hand detection based on the haar cascade before performing the background subtraction. In an experimental setup, we achieved a latency of only 100ms in air drawing. The code used to this research are available in GitHub as https://github.com/q-viper/Contour-Based-Writing
Authors: Carlos Gomes, Isabelle Wittmann, Damien Robert, Johannes Jakubik, Tim Reichelt, Michele Martone, Stefano Maurogiovanni, Rikard Vinge, Jonas Hurst, Erik Scheurer, Rocco Sedona, Thomas Brunschwiler, Stefan Kesselheim, Matej Batic, Philip Stier, Jan Dirk Wegner, Gabriele Cavallaro, Edzer Pebesma, Michael Marszalek, Miguel A Belenguer-Plomer, Kennedy Adriko, Paolo Fraccaro, Romeo Kienzler, Rania Briq, Sabrina Benassou, Michele Lazzarini, Conrad M Albrecht
Abstract: Over the past decades, there has been an explosion in the amount of available Earth Observation (EO) data. The unprecedented coverage of the Earth's surface and atmosphere by satellite imagery has resulted in large volumes of data that must be transmitted to ground stations, stored in data centers, and distributed to end users. Modern Earth System Models (ESMs) face similar challenges, operating at high spatial and temporal resolutions, producing petabytes of data per simulated day. Data compression has gained relevance over the past decade, with neural compression (NC) emerging from deep learning and information theory, making EO data and ESM outputs ideal candidates due to their abundance of unlabeled data. In this review, we outline recent developments in NC applied to geospatial data. We introduce the fundamental concepts of NC including seminal works in its traditional applications to image and video compression domains with focus on lossy compression. We discuss the unique characteristics of EO and ESM data, contrasting them with "natural images", and explain the additional challenges and opportunities they present. Moreover, we review current applications of NC across various EO modalities and explore the limited efforts in ESM compression to date. The advent of self-supervised learning (SSL) and foundation models (FM) has advanced methods to efficiently distill representations from vast unlabeled data. We connect these developments to NC for EO, highlighting the similarities between the two fields and elaborate on the potential of transferring compressed feature representations for machine--to--machine communication. Based on insights drawn from this review, we devise future directions relevant to applications in EO and ESM.
Authors: Muhammad Aqeel, Shakiba Sharifi, Marco Cristani, Francesco Setti
Abstract: This study investigates the performance of robust anomaly detection models in industrial inspection, focusing particularly on their ability to handle noisy data. We propose to leverage the adaptation ability of meta learning approaches to identify and reject noisy training data to improve the learning process. In our model, we employ Model Agnostic Meta Learning (MAML) and an iterative refinement process through an Inter-Quartile Range rejection scheme to enhance their adaptability and robustness. This approach significantly improves the models capability to distinguish between normal and defective conditions. Our results of experiments conducted on well known MVTec and KSDD2 datasets demonstrate that the proposed method not only excels in environments with substantial noise but can also contribute in case of a clear training set, isolating those samples that are relatively out of distribution, thus offering significant improvements over traditional models.
Authors: Pedro Reis, Ana Paula Serra, Jo\~ao Gama
Abstract: This review systematically examines deep learning applications in financial asset management. Unlike prior reviews, this study focuses on identifying emerging trends, such as the integration of explainable artificial intelligence (XAI) and deep reinforcement learning (DRL), and their transformative potential. It highlights new developments, including hybrid models (e.g., transformer-based architectures) and the growing use of alternative data sources such as ESG indicators and sentiment analysis. These advancements challenge traditional financial paradigms and set the stage for a deeper understanding of the evolving landscape. We use the Scopus database to select the most relevant articles published from 2018 to 2023. The inclusion criteria encompassed articles that explicitly apply deep learning models within financial asset management. We excluded studies focused on physical assets. This review also outlines our methodology for evaluating the relevance and impact of the included studies, including data sources and analytical methods. Our search identified 934 articles, with 612 meeting the inclusion criteria based on their focus and methodology. The synthesis of results from these articles provides insights into the effectiveness of deep learning models in improving portfolio performance and price forecasting accuracy. The review highlights the broad applicability and potential enhancements deep learning offers to financial asset management. Despite some limitations due to the scope of model application and variation in methodological rigour, the overall evidence supports deep learning as a valuable tool in this field. Our systematic review underscores the progressive integration of deep learning in financial asset management, suggesting a trajectory towards more sophisticated and impactful applications.
Authors: Saeed Shakuri, Alireza Rezvanian
Abstract: Lung cancer has the highest rate of cancer-caused deaths, and early-stage diagnosis could increase the survival rate. Lung nodules are common indicators of lung cancer, making their detection crucial. Various lung nodule detection models exist, but many lack efficiency. Hence, we propose a more efficient approach by leveraging 2D CT slices, reducing computational load and complexity in training and inference. We employ the tiny version of Swin Transformer to benefit from Vision Transformers (ViT) while maintaining low computational cost. A Feature Pyramid Network is added to enhance detection, particularly for small nodules. Additionally, Transfer Learning is used to accelerate training. Our experimental results show that the proposed model outperforms state-of-the-art methods, achieving higher mAP and mAR for small nodules by 1.3% and 1.6%, respectively. Overall, our model achieves the highest mAP of 94.7% and mAR of 94.9%.
Authors: Zhanghao Hu, Hanqi Yan, Qingling Zhu, Zhenyi Shen, Yulan He, Lin Gui
Abstract: Large language models have recently pushed open domain question answering (ODQA) to new frontiers. However, prevailing retriever-reader pipelines often depend on multiple rounds of prompt level instructions, leading to high computational overhead, instability, and suboptimal retrieval coverage. In this paper, we propose EmbQA, an embedding-level framework that alleviates these shortcomings by enhancing both the retriever and the reader. Specifically, we refine query representations via lightweight linear layers under an unsupervised contrastive learning objective, thereby reordering retrieved passages to highlight those most likely to contain correct answers. Additionally, we introduce an exploratory embedding that broadens the model's latent semantic space to diversify candidate generation and employs an entropy-based selection mechanism to choose the most confident answer automatically. Extensive experiments across three open-source LLMs, three retrieval methods, and four ODQA benchmarks demonstrate that EmbQA substantially outperforms recent baselines in both accuracy and efficiency.
Authors: Tong Ge, Yashu Liu, Jieping Ye, Tianyi Li, Chao Wang
Abstract: Modern front-end (FE) development, especially when leveraging the unique features of frameworks like React and Vue, presents distinctive challenges. These include managing modular architectures, ensuring synchronization between data and visual outputs for declarative rendering, and adapting reusable components to various scenarios. Such complexities make it particularly difficult for state-of-the-art large vision-language models (VLMs) to generate accurate and functional code directly from design images. To address these challenges, we propose a reflective agentic workflow that synthesizes high-quality image-text data to capture the diverse characteristics of FE development. This workflow automates the extraction of self-contained\footnote{A \textbf{self-contained} code snippet is one that encapsulates all necessary logic, styling, and dependencies, ensuring it functions independently without requiring external imports or context.} code snippets from real-world projects, renders the corresponding visual outputs, and generates detailed descriptions that link design elements to functional code. To further expand the scope and utility of the synthesis, we introduce three data synthesis strategies: Evolution-based synthesis, which enables scalable and diverse dataset expansion; Waterfall-Model-based synthesis, which generates logically coherent code derived from system requirements; and Additive Development synthesis, which iteratively increases the complexity of human-authored components. We build a large vision-language model, Flame, trained on the synthesized datasets and demonstrate its effectiveness in generating React code via the $\text{pass}@k$ metric. Our results suggest that a code VLM trained to interpret images before code generation may achieve better performance.
Authors: Sueda Taner, Ziyi Wang, Christoph Studer
Abstract: We introduce a novel class of regularization functions, called Cauchy-Schwarz (CS) regularizers, which can be designed to induce a wide range of properties in solution vectors of optimization problems. To demonstrate the versatility of CS regularizers, we derive regularization functions that promote discrete-valued vectors, eigenvectors of a given matrix, and orthogonal matrices. The resulting CS regularizers are simple, differentiable, and can be free of spurious stationary points, making them suitable for gradient-based solvers and large-scale optimization problems. In addition, CS regularizers automatically adapt to the appropriate scale, which is, for example, beneficial when discretizing the weights of neural networks. To demonstrate the efficacy of CS regularizers, we provide results for solving underdetermined systems of linear equations and weight quantization in neural networks. Furthermore, we discuss specializations, variations, and generalizations, which lead to an even broader class of new and possibly more powerful regularizers.
Authors: Siya Qi, Rui Cao, Yulan He, Zheng Yuan
Abstract: With the rapid development of large language models (LLMs), LLM-as-a-judge has emerged as a widely adopted approach for text quality evaluation, including hallucination evaluation. While previous studies have focused exclusively on single-context evaluation (e.g., discourse faithfulness or world factuality), real-world hallucinations typically involve mixed contexts, which remains inadequately evaluated. In this study, we use summarization as a representative task to comprehensively evaluate LLMs' capability in detecting mixed-context hallucinations, specifically distinguishing between factual and non-factual hallucinations. Through extensive experiments across direct generation and retrieval-based models of varying scales, our main observations are: (1) LLMs' intrinsic knowledge introduces inherent biases in hallucination evaluation; (2) These biases particularly impact the detection of factual hallucinations, yielding a significant performance bottleneck; (3) The fundamental challenge lies in effective knowledge utilization, balancing between LLMs' intrinsic knowledge and external context for accurate mixed-context hallucination evaluation.
Authors: Elahe Delavari, John Moore, Junho Hong, Jaerock Kwon
Abstract: This paper presents a novel Perceptual Motor Learning (PML) framework integrated with Active Inference (AIF) to enhance lateral control in Highly Automated Vehicles (HAVs). PML, inspired by human motor learning, emphasizes the seamless integration of perception and action, enabling efficient decision-making in dynamic environments. Traditional autonomous driving approaches--including modular pipelines, imitation learning, and reinforcement learning--struggle with adaptability, generalization, and computational efficiency. In contrast, PML with AIF leverages a generative model to minimize prediction error ("surprise") and actively shape vehicle control based on learned perceptual-motor representations. Our approach unifies deep learning with active inference principles, allowing HAVs to perform lane-keeping maneuvers with minimal data and without extensive retraining across different environments. Extensive experiments in the CARLA simulator demonstrate that PML with AIF enhances adaptability without increasing computational overhead while achieving performance comparable to conventional methods. These findings highlight the potential of PML-driven active inference as a robust alternative for real-world autonomous driving applications.
Authors: Honglin Fu, Yebo Feng, Cong Wu, Jiahua Xu
Abstract: Masterminds are entities organizing, coordinating, and orchestrating cryptocurrency pump-and-dump schemes, a form of trade-based manipulation undermining market integrity and causing financial losses for unwitting investors. Previous research detects pump-and-dump activities in the market, predicts the target cryptocurrency, and examines investors and \ac{osn} entities. However, these solutions do not address the root cause of the problem. There is a critical gap in identifying and tracing the masterminds involved in these schemes. In this research, we develop a detection system \textsc{Perseus}, which collects real-time data from the \acs{osn} and cryptocurrency markets. \textsc{Perseus} then constructs temporal attributed graphs that preserve the direction of information diffusion and the structure of the community while leveraging \ac{gnn} to identify the masterminds behind pump-and-dump activities. Our design of \textsc{Perseus} leads to higher F1 scores and precision than the \ac{sota} fraud detection method, achieving fast training and inferring speeds. Deployed in the real world from February 16 to October 9 2024, \textsc{Perseus} successfully detects $438$ masterminds who are efficient in the pump-and-dump information diffusion networks. \textsc{Perseus} provides regulators with an explanation of the risks of masterminds and oversight capabilities to mitigate the pump-and-dump schemes of cryptocurrency.
Authors: Petr Sychev, Andrey Goncharov, Daniil Vyazhev, Edvard Khalafyan, Alexey Zaytsev
Abstract: Uncertainty estimation is crucial for evaluating Large Language Models (LLMs), particularly in high-stakes domains where incorrect answers result in significant consequences. Numerous approaches consider this problem, while focusing on a specific type of uncertainty, ignoring others. We investigate what estimates, specifically token-wise entropy and model-as-judge (MASJ), would work for multiple-choice question-answering tasks for different question topics. Our experiments consider three LLMs: Phi-4, Mistral, and Qwen of different sizes from 1.5B to 72B and $14$ topics. While MASJ performs similarly to a random error predictor, the response entropy predicts model error in knowledge-dependent domains and serves as an effective indicator of question difficulty: for biology ROC AUC is $0.73$. This correlation vanishes for the reasoning-dependent domain: for math questions ROC-AUC is $0.55$. More principally, we found out that the entropy measure required a reasoning amount. Thus, data-uncertainty related entropy should be integrated within uncertainty estimates frameworks, while MASJ requires refinement. Moreover, existing MMLU-Pro samples are biased, and should balance required amount of reasoning for different subdomains to provide a more fair assessment of LLMs performance.
Authors: Yuyan Chen, Nico Lang, B. Christian Schmidt, Aditya Jain, Yves Basset, Sara Beery, Maxim Larriv\'ee, David Rolnick
Abstract: Machine learning is increasingly being applied to facilitate long-term, large-scale biodiversity monitoring. With most species on Earth still undiscovered or poorly documented, species-recognition models are expected to encounter new species during deployment. We introduce Open-Insects, a fine-grained image recognition benchmark dataset for open-set recognition and out-of-distribution detection in biodiversity monitoring. Open-Insects makes it possible to evaluate algorithms for new species detection on several geographical open-set splits with varying difficulty. Furthermore, we present a test set recently collected in the wild with 59 species that are likely new to science. We evaluate a variety of open-set recognition algorithms, including post-hoc methods, training-time regularization, and training with auxiliary data, finding that the simple post-hoc approach of utilizing softmax scores remains a strong baseline. We also demonstrate how to leverage auxiliary data to improve the detection performance when the training dataset is limited. Our results provide timely insights to guide the development of computer vision methods for biodiversity monitoring and species discovery.
Authors: Francesco Bacchiocchi, Matteo Castiglioni, Alberto Marchesi, Nicola Gatti
Abstract: Most microeconomic models of interest involve optimizing a piecewise linear function. These include contract design in hidden-action principal-agent problems, selling an item in posted-price auctions, and bidding in first-price auctions. When the relevant model parameters are unknown and determined by some (unknown) probability distributions, the problem becomes learning how to optimize an unknown and stochastic piecewise linear reward function. Such a problem is usually framed within an online learning framework, where the decision-maker (learner) seeks to minimize the regret of not knowing an optimal decision in hindsight. This paper introduces a general online learning framework that offers a unified approach to tackle regret minimization for piecewise linear rewards, under a suitable monotonicity assumption commonly satisfied by microeconomic models. We design a learning algorithm that attains a regret of $\widetilde{O}(\sqrt{nT})$, where $n$ is the number of ``pieces'' of the reward function and $T$ is the number of rounds. This result is tight when $n$ is \emph{small} relative to $T$, specifically when $n \leq T^{1/3}$. Our algorithm solves two open problems in the literature on learning in microeconomic settings. First, it shows that the $\widetilde{O}(T^{2/3})$ regret bound obtained by Zhu et al. [Zhu+23] for learning optimal linear contracts in hidden-action principal-agent problems is not tight when the number of agent's actions is small relative to $T$. Second, our algorithm demonstrates that, in the problem of learning to set prices in posted-price auctions, it is possible to attain suitable (and desirable) instance-independent regret bounds, addressing an open problem posed by Cesa-Bianchi et al. [CBCP19].
Authors: Shishir Adhikari, Sourav Medya, Elena Zheleva
Abstract: In causal inference, interference refers to the phenomenon in which the actions of peers in a network can influence an individual's outcome. Peer effect refers to the difference in counterfactual outcomes of an individual for different levels of peer exposure, the extent to which an individual is exposed to the treatments, actions, or behaviors of peers. Estimating peer effects requires deciding how to represent peer exposure. Typically, researchers define an exposure mapping function that aggregates peer treatments and outputs peer exposure. Most existing approaches for defining exposure mapping functions assume peer exposure based on the number or fraction of treated peers. Recent studies have investigated more complex functions of peer exposure which capture that different peers can exert different degrees of influence. However, none of these works have explicitly considered the problem of automatically learning the exposure mapping function. In this work, we focus on learning this function for the purpose of estimating heterogeneous peer effects, where heterogeneity refers to the variation in counterfactual outcomes for the same peer exposure but different individual's contexts. We develop EgoNetGNN, a graph neural network (GNN)-based method, to automatically learn the appropriate exposure mapping function allowing for complex peer influence mechanisms that, in addition to peer treatments, can involve the local neighborhood structure and edge attributes. We show that GNN models that use peer exposure based on the number or fraction of treated peers or learn peer exposure naively face difficulty accounting for such influence mechanisms. Our comprehensive evaluation on synthetic and semi-synthetic network data shows that our method is more robust to different unknown underlying influence mechanisms when estimating heterogeneous peer effects when compared to state-of-the-art baselines.
Authors: Abdelrahman Abouelenin, Atabak Ashfaq, Adam Atkinson, Hany Awadalla, Nguyen Bach, Jianmin Bao, Alon Benhaim, Martin Cai, Vishrav Chaudhary, Congcong Chen, Dong Chen, Dongdong Chen, Junkun Chen, Weizhu Chen, Yen-Chun Chen, Yi-ling Chen, Qi Dai, Xiyang Dai, Ruchao Fan, Mei Gao, Min Gao, Amit Garg, Abhishek Goswami, Junheng Hao, Amr Hendy, Yuxuan Hu, Xin Jin, Mahmoud Khademi, Dongwoo Kim, Young Jin Kim, Gina Lee, Jinyu Li, Yunsheng Li, Chen Liang, Xihui Lin, Zeqi Lin, Mengchen Liu, Yang Liu, Gilsinia Lopez, Chong Luo, Piyush Madan, Vadim Mazalov, Ali Mousavi, Anh Nguyen, Jing Pan, Daniel Perez-Becker, Jacob Platin, Thomas Portet, Kai Qiu, Bo Ren, Liliang Ren, Sambuddha Roy, Ning Shang, Yelong Shen, Saksham Singhal, Subhojit Som, Xia Song, Tetyana Sych, Praneetha Vaddamanu, Shuohang Wang, Yiming Wang, Zhenghao Wang, Haibin Wu, Haoran Xu, Weijian Xu, Yifan Yang, Ziyi Yang, Donghan Yu, Ishmam Zabir, Jianwen Zhang, Li Lyna Zhang, Yunan Zhang, Xiren Zhou
Abstract: We introduce Phi-4-Mini and Phi-4-Multimodal, compact yet highly capable language and multimodal models. Phi-4-Mini is a 3.8-billion-parameter language model trained on high-quality web and synthetic data, significantly outperforming recent open-source models of similar size and matching the performance of models twice its size on math and coding tasks requiring complex reasoning. This achievement is driven by a carefully curated synthetic data recipe emphasizing high-quality math and coding datasets. Compared to its predecessor, Phi-3.5-Mini, Phi-4-Mini features an expanded vocabulary size of 200K tokens to better support multilingual applications, as well as group query attention for more efficient long-sequence generation. Phi-4-Multimodal is a multimodal model that integrates text, vision, and speech/audio input modalities into a single model. Its novel modality extension approach leverages LoRA adapters and modality-specific routers to allow multiple inference modes combining various modalities without interference. For example, it now ranks first in the OpenASR leaderboard to date, although the LoRA component of the speech/audio modality has just 460 million parameters. Phi-4-Multimodal supports scenarios involving (vision + language), (vision + speech), and (speech/audio) inputs, outperforming larger vision-language and speech-language models on a wide range of tasks. Additionally, we experiment to further train Phi-4-Mini to enhance its reasoning capabilities. Despite its compact 3.8-billion-parameter size, this experimental version achieves reasoning performance on par with or surpassing significantly larger models, including DeepSeek-R1-Distill-Qwen-7B and DeepSeek-R1-Distill-Llama-8B.
Authors: Sam Bowyer, Laurence Aitchison, Desi R. Ivanova
Abstract: Rigorous statistical evaluations of large language models (LLMs), including valid error bars and significance testing, are essential for meaningful and reliable performance assessment. Currently, when such statistical measures are reported, they typically rely on the Central Limit Theorem (CLT). In this position paper, we argue that while CLT-based methods for uncertainty quantification are appropriate when benchmarks consist of thousands of examples, they fail to provide adequate uncertainty estimates for LLM evaluations that rely on smaller, highly specialized benchmarks. In these small-data settings, we demonstrate that CLT-based methods perform very poorly, usually dramatically underestimating uncertainty (i.e. producing error bars that are too small). We give recommendations for alternative frequentist and Bayesian methods that are both easy to implement and more appropriate in these increasingly common scenarios. We provide a simple Python library for these Bayesian methods at https://github.com/sambowyer/bayes_evals .
Authors: Marco Scialanga, Thibault Laugel, Vincent Grari, Marcin Detyniecki
Abstract: As Large Langue Models have been shown to memorize real-world facts, the need to update this knowledge in a controlled and efficient manner arises. Designed with these constraints in mind, Knowledge Editing (KE) approaches propose to alter specific facts in pretrained models. However, they have been shown to suffer from several limitations, including their lack of contextual robustness and their failure to generalize to logical implications related to the fact. To overcome these issues, we propose SAKE, a steering activation method that models a fact to be edited as a distribution rather than a single prompt. Leveraging Optimal Transport, SAKE alters the LLM behavior over a whole fact-related distribution, defined as paraphrases and logical implications. Several numerical experiments demonstrate the effectiveness of this method: SAKE is thus able to perform more robust edits than its existing counterparts.
Authors: Valentio Iverson, Gautam Kamath, Argyris Mouzakis
Abstract: We provide the first $\widetilde{\mathcal{O}}\left(d\right)$-sample algorithm for sampling from unbounded Gaussian distributions under the constraint of $\left(\varepsilon, \delta\right)$-differential privacy. This is a quadratic improvement over previous results for the same problem, settling an open question of Ghazi, Hu, Kumar, and Manurangsi.
Authors: Chenning Li, Anton A. Zabreyko, Arash Nasr-Esfahany, Kevin Zhao, Prateesh Goyal, Mohammad Alizadeh, Thomas Anderson
Abstract: Flow-level simulation is widely used to model large-scale data center networks due to its scalability. Unlike packet-level simulators that model individual packets, flow-level simulators abstract traffic as continuous flows with dynamically assigned transmission rates. While this abstraction enables orders-of-magnitude speedup, it is inaccurate by omitting critical packet-level effects such as queuing, congestion control, and retransmissions. We present m4, an accurate and scalable flow-level simulator that uses machine learning to learn the dynamics of the network of interest. At the core of m4 lies a novel ML architecture that decomposes state transition computations into distinct spatial and temporal components, each represented by a suitable neural network. To efficiently learn the underlying flow-level dynamics, m4 adds dense supervision signals by predicting intermediate network metrics such as remaining flow size and queue length during training. m4 achieves a speedup of up to 104$\times$ over packet-level simulation. Relative to a traditional flow-level simulation, m4 reduces per-flow estimation errors by 45.3% (mean) and 53.0% (p90). For closed-loop applications, m4 accurately predicts network throughput under various congestion control schemes and workloads.
Authors: Md. Farhan Shahriyar, Gazi Tanbhir, Abdullah Md Raihan Chy, Mohammed Abdul Al Arafat Tanzin, Md. Jisan Mashrafi
Abstract: Phishing URL detection is crucial in cybersecurity as malicious websites disguise themselves to steal sensitive infor mation. Traditional machine learning techniques struggle to per form well in complex real-world scenarios due to large datasets and intricate patterns. Motivated by quantum computing, this paper proposes using Variational Quantum Classifiers (VQC) to enhance phishing URL detection. We present PhishVQC, a quantum model that combines quantum feature maps and vari ational ansatzes such as RealAmplitude and EfficientSU2. The model is evaluated across two experimental setups with varying dataset sizes and feature map repetitions. PhishVQC achieves a maximum macro average F1-score of 0.89, showing a 22% improvement over prior studies. This highlights the potential of quantum machine learning to improve phishing detection accuracy. The study also notes computational challenges, with execution wall times increasing as dataset size grows.
Authors: Mohammad Albinhassan, Pranava Madhyastha, Alessandra Russo
Abstract: Ensuring both syntactic and semantic correctness in Large Language Model (LLM) outputs remains a significant challenge, despite being critical for real-world deployment. In this paper, we introduce $\texttt{SEM-CTRL}$, a unified approach that enforces rich context-sensitive constraints and task- and instance-specific semantics directly on an LLM decoder. Our approach integrates token-level MCTS, which is guided by specific syntactic and semantic constraints. The constraints over the desired outputs are expressed using Answer Set Grammars -- a logic-based formalism that generalizes context-sensitive grammars while incorporating background knowledge to represent task-specific semantics. We show that our approach guarantees correct completions for any off-the-shelf LLM without the need for fine-tuning. We evaluate $\texttt{SEM-CTRL}$ on a range of tasks, including synthetic grammar synthesis, combinatorial reasoning, and planning. Our results demonstrate that $\texttt{SEM-CTRL}$ allows small pre-trained LLMs to efficiently outperform larger variants and state-of-the-art reasoning models (e.g., o1-preview) while simultaneously guaranteeing solution correctness.
Authors: Nicholas Carlini, Javier Rando, Edoardo Debenedetti, Milad Nasr, Florian Tram\`er
Abstract: We introduce AutoAdvExBench, a benchmark to evaluate if large language models (LLMs) can autonomously exploit defenses to adversarial examples. Unlike existing security benchmarks that often serve as proxies for real-world tasks, bench directly measures LLMs' success on tasks regularly performed by machine learning security experts. This approach offers a significant advantage: if a LLM could solve the challenges presented in bench, it would immediately present practical utility for adversarial machine learning researchers. We then design a strong agent that is capable of breaking 75% of CTF-like ("homework exercise") adversarial example defenses. However, we show that this agent is only able to succeed on 13% of the real-world defenses in our benchmark, indicating the large gap between difficulty in attacking "real" code, and CTF-like code. In contrast, a stronger LLM that can attack 21% of real defenses only succeeds on 54% of CTF-like defenses. We make this benchmark available at https://github.com/ethz-spylab/AutoAdvExBench.
Authors: Weizhi Zhang, Liangwei Yang, Wooseong Yang, Henry Peng Zou, Yuqing Liu, Ke Xu, Sourav Medya, Philip S. Yu
Abstract: Collaborative filtering models, particularly graph-based approaches, have demonstrated strong performance in capturing user-item interactions for recommendation systems. However, they continue to struggle in cold-start and data-sparse scenarios. The emergence of large language models (LLMs) like GPT and LLaMA presents new possibilities for enhancing recommendation performance, especially in cold-start settings. Despite their promise, LLMs pose challenges related to scalability and efficiency due to their high computational demands and limited ability to model complex user-item relationships effectively. In this work, we introduce a novel perspective on leveraging LLMs for CF model initialization. Through experiments, we uncover an embedding collapse issue when scaling CF models to larger embedding dimensions. To effectively harness large-scale LLM embeddings, we propose innovative selective initialization strategies utilizing random, uniform, and variance-based index sampling. Our comprehensive evaluation on multiple real-world datasets demonstrates significant performance gains across various CF models while maintaining a lower computational cost compared to existing LLM-based recommendation approaches.
Authors: Nimet Beyza Bozdag, Shuhaib Mehri, Gokhan Tur, Dilek Hakkani-T\"ur
Abstract: Large Language Models (LLMs) demonstrate persuasive capabilities that rival human-level persuasion. While these capabilities can be used for social good, they also present risks of potential misuse. Moreover, LLMs' susceptibility to persuasion raises concerns about alignment with ethical principles. To study these dynamics, we introduce Persuade Me If You Can (PMIYC), an automated framework for evaluating persuasion through multi-agent interactions. Here, Persuader agents engage in multi-turn conversations with the Persuadee agents, allowing us to measure LLMs' persuasive effectiveness and their susceptibility to persuasion. We conduct comprehensive evaluations across diverse LLMs, ensuring each model is assessed against others in both subjective and misinformation contexts. We validate the efficacy of our framework through human evaluations and show alignment with prior work. PMIYC offers a scalable alternative to human annotation for studying persuasion in LLMs. Through PMIYC, we find that Llama-3.3-70B and GPT-4o exhibit similar persuasive effectiveness, outperforming Claude 3 Haiku by 30%. However, GPT-4o demonstrates over 50% greater resistance to persuasion for misinformation compared to Llama-3.3-70B. These findings provide empirical insights into the persuasive dynamics of LLMs and contribute to the development of safer AI systems.
Authors: Andr\'e Jonasson
Abstract: Transformer-based Large Language Models (LLMs) rely on positional encodings to provide sequence position information to their attention mechanism. Rotary Positional Encodings (RoPE), which encode relative position by rotating queries and keys, have become widely used in modern LLMs. We study the features and patterns that emerge in queries and keys when using rotary embeddings. Our analysis reveals consistent patterns within the same model across layers and attention heads and across different models and architectures. We present and apply analysis techniques and show how the queries and keys use RoPE to construct various attention patterns, including attention sinks. We find and analyze outliers across models in queries and keys and find that they are likely to be found in rotary features with partial cycles. We derive bounds that tell us what rotary frequencies are likely to be selected as outlier features and at what minimum angle the query-key rotary pairs in these features tend to be above and verify the bounds empirically with models of significant architectural differences.
Authors: Haoran You, Xiaohan Chen, Yongan Zhang, Chaojian Li, Sicheng Li, Zihao Liu, Zhangyang Wang, Yingyan Celine Lin
Abstract: Multiplication (e.g., convolution) is arguably a cornerstone of modern deep neural networks (DNNs). However, intensive multiplications cause expensive resource costs that challenge DNNs' deployment on resource-constrained edge devices, driving several attempts for multiplication-less deep networks. This paper presented ShiftAddNet, whose main inspiration is drawn from a common practice in energy-efficient hardware implementation, that is, multiplication can be instead performed with additions and logical bit-shifts. We leverage this idea to explicitly parameterize deep networks in this way, yielding a new type of deep network that involves only bit-shift and additive weight layers. This hardware-inspired ShiftAddNet immediately leads to both energy-efficient inference and training, without compromising the expressive capacity compared to standard DNNs. The two complementary operation types (bit-shift and add) additionally enable finer-grained control of the model's learning capacity, leading to more flexible trade-off between accuracy and (training) efficiency, as well as improved robustness to quantization and pruning. We conduct extensive experiments and ablation studies, all backed up by our FPGA-based ShiftAddNet implementation and energy measurements. Compared to existing DNNs or other multiplication-less models, ShiftAddNet aggressively reduces over 80% hardware-quantified energy cost of DNNs training and inference, while offering comparable or better accuracies. Codes and pre-trained models are available at https://github.com/RICE-EIC/ShiftAddNet.
Authors: Haoran You, Randall Balestriero, Zhihan Lu, Yutong Kou, Huihong Shi, Shunyao Zhang, Shang Wu, Yingyan Celine Lin, Richard Baraniuk
Abstract: In this paper, we study the importance of pruning in Deep Networks (DNs) and the yin & yang relationship between (1) pruning highly overparametrized DNs that have been trained from random initialization and (2) training small DNs that have been "cleverly" initialized. As in most cases practitioners can only resort to random initialization, there is a strong need to develop a grounded understanding of DN pruning. Current literature remains largely empirical, lacking a theoretical understanding of how pruning affects DNs' decision boundary, how to interpret pruning, and how to design corresponding principled pruning techniques. To tackle those questions, we propose to employ recent advances in the theoretical analysis of Continuous Piecewise Affine (CPA) DNs. From this perspective, we will be able to detect the early-bird (EB) ticket phenomenon, provide interpretability into current pruning techniques, and develop a principled pruning strategy. In each step of our study, we conduct extensive experiments supporting our claims and results; while our main goal is to enhance the current understanding towards DN pruning instead of developing a new pruning method, our spline pruning criteria in terms of layerwise and global pruning is on par with or even outperforms state-of-the-art pruning methods.
Authors: Haoran You, Zhihan Lu, Zijian Zhou, Yonggan Fu, Yingyan Celine Lin
Abstract: Graph Convolutional Networks (GCNs) have emerged as the state-of-the-art deep learning model for representation learning on graphs. However, it remains notoriously challenging to train and inference GCNs over large graph datasets, limiting their application to large real-world graphs and hindering the exploration of deeper and more sophisticated GCN graphs. This is because as the graph size grows, the sheer number of node features and the large adjacency matrix can easily explode the required memory and data movements. To tackle the aforementioned challenges, we explore the possibility of drawing lottery tickets when sparsifying GCN graphs, i.e., subgraphs that largely shrink the adjacency matrix yet are capable of achieving accuracy comparable to or even better than their full graphs. Specifically, we for the first time discover the existence of graph early-bird (GEB) tickets that emerge at the very early stage when sparsifying GCN graphs, and propose a simple yet effective detector to automatically identify the emergence of such GEB tickets. Furthermore, we advocate graph-model co-optimization and develop a generic efficient GCN early-bird training framework dubbed GEBT that can significantly boost the efficiency of GCN training by (1) drawing joint early-bird tickets between the GCN graphs and models and (2) enabling simultaneously sparsification of both the GCN graphs and models. Experiments on various GCN models and datasets consistently validate our GEB finding and the effectiveness of our GEBT, e.g., our GEBT achieves up to 80.2% ~ 85.6% and 84.6% ~ 87.5% savings of GCN training and inference costs while offering a comparable or even better accuracy as compared to state-of-the-art methods. Our source code and supplementary appendix are available at https://github.com/RICE-EIC/Early-Bird-GCN.
Authors: Haoran You, Baopu Li, Huihong Shi, Yonggan Fu, Yingyan Celine Lin
Abstract: Neural networks (NNs) with intensive multiplications (e.g., convolutions and transformers) are capable yet power hungry, impeding their more extensive deployment into resource-constrained devices. As such, multiplication-free networks, which follow a common practice in energy-efficient hardware implementation to parameterize NNs with more efficient operators (e.g., bitwise shifts and additions), have gained growing attention. However, multiplication-free networks usually under-perform their vanilla counterparts in terms of the achieved accuracy. To this end, this work advocates hybrid NNs that consist of both powerful yet costly multiplications and efficient yet less powerful operators for marrying the best of both worlds, and proposes ShiftAddNAS, which can automatically search for more accurate and more efficient NNs. Our ShiftAddNAS highlights two enablers. Specifically, it integrates (1) the first hybrid search space that incorporates both multiplication-based and multiplication-free operators for facilitating the development of both accurate and efficient hybrid NNs; and (2) a novel weight sharing strategy that enables effective weight sharing among different operators that follow heterogeneous distributions (e.g., Gaussian for convolutions vs. Laplacian for add operators) and simultaneously leads to a largely reduced supernet size and much better searched networks. Extensive experiments and ablation studies on various models, datasets, and tasks consistently validate the efficacy of ShiftAddNAS, e.g., achieving up to a +7.7% higher accuracy or a +4.9 better BLEU score compared to state-of-the-art NN, while leading to up to 93% or 69% energy and latency savings, respectively. Codes and pretrained models are available at https://github.com/RICE-EIC/ShiftAddNAS.
Authors: Sarem Seitz
Abstract: Gaussian Process (GP) models are a powerful tool in probabilistic machine learning with a solid theoretical foundation. Thanks to current advances, modeling complex data with GPs is becoming increasingly feasible, which makes them an interesting alternative to deep learning and related approaches. As the latter are getting more and more influential on society, the need for making a model's decision making process transparent and explainable is now a major focus of research. A major direction in interpretable machine learning is the use of gradient-based approaches, such as Integrated Gradients, to quantify feature attribution, locally for a given datapoint of interest. Since GPs and the behavior of their partial derivatives are well studied and straightforward to derive, studying gradient-based explainability for GPs is a promising direction of research. Unfortunately, partial derivatives for GPs become less trivial to handle when dealing with non-Gaussian target data as in classification or more sophisticated regression problems. This paper therefore proposes an approach for applying Integrated Gradient-based explainability to non-Gaussian GP models, offering both analytical and approximate solutions. This extends gradient-based explainability to probabilistic models with complex likelihoods to extend their practical applicability.
Authors: Pierre Marion, Adeline Fermanian, G\'erard Biau, Jean-Philippe Vert
Abstract: Deep ResNets are recognized for achieving state-of-the-art results in complex machine learning tasks. However, the remarkable performance of these architectures relies on a training procedure that needs to be carefully crafted to avoid vanishing or exploding gradients, particularly as the depth $L$ increases. No consensus has been reached on how to mitigate this issue, although a widely discussed strategy consists in scaling the output of each layer by a factor $\alpha_L$. We show in a probabilistic setting that with standard i.i.d.~initializations, the only non-trivial dynamics is for $\alpha_L = \frac{1}{\sqrt{L}}$; other choices lead either to explosion or to identity mapping. This scaling factor corresponds in the continuous-time limit to a neural stochastic differential equation, contrarily to a widespread interpretation that deep ResNets are discretizations of neural ordinary differential equations. By contrast, in the latter regime, stability is obtained with specific correlated initializations and $\alpha_L = \frac{1}{L}$. Our analysis suggests a strong interplay between scaling and regularity of the weights as a function of the layer index. Finally, in a series of experiments, we exhibit a continuous range of regimes driven by these two parameters, which jointly impact performance before and after training.
Authors: Junhao Liu, Xin Zhang
Abstract: Existing local model-agnostic explanation techniques are ineffective for machine learning models that consider inputs of variable lengths, as they do not consider temporal information embedded in these models. To address this limitation, we propose \textsc{ReX}, a general framework for incorporating temporal information in these techniques. Our key insight is that these techniques typically learn a model surrogate by sampling model inputs and outputs, and we can incorporate temporal information in a uniform way by only changing the sampling process and the surrogate features. We instantiate our approach on three popular explanation techniques: Anchors, LIME, and Kernel SHAP. To evaluate the effectiveness of \textsc{ReX}, we apply our approach to six models in three different tasks. Our evaluation results demonstrate that our approach 1) significantly improves the fidelity of explanations, making model-agnostic techniques outperform a state-of-the-art model-specific technique on its target model, and 2) helps end users better understand the models' behaviors.
Authors: Haoran You, Zhanyi Sun, Huihong Shi, Zhongzhi Yu, Yang Zhao, Yongan Zhang, Chaojian Li, Baopu Li, Yingyan Celine Lin
Abstract: Vision Transformers (ViTs) have achieved state-of-the-art performance on various vision tasks. However, ViTs' self-attention module is still arguably a major bottleneck, limiting their achievable hardware efficiency. Meanwhile, existing accelerators dedicated to NLP Transformers are not optimal for ViTs. This is because there is a large difference between ViTs and NLP Transformers: ViTs have a relatively fixed number of input tokens, whose attention maps can be pruned by up to 90% even with fixed sparse patterns; while NLP Transformers need to handle input sequences of varying numbers of tokens and rely on on-the-fly predictions of dynamic sparse attention patterns for each input to achieve a decent sparsity (e.g., >=50%). To this end, we propose a dedicated algorithm and accelerator co-design framework dubbed ViTCoD for accelerating ViTs. Specifically, on the algorithm level, ViTCoD prunes and polarizes the attention maps to have either denser or sparser fixed patterns for regularizing two levels of workloads without hurting the accuracy, largely reducing the attention computations while leaving room for alleviating the remaining dominant data movements; on top of that, we further integrate a lightweight and learnable auto-encoder module to enable trading the dominant high-cost data movements for lower-cost computations. On the hardware level, we develop a dedicated accelerator to simultaneously coordinate the enforced denser/sparser workloads and encoder/decoder engines for boosted hardware utilization. Extensive experiments and ablation studies validate that ViTCoD largely reduces the dominant data movement costs, achieving speedups of up to 235.3x, 142.9x, 86.0x, 10.1x, and 6.8x over general computing platforms CPUs, EdgeGPUs, GPUs, and prior-art Transformer accelerators SpAtten and Sanger under an attention sparsity of 90%, respectively.
Authors: Shiyu Wang, Fan Zhou, Yinbo Sun, Lintao Ma, James Zhang, Yangfei Zheng
Abstract: Multivariate time series forecasting with hierarchical structure is pervasive in real-world applications, demanding not only predicting each level of the hierarchy, but also reconciling all forecasts to ensure coherency, i.e., the forecasts should satisfy the hierarchical aggregation constraints. Moreover, the disparities of statistical characteristics between levels can be huge, worsened by non-Gaussian distributions and non-linear correlations. To this extent, we propose a novel end-to-end hierarchical time series forecasting model, based on conditioned normalizing flow-based autoregressive transformer reconciliation, to represent complex data distribution while simultaneously reconciling the forecasts to ensure coherency. Unlike other state-of-the-art methods, we achieve the forecasting and reconciliation simultaneously without requiring any explicit post-processing step. In addition, by harnessing the power of deep model, we do not rely on any assumption such as unbiased estimates or Gaussian distribution. Our evaluation experiments are conducted on four real-world hierarchical datasets from different industrial domains (three public ones and a dataset from the application servers of Alipay's data center) and the preliminary results demonstrate efficacy of our proposed method.
Authors: Pouya Hamadanian, Arash Nasr-Esfahany, Malte Schwarzkopf, Siddartha Sen, Mohammad Alizadeh
Abstract: We study online reinforcement learning (RL) in non-stationary environments, where a time-varying exogenous context process affects the environment dynamics. Online RL is challenging in such environments due to "catastrophic forgetting" (CF). The agent tends to forget prior knowledge as it trains on new experiences. Prior approaches to mitigate this issue assume task labels (which are often not available in practice), employ brittle regularization heuristics or use off-policy methods that suffer from instability and poor performance. We present Locally Constrained Policy Optimization (LCPO), an online RL approach that combats CF by anchoring policy outputs on old experiences while optimizing the return on current experiences. To perform this anchoring, LCPO locally constrains policy optimization using samples from experiences that lie outside of the current context distribution. We evaluate LCPO in Mujoco, classic control and computer systems environments with a variety of synthetic and real context traces, and find that it outperforms a variety of baselines in the non-stationary setting, while achieving results on-par with a "prescient" agent trained offline across all context traces. LCPO's source code is available at https://github.com/pouyahmdn/LCPO.
Authors: Max Hopkins, Daniel M. Kane, Shachar Lovett, Gaurav Mahajan
Abstract: The Fundamental Theorem of PAC Learning asserts that learnability of a concept class $H$ is equivalent to the $\textit{uniform convergence}$ of empirical error in $H$ to its mean, or equivalently, to the problem of $\textit{density estimation}$, learnability of the underlying marginal distribution with respect to events in $H$. This seminal equivalence relies strongly on PAC learning's `distribution-free' assumption, that the adversary may choose any marginal distribution over data. Unfortunately, the distribution-free model is known to be overly adversarial in practice, failing to predict the success of modern machine learning algorithms, but without the Fundamental Theorem our theoretical understanding of learning under distributional constraints remains highly limited. In this work, we revisit the connection between PAC learning, uniform convergence, and density estimation beyond the distribution-free setting when the adversary is restricted to choosing a marginal distribution from a known family $\mathscr{P}$. We prove that while the traditional Fundamental Theorem indeed fails, a finer-grained connection between the three fundamental notions continues to hold: 1. PAC-Learning is strictly sandwiched between two refined models of density estimation, differing only in whether the learner $\textit{knows}$ the set of well-estimated events in $H$. 2. Under reasonable assumptions on $H$ and $\mathscr{P}$, density estimation is equivalent to $\textit{uniform estimation}$, a relaxation of uniform convergence allowing non-empirical estimators. Together, our results give a clearer picture of how the Fundamental Theorem extends beyond the distribution-free setting and shed new light on the classically challenging problem of learning under arbitrary distributional assumptions.
Authors: Jaemin Yoo, Lingxiao Zhao, Leman Akoglu
Abstract: Self-supervised learning (SSL) has emerged as a promising paradigm that presents supervisory signals to real-world problems, bypassing the extensive cost of manual labeling. Consequently, self-supervised anomaly detection (SSAD) has seen a recent surge of interest, since SSL is especially attractive for unsupervised tasks. However, recent works have reported that the choice of a data augmentation function has significant impact on the accuracy of SSAD, posing augmentation search as an essential but nontrivial problem with the lack of labeled validation data. In this paper, we introduce ST-SSAD, the first systematic approach for rigorous augmentation tuning on SSAD. To this end, our work presents two key contributions. The first is a new unsupervised validation loss that quantifies the alignment between augmented training data and unlabeled validation data. The second is new differentiable augmentation functions, allowing data augmentation hyperparameter(s) to be tuned in an end-to-end manner. Experiments on two testbeds with semantic class anomalies and subtle industrial defects show that ST-SSAD gives significant performance gains over existing works.
Authors: Yuwei Luo, Mohsen Bayati
Abstract: This paper is motivated by recent research in the $d$-dimensional stochastic linear bandit literature, which has revealed an unsettling discrepancy: algorithms like Thompson sampling and Greedy demonstrate promising empirical performance, yet this contrasts with their pessimistic theoretical regret bounds. The challenge arises from the fact that while these algorithms may perform poorly in certain problem instances, they generally excel in typical instances. To address this, we propose a new data-driven technique that tracks the geometric properties of the uncertainty ellipsoid around the main problem parameter. This methodology enables us to formulate a data-driven frequentist regret bound, which incorporates the geometric information, for a broad class of base algorithms, including Greedy, OFUL, and Thompson sampling. This result allows us to identify and ``course-correct" problem instances in which the base algorithms perform poorly. The course-corrected algorithms achieve the minimax optimal regret of order $\tilde{\mathcal{O}}(d\sqrt{T})$ for a $T$-period decision-making scenario, effectively maintaining the desirable attributes of the base algorithms, including their empirical efficacy. We present simulation results to validate our findings using synthetic and real data.
Authors: Jung Hwan Heo, Jeonghoon Kim, Beomseok Kwon, Byeongwook Kim, Se Jung Kwon, Dongsoo Lee
Abstract: Large Language Models (LLMs) have recently demonstrated remarkable success across various tasks. However, efficiently serving LLMs has been a challenge due to the large memory bottleneck, specifically in small batch inference settings (e.g. mobile devices). Weight-only quantization can be a promising approach, but sub-4 bit quantization remains a challenge due to large-magnitude activation outliers. To mitigate the undesirable outlier effect, we first propose per-IC quantization, a simple yet effective method that creates quantization groups within each input channel (IC) rather than the conventional per-output-channel (per-OC). Our method is motivated by the observation that activation outliers affect the input dimension of the weight matrix, so similarly grouping the weights in the IC direction can isolate outliers within a group. We also find that activation outliers do not dictate quantization difficulty, and inherent weight sensitivities also exist. With per-IC quantization as a new outlier-friendly scheme, we propose Adaptive Dimensions (AdaDim), a versatile quantization framework that can adapt to various weight sensitivity patterns. We demonstrate the effectiveness of AdaDim by augmenting prior methods such as Round-To-Nearest and GPTQ, showing significant improvements across various language modeling benchmarks for both base (up to +4.7% on MMLU) and instruction-tuned (up to +10% on HumanEval) LLMs. Code is available at https://github.com/johnheo/adadim-llm
Authors: Andy Zou, Long Phan, Sarah Chen, James Campbell, Phillip Guo, Richard Ren, Alexander Pan, Xuwang Yin, Mantas Mazeika, Ann-Kathrin Dombrowski, Shashwat Goel, Nathaniel Li, Michael J. Byun, Zifan Wang, Alex Mallen, Steven Basart, Sanmi Koyejo, Dawn Song, Matt Fredrikson, J. Zico Kolter, Dan Hendrycks
Abstract: In this paper, we identify and characterize the emerging area of representation engineering (RepE), an approach to enhancing the transparency of AI systems that draws on insights from cognitive neuroscience. RepE places population-level representations, rather than neurons or circuits, at the center of analysis, equipping us with novel methods for monitoring and manipulating high-level cognitive phenomena in deep neural networks (DNNs). We provide baselines and an initial analysis of RepE techniques, showing that they offer simple yet effective solutions for improving our understanding and control of large language models. We showcase how these methods can provide traction on a wide range of safety-relevant problems, including honesty, harmlessness, power-seeking, and more, demonstrating the promise of top-down transparency research. We hope that this work catalyzes further exploration of RepE and fosters advancements in the transparency and safety of AI systems.
Authors: Xu Ouyang, Felix Xiaozhu Lin, Yangfeng Ji
Abstract: Critical to a free data market is $\textit{private data selection}$, i.e. the model owner selects and then appraises training data from the data owner before both parties commit to a transaction. To keep the data and model private, this process shall evaluate the target model to be trained over Multi-Party Computation (MPC). While prior work suggests that evaluating Transformer-based models over MPC is prohibitively expensive, this paper makes it practical for the purpose of data selection. Our contributions are three: (1) a new pipeline for private data selection over MPC; (2) emulating high-dimensional nonlinear operators with low-dimension MLPs, which are trained on a small sample of the data of interest; (3) scheduling MPC in a parallel, multiphase fashion. We evaluate our method on diverse Transformer models and NLP/CV benchmarks. Compared to directly evaluating the target model over MPC, our method reduces the delay from thousands of hours to tens of hours, while only seeing around 0.20% accuracy degradation from training with the selected data.
Authors: Yongxin Guo, Xiaoying Tang, Tao Lin
Abstract: Federated Learning (FL) is an evolving distributed machine learning approach that safeguards client privacy by keeping data on edge devices. However, the variation in data among clients poses challenges in training models that excel across all local distributions. Recent studies suggest clustering as a solution to address client heterogeneity in FL by grouping clients with distribution shifts into distinct clusters. Nonetheless, the diverse learning frameworks used in current clustered FL methods create difficulties in integrating these methods, leveraging their advantages, and making further enhancements. To this end, this paper conducts a thorough examination of existing clustered FL methods and introduces a four-tier framework, named HCFL, to encompass and extend the existing approaches. Utilizing the HCFL, we identify persistent challenges associated with current clustering methods in each tier and propose an enhanced clustering method called HCFL$^{+}$ to overcome these challenges. Through extensive numerical evaluations, we demonstrate the effectiveness of our clustering framework and the enhanced components. Our code is available at https://github.com/LINs-lab/HCFL.
Authors: Thomas Chen
Abstract: We consider the scenario of supervised learning in Deep Learning (DL) networks, and exploit the arbitrariness of choice in the Riemannian metric relative to which the gradient descent flow can be defined (a general fact of differential geometry). In the standard approach to DL, the gradient flow on the space of parameters (weights and biases) is defined with respect to the Euclidean metric. Here instead, we choose the gradient flow with respect to the Euclidean metric in the output layer of the DL network. This naturally induces two modified versions of the gradient descent flow in the parameter space, one adapted for the overparametrized setting, and the other for the underparametrized setting. In the overparametrized case, we prove that, provided that a rank condition holds, all orbits of the modified gradient descent drive the ${\mathcal L}^2$ cost to its global minimum at a uniform exponential convergence rate; one thereby obtains an a priori stopping time for any prescribed proximity to the global minimum. We point out relations of the latter to sub-Riemannian geometry. Moreover, we generalize the above framework to the situation in which the rank condition does not hold; in particular, we show that local equilibria can only exist if a rank loss occurs, and that generically, they are not isolated points, but elements of a critical submanifold of parameter space.
Authors: Tingting Ni, Maryam Kamgarpour
Abstract: We consider discounted infinite-horizon constrained Markov decision processes (CMDPs), where the goal is to find an optimal policy that maximizes the expected cumulative reward while satisfying expected cumulative constraints. Motivated by the application of CMDPs in online learning for safety-critical systems, we focus on developing a model-free and \emph{simulator-free} algorithm that ensures \emph{constraint satisfaction during learning}. To this end, we employ the LB-SGD algorithm proposed in \cite{usmanova2022log}, which utilizes an interior-point approach based on the log-barrier function of the CMDP. Under the commonly assumed conditions of relaxed Fisher non-degeneracy and bounded transfer error in policy parameterization, we establish the theoretical properties of the LB-SGD algorithm. In particular, unlike existing CMDP approaches that ensure policy feasibility only upon convergence, the LB-SGD algorithm guarantees feasibility throughout the learning process and converges to the $\varepsilon$-optimal policy with a sample complexity of $\tilde{\mathcal{O}}(\varepsilon^{-6})$. Compared to the state-of-the-art policy gradient-based algorithm, C-NPG-PDA \cite{bai2022achieving2}, the LB-SGD algorithm requires an additional $\mathcal{O}(\varepsilon^{-2})$ samples to ensure policy feasibility during learning with the same Fisher non-degenerate parameterization.
Authors: Asela Hevapathige, Qing Wang
Abstract: Graph Neural Networks (GNNs) have paved the way for being a cornerstone in graph-related learning tasks. Yet, the ability of GNNs to capture structural interactions within graphs remains under-explored. In this work, we address this gap by drawing on the insight that permutation invariant graph partitioning enables a powerful way of exploring structural interactions. We establish theoretical connections between permutation invariant graph partitioning and graph isomorphism, and then propose Graph Partitioning Neural Networks (GPNNs), a novel architecture that efficiently enhances the expressive power of GNNs in learning structural interactions. We analyze how partitioning schemes and structural interactions contribute to GNN expressivity and their trade-offs with complexity. Empirically, we demonstrate that GPNNs outperform existing GNN models in capturing structural interactions across diverse graph benchmark tasks.
Authors: Yun Young Choi, Sun Woo Park, Minho Lee, Youngho Woo
Abstract: Transformers have revolutionized performance in Natural Language Processing and Vision, paving the way for their integration with Graph Neural Networks (GNNs). One key challenge in enhancing graph transformers is strengthening the discriminative power of distinguishing isomorphisms of graphs, which plays a crucial role in boosting their predictive performances. To address this challenge, we introduce 'Topology-Informed Graph Transformer (TIGT)', a novel transformer enhancing both discriminative power in detecting graph isomorphisms and the overall performance of Graph Transformers. TIGT consists of four components: A topological positional embedding layer using non-isomorphic universal covers based on cyclic subgraphs of graphs to ensure unique graph representation: A dual-path message-passing layer to explicitly encode topological characteristics throughout the encoder layers: A global attention mechanism: And a graph information layer to recalibrate channel-wise graph features for better feature representation. TIGT outperforms previous Graph Transformers in classifying synthetic dataset aimed at distinguishing isomorphism classes of graphs. Additionally, mathematical analysis and empirical evaluations highlight our model's competitive edge over state-of-the-art Graph Transformers across various benchmark datasets.
Authors: Bohan Tang, Zexi Liu, Keyue Jiang, Siheng Chen, Xiaowen Dong
Abstract: Hypergraphs are crucial for modelling higher-order interactions in real-world data. Hypergraph neural networks (HNNs) effectively utilise these structures by message passing to generate informative node features for various downstream tasks like node classification. However, the message passing module in existing HNNs typically requires a computationally intensive training process, which limits their practical use. To tackle this challenge, we propose an alternative approach by decoupling the usage of hypergraph structural information from the model learning stage. This leads to a novel training-free message passing module, named TF-MP-Module, which can be precomputed in the data preprocessing stage, thereby reducing the computational burden. We refer to the hypergraph neural network equipped with our TF-MP-Module as TF-HNN. We theoretically support the efficiency and effectiveness of TF-HNN by showing that: 1) It is more training-efficient compared to existing HNNs; 2) It utilises as much information as existing HNNs for node feature generation; and 3) It is robust against the oversmoothing issue while using long-range interactions. Experiments based on seven real-world hypergraph benchmarks in node classification and hyperlink prediction show that, compared to state-of-the-art HNNs, TF-HNN exhibits both competitive performance and superior training efficiency. Specifically, on the large-scale benchmark, Trivago, TF-HNN outperforms the node classification accuracy of the best baseline by 10% with just 1% of the training time of that baseline.
Authors: Clara Punzi, Roberto Pellungrini, Mattia Setzu, Fosca Giannotti, Dino Pedreschi
Abstract: Everyday we increasingly rely on machine learning models to automate and support high-stake tasks and decisions. This growing presence means that humans are now constantly interacting with machine learning-based systems, training and using models everyday. Several different techniques in computer science literature account for the human interaction with machine learning systems, but their classification is sparse and the goals varied. This survey proposes a taxonomy of Hybrid Decision Making Systems, providing both a conceptual and technical framework for understanding how current computer science literature models interaction between humans and machines.
Authors: Marios Papachristou, M. Amin Rahimian
Abstract: How can agents exchange information to learn while protecting privacy? Healthcare centers collaborating on clinical trials must balance knowledge sharing with safeguarding sensitive patient data. We address this challenge by using differential privacy (DP) to control information leakage. Agents update belief statistics via log-linear rules, and DP noise provides plausible deniability and rigorous performance guarantees. We study two settings: distributed maximum likelihood estimation (MLE) with a finite set of private signals and online learning from an intermittent signal stream. Noisy aggregation introduces trade-offs between rejecting low-quality states and accepting high-quality ones. The MLE setting naturally applies to binary hypothesis testing with formal statistical guarantees. Through simulations, we demonstrate differentially private, distributed survival analysis on real-world clinical trial data, evaluating treatment efficacy and the impact of biomedical indices on patient survival. Our methods enable privacy-preserving inference with greater efficiency and lower error rates than homomorphic encryption and first-order DP optimization approaches.
Authors: Andrzej Mizera, Jakub Zarzycki
Abstract: Cellular reprogramming can be used for both the prevention and cure of different diseases. However, the efficiency of discovering reprogramming strategies with classical wet-lab experiments is hindered by lengthy time commitments and high costs. In this study, we develop a novel computational framework based on deep reinforcement learning that facilitates the identification of reprogramming strategies. For this aim, we formulate a control problem in the context of cellular reprogramming for the frameworks of BNs and PBNs under the asynchronous update mode. Furthermore, we introduce the notion of a pseudo-attractor and a procedure for identification of pseudo-attractor state during training. Finally, we devise a computational framework for solving the control problem, which we test on a number of different models.
Authors: Simon Geisler, Tom Wollschl\"ager, M. H. I. Abdalla, Johannes Gasteiger, Stephan G\"unnemann
Abstract: Current LLM alignment methods are readily broken through specifically crafted adversarial prompts. While crafting adversarial prompts using discrete optimization is highly effective, such attacks typically use more than 100,000 LLM calls. This high computational cost makes them unsuitable for, e.g., quantitative analyses and adversarial training. To remedy this, we revisit Projected Gradient Descent (PGD) on the continuously relaxed input prompt. Although previous attempts with ordinary gradient-based attacks largely failed, we show that carefully controlling the error introduced by the continuous relaxation tremendously boosts their efficacy. Our PGD for LLMs is up to one order of magnitude faster than state-of-the-art discrete optimization to achieve the same devastating attack results.
Authors: Mingyu Guan, Jack W. Stokes, Qinlong Luo, Fuchen Liu, Purvanshi Mehta, Elnaz Nouri, Taesoo Kim
Abstract: The recent past has seen an increasing interest in Heterogeneous Graph Neural Networks (HGNNs), since many real-world graphs are heterogeneous in nature, from citation graphs to email graphs. However, existing methods ignore a tree hierarchy among metapaths, naturally constituted by different node types and relation types. In this paper, we present HetTree, a novel HGNN that models both the graph structure and heterogeneous aspects in a scalable and effective manner. Specifically, HetTree builds a semantic tree data structure to capture the hierarchy among metapaths. To effectively encode the semantic tree, HetTree uses a novel subtree attention mechanism to emphasize metapaths that are more helpful in encoding parent-child relationships. Moreover, HetTree proposes carefully matching pre-computed features and labels correspondingly, constituting a complete metapath representation. Our evaluation of HetTree on a variety of real-world datasets demonstrates that it outperforms all existing baselines on open benchmarks and efficiently scales to large real-world graphs with millions of nodes and edges.
Authors: Georg Manten, Cecilia Casolo, Emilio Ferrucci, S{\o}ren Wengel Mogensen, Cristopher Salvi, Niki Kilbertus
Abstract: Inferring the causal structure underlying stochastic dynamical systems from observational data holds great promise in domains ranging from science and health to finance. Such processes can often be accurately modeled via stochastic differential equations (SDEs), which naturally imply causal relationships via "which variables enter the differential of which other variables". In this paper, we develop conditional independence (CI) constraints on coordinate processes over selected intervals that are Markov with respect to the acyclic dependence graph (allowing self-loops) induced by a general SDE model. We then provide a sound and complete causal discovery algorithm, capable of handling both fully and partially observed data, and uniquely recovering the underlying or induced ancestral graph by exploiting time directionality assuming a CI oracle. Finally, to make our algorithm practically usable, we also propose a flexible, consistent signature kernel-based CI test to infer these constraints from data. We extensively benchmark the CI test in isolation and as part of our causal discovery algorithms, outperforming existing approaches in SDE models and beyond.
Authors: Th\'eo Vincent, Daniel Palenicek, Boris Belousov, Jan Peters, Carlo D'Eramo
Abstract: The vast majority of Reinforcement Learning methods is largely impacted by the computation effort and data requirements needed to obtain effective estimates of action-value functions, which in turn determine the quality of the overall performance and the sample-efficiency of the learning procedure. Typically, action-value functions are estimated through an iterative scheme that alternates the application of an empirical approximation of the Bellman operator and a subsequent projection step onto a considered function space. It has been observed that this scheme can be potentially generalized to carry out multiple iterations of the Bellman operator at once, benefiting the underlying learning algorithm. However, till now, it has been challenging to effectively implement this idea, especially in high-dimensional problems. In this paper, we introduce iterated $Q$-Network (i-QN), a novel principled approach that enables multiple consecutive Bellman updates by learning a tailored sequence of action-value functions where each serves as the target for the next. We show that i-QN is theoretically grounded and that it can be seamlessly used in value-based and actor-critic methods. We empirically demonstrate the advantages of i-QN in Atari $2600$ games and MuJoCo continuous control problems.
Authors: Benedikt Fesl, Benedikt B\"ock, Florian Strasser, Michael Baur, Michael Joham, Wolfgang Utschick
Abstract: Diffusion models (DMs) as generative priors have recently shown great potential for denoising tasks but lack theoretical understanding with respect to their mean square error (MSE) optimality. This paper proposes a novel denoising strategy inspired by the structure of the MSE-optimal conditional mean estimator (CME). The resulting DM-based denoiser can be conveniently employed using a pre-trained DM, being particularly fast by truncating reverse diffusion steps and not requiring stochastic re-sampling. We present a comprehensive (non-)asymptotic optimality analysis of the proposed diffusion-based denoiser, demonstrating polynomial-time convergence to the CME under mild conditions. Our analysis also derives a novel Lipschitz constant that depends solely on the DM's hyperparameters. Further, we offer a new perspective on DMs, showing that they inherently combine an asymptotically optimal denoiser with a powerful generator, modifiable by switching re-sampling in the reverse process on or off. The theoretical findings are thoroughly validated with experiments based on various benchmark datasets
Authors: Titouan Renard, Andreas Schlaginhaufen, Tingting Ni, Maryam Kamgarpour
Abstract: Given a dataset of expert demonstrations, inverse reinforcement learning (IRL) aims to recover a reward for which the expert is optimal. This work proposes a model-free algorithm to solve entropy-regularized IRL problem. In particular, we employ a stochastic gradient descent update for the reward and a stochastic soft policy iteration update for the policy. Assuming access to a generative model, we prove that our algorithm is guaranteed to recover a reward for which the expert is $\varepsilon$-optimal using $\mathcal{O}(1/\varepsilon^{2})$ samples of the Markov decision process (MDP). Furthermore, with $\mathcal{O}(1/\varepsilon^{4})$ samples we prove that the optimal policy corresponding to the recovered reward is $\varepsilon$-close to the expert policy in total variation distance.
Authors: Liangchen Li, Jiajun He
Abstract: Diffusion models (DMs) are capable of generating remarkably high-quality samples by iteratively denoising a random vector, a process that corresponds to moving along the probability flow ordinary differential equation (PF ODE). Interestingly, DMs can also invert an input image to noise by moving backward along the PF ODE, a key operation for downstream tasks such as interpolation and image editing. However, the iterative nature of this process restricts its speed, hindering its broader application. Recently, Consistency Models (CMs) have emerged to address this challenge by approximating the integral of the PF ODE, largely reducing the number of iterations. Yet, the absence of an explicit ODE solver complicates the inversion process. To resolve this, we introduce Bidirectional Consistency Model (BCM), which learns a single neural network that enables both forward and backward traversal along the PF ODE, efficiently unifying generation and inversion tasks within one framework. We can train BCM from scratch or tune it using a pretrained consistency model, which reduces the training cost and increases scalability. We demonstrate that BCM enables one-step generation and inversion while also allowing the use of additional steps to enhance generation quality or reduce reconstruction error. We further showcase BCM's capability in downstream tasks, such as interpolation and inpainting. Our code and weights are available at https://github.com/Mosasaur5526/BCM-iCT-torch.
Authors: Yiping Ji, Hemanth Saratchandran, Cameron Gordon, Zeyu Zhang, Simon Lucey
Abstract: Low-rank decomposition has emerged as a vital tool for enhancing parameter efficiency in neural network architectures, gaining traction across diverse applications in machine learning. These techniques significantly lower the number of parameters, striking a balance between compactness and performance. However, a common challenge has been the compromise between parameter efficiency and the accuracy of the model, where reduced parameters often lead to diminished accuracy compared to their full-rank counterparts. In this work, we propose a novel theoretical framework that integrates a sinusoidal function within the low-rank decomposition process. This approach not only preserves the benefits of the parameter efficiency characteristic of low-rank methods but also increases the decomposition's rank, thereby enhancing model performance. Our method proves to be a plug in enhancement for existing low-rank models, as evidenced by its successful application in Vision Transformers (ViT), Large Language Models (LLMs), Neural Radiance Fields (NeRF) and 3D shape modelling.
Authors: Soichiro Nishimori, Xin-Qiang Cai, Johannes Ackermann, Masashi Sugiyama
Abstract: Offline reinforcement learning (RL) is vital in areas where active data collection is expensive or infeasible, such as robotics or healthcare. In the real world, offline datasets often involve multiple domains that share the same state and action spaces but have distinct dynamics, and only a small fraction of samples are clearly labeled as belonging to the target domain we are interested in. For example, in robotics, precise system identification may only have been performed for part of the deployments. To address this challenge, we consider Positive-Unlabeled Offline RL (PUORL), a novel offline RL setting in which we have a small amount of labeled target-domain data and a large amount of domain-unlabeled data from multiple domains, including the target domain. For PUORL, we propose a plug-and-play approach that leverages positive-unlabeled (PU) learning to train a domain classifier. The classifier then extracts target-domain samples from the domain-unlabeled data, augmenting the scarce target-domain data. Empirical results on a modified version of the D4RL benchmark demonstrate the effectiveness of our method: even when only 1 to 3 percent of the dataset is domain-labeled, our approach accurately identifies target-domain samples and achieves high performance, even under substantial dynamics shift. Our plug-and-play algorithm seamlessly integrates PU learning with existing offline RL pipelines, enabling effective multi-domain data utilization in scenarios where comprehensive domain labeling is prohibitive.
Authors: Dipika Jha, Ankit K. Bhagat, Raju Halder, Rajendra N. Paramanik, Chandra M. Kumar
Abstract: This paper presents a comprehensive Decentraland parcels dataset, called IITP-VDLand, sourced from diverse platforms such as Decentraland, OpenSea, Etherscan, Google BigQuery, and various Social Media Platforms. Unlike existing datasets which have limited attributes and records, IITP-VDLand offers a rich array of attributes, encompassing parcel characteristics, trading history, past activities, transactions, and social media interactions. Alongside, we introduce a key attribute in the dataset, namely Rarity score, which measures the uniqueness of each parcel within the virtual world. Addressing the significant challenge posed by the dispersed nature of this data across various sources, we employ a systematic approach, utilizing both available APIs and custom scripts, to gather it. Subsequently, we meticulously curate and organize the information into four distinct fragments: (1) Characteristics, (2) OpenSea Trading History, (3) Ethereum Activity Transactions, and (4) Social Media. We envisage that this dataset would serve as a robust resource for training machine- and deep-learning models specifically designed to address real-world challenges within the domain of Decentraland parcels. The performance benchmarking of more than 20 state-of-the-art price prediction models on our dataset yields promising results, achieving a maximum R2 score of 0.8251 and an accuracy of 74.23% in case of Extra Trees Regressor and Classifier. The key findings reveal that the ensemble models perform better than both deep learning and linear models for our dataset. We observe a significant impact of coordinates, geographical proximity, rarity score, and few other economic indicators on the prediction of parcel prices.
Authors: Hao-Lun Hsu, Weixin Wang, Miroslav Pajic, Pan Xu
Abstract: We present the first study on provably efficient randomized exploration in cooperative multi-agent reinforcement learning (MARL). We propose a unified algorithm framework for randomized exploration in parallel Markov Decision Processes (MDPs), and two Thompson Sampling (TS)-type algorithms, CoopTS-PHE and CoopTS-LMC, incorporating the perturbed-history exploration (PHE) strategy and the Langevin Monte Carlo exploration (LMC) strategy, respectively, which are flexible in design and easy to implement in practice. For a special class of parallel MDPs where the transition is (approximately) linear, we theoretically prove that both CoopTS-PHE and CoopTS-LMC achieve a $\widetilde{\mathcal{O}}(d^{3/2}H^2\sqrt{MK})$ regret bound with communication complexity $\widetilde{\mathcal{O}}(dHM^2)$, where $d$ is the feature dimension, $H$ is the horizon length, $M$ is the number of agents, and $K$ is the number of episodes. This is the first theoretical result for randomized exploration in cooperative MARL. We evaluate our proposed method on multiple parallel RL environments, including a deep exploration problem (i.e., $N$-chain), a video game, and a real-world problem in energy systems. Our experimental results support that our framework can achieve better performance, even under conditions of misspecified transition models. Additionally, we establish a connection between our unified framework and the practical application of federated learning.
Authors: Ted Edward Holmberg, Mahdi Abdelguerfi, Elias Ioup
Abstract: Spatiotemporal networks' observational capabilities are crucial for accurate data gathering and informed decisions across multiple sectors. This study focuses on the Spatiotemporal Ranged Observer-Observable Bipartite Network (STROOBnet), linking observational nodes (e.g., surveillance cameras) to events within defined geographical regions, enabling efficient monitoring. Using data from Real-Time Crime Camera (RTCC) systems and Calls for Service (CFS) in New Orleans, where RTCC combats rising crime amidst reduced police presence, we address the network's initial observational imbalances. Aiming for uniform observational efficacy, we propose the Proximal Recurrence approach. It outperformed traditional clustering methods like k-means and DBSCAN by offering holistic event frequency and spatial consideration, enhancing observational coverage.
Authors: Toshimitsu Uesaka, Taiji Suzuki, Yuhta Takida, Chieh-Hsin Lai, Naoki Murata, Yuki Mitsufuji
Abstract: In typical multimodal contrastive learning, such as CLIP, encoders produce one point in the latent representation space for each input. However, one-point representation has difficulty in capturing the relationship and the similarity structure of a huge amount of instances in the real world. For richer classes of the similarity, we propose the use of weighted point sets, namely, sets of pairs of weight and vector, as representations of instances. In this work, we theoretically show the benefit of our proposed method through a new understanding of the contrastive loss of CLIP, which we call symmetric InfoNCE. We clarify that the optimal similarity that minimizes symmetric InfoNCE is the pointwise mutual information, and show an upper bound of excess risk on downstream classification tasks of representations that achieve the optimal similarity. In addition, we show that our proposed similarity based on weighted point sets consistently achieves the optimal similarity. To verify the effectiveness of our proposed method, we demonstrate pretraining of text-image representation models and classification tasks on common benchmarks.
Authors: Yihao Zhang, Zeming Wei
Abstract: Large Language Models (LLMs) have achieved remarkable success across diverse tasks, yet they remain vulnerable to adversarial attacks, notably the well-known jailbreak attack. In particular, the Greedy Coordinate Gradient (GCG) attack has demonstrated efficacy in exploiting this vulnerability by optimizing adversarial prompts through a combination of gradient heuristics and greedy search. However, the efficiency of this attack has become a bottleneck in the attacking process. To mitigate this limitation, in this paper we rethink the generation of the adversarial prompts through an optimization lens, aiming to stabilize the optimization process and harness more heuristic insights from previous optimization iterations. Specifically, we propose the \textbf{M}omentum \textbf{A}ccelerated G\textbf{C}G (\textbf{MAC}) attack, which integrates a momentum term into the gradient heuristic to boost and stabilize the random search for tokens in adversarial prompts. Experimental results showcase the notable enhancement achieved by MAC over baselines in terms of attack success rate and optimization efficiency. Moreover, we demonstrate that MAC can still exhibit superior performance for transfer attacks and models under defense mechanisms. Our code is available at https://github.com/weizeming/momentum-attack-llm.
Authors: Shangding Gu, Bilgehan Sel, Yuhao Ding, Lu Wang, Qingwei Lin, Ming Jin, Alois Knoll
Abstract: Ensuring the safety of Reinforcement Learning (RL) is crucial for its deployment in real-world applications. Nevertheless, managing the trade-off between reward and safety during exploration presents a significant challenge. Improving reward performance through policy adjustments may adversely affect safety performance. In this study, we aim to address this conflicting relation by leveraging the theory of gradient manipulation. Initially, we analyze the conflict between reward and safety gradients. Subsequently, we tackle the balance between reward and safety optimization by proposing a soft switching policy optimization method, for which we provide convergence analysis. Based on our theoretical examination, we provide a safe RL framework to overcome the aforementioned challenge, and we develop a Safety-MuJoCo Benchmark to assess the performance of safe RL algorithms. Finally, we evaluate the effectiveness of our method on the Safety-MuJoCo Benchmark and a popular safe RL benchmark, Omnisafe. Experimental results demonstrate that our algorithms outperform several state-of-the-art baselines in terms of balancing reward and safety optimization.
Authors: Sixin Zhang
Abstract: We study min-max algorithms to solve zero-sum differential games on Riemannian manifold. Based on the notions of differential Stackelberg equilibrium and differential Nash equilibrium on Riemannian manifold, we analyze the local convergence of two representative deterministic simultaneous algorithms $\tau$-GDA and $\tau$-SGA to such equilibria. Sufficient conditions are obtained to establish the linear convergence rate of $\tau$-GDA based on the Ostrowski theorem on manifold and spectral analysis. To avoid strong rotational dynamics in $\tau$-GDA, $\tau$-SGA is extended from the symplectic gradient-adjustment method in Euclidean space. We analyze an asymptotic approximation of $\tau$-SGA when the learning rate ratio $\tau$ is big. In some cases, it can achieve a faster convergence rate to differential Stackelberg equilibrium compared to $\tau$-GDA. We show numerically how the insights obtained from the convergence analysis may improve the training of orthogonal Wasserstein GANs using stochastic $\tau$-GDA and $\tau$-SGA on simple benchmarks.
Authors: Haoxian Chen, Hanyang Zhao, Henry Lam, David Yao, Wenpin Tang
Abstract: Direct Preference Optimization (DPO) has recently emerged as a popular approach to improve reinforcement learning with human feedback (RLHF), leading to better techniques to fine-tune large language models (LLM). A weakness of DPO, however, lies in its lack of capability to characterize the diversity of human preferences. Inspired by Mallows' theory of preference ranking, we develop in this paper a new approach, the MallowsPO. A distinct feature of this approach is a dispersion index, which reflects the dispersion of human preference to prompts. We show that existing DPO models can be reduced to special cases of this dispersion index, thus unified with MallowsPO. More importantly, we demonstrate (empirically) how to use this dispersion index to enhance the performance of DPO in a broad array of benchmark tasks, from synthetic bandit selection to controllable generations and dialogues, while maintaining great generalization capabilities. MallowsPO is also compatible with other SOTA offline preference optimization methods, boosting nearly 2\% extra LC win rate when used as a plugin for fine-tuning Llama3-Instruct.
Authors: Haokai Hong, Wanyu Lin, Kay Chen Tan
Abstract: This paper proposes a new 3D molecule generation framework, called GOAT, for fast and effective 3D molecule generation based on the flow-matching optimal transport objective. Specifically, we formulate a geometric transport formula for measuring the cost of mapping multi-modal features (e.g., continuous atom coordinates and categorical atom types) between a base distribution and a target data distribution. Our formula is solved within a joint, equivariant, and smooth representation space. This is achieved by transforming the multi-modal features into a continuous latent space with equivariant networks. In addition, we find that identifying optimal distributional coupling is necessary for fast and effective transport between any two distributions. We further propose a mechanism for estimating and purifying optimal coupling to train the flow model with optimal transport. By doing so, GOAT can turn arbitrary distribution couplings into new deterministic couplings, leading to an estimated optimal transport plan for fast 3D molecule generation. The purification filters out the subpar molecules to ensure the ultimate generation quality. We theoretically and empirically prove that the proposed optimal coupling estimation and purification yield transport plan with non-increasing cost. Finally, extensive experiments show that GOAT enjoys the efficiency of solving geometric optimal transport, leading to a double speedup compared to the sub-optimal method while achieving the best generation quality regarding validity, uniqueness, and novelty. The code is available at https://github.com/WanyuGroup/ICLR2025-GOAT.
Authors: Qichao Shentu, Beibu Li, Kai Zhao, Yang Shu, Zhongwen Rao, Lujia Pan, Bin Yang, Chenjuan Guo
Abstract: Time series anomaly detection plays a vital role in a wide range of applications. Existing methods require training one specific model for each dataset, which exhibits limited generalization capability across different target datasets, hindering anomaly detection performance in various scenarios with scarce training data. Aiming at this problem, we propose constructing a general time series anomaly detection model, which is pre-trained on extensive multi-domain datasets and can subsequently apply to a multitude of downstream scenarios. The significant divergence of time series data across different domains presents two primary challenges in building such a general model: (1) meeting the diverse requirements of appropriate information bottlenecks tailored to different datasets in one unified model, and (2) enabling distinguishment between multiple normal and abnormal patterns, both are crucial for effective anomaly detection in various target scenarios. To tackle these two challenges, we propose a General time series anomaly Detector with Adaptive Bottlenecks and Dual Adversarial Decoders (DADA), which enables flexible selection of bottlenecks based on different data and explicitly enhances clear differentiation between normal and abnormal series. We conduct extensive experiments on nine target datasets from different domains. After pre-training on multi-domain data, DADA, serving as a zero-shot anomaly detector for these datasets, still achieves competitive or even superior results compared to those models tailored to each specific dataset. The code is made available at https://github.com/decisionintelligence/DADA.
Authors: Md Abrar Jahin, Asef Shahriar, Md Al Amin
Abstract: Accurate demand forecasting is crucial for optimizing supply chain management. Traditional methods often fail to capture complex patterns from seasonal variability and special events. Despite advancements in deep learning, interpretable forecasting models remain a challenge. To address this, we introduce the Multi-Channel Data Fusion Network (MCDFN), a hybrid architecture that integrates Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and Gated Recurrent Units (GRU) to enhance predictive performance by extracting spatial and temporal features from time series data. Our comparative benchmarking demonstrates that MCDFN outperforms seven other deep-learning models, achieving superior metrics: MSE (23.5738), RMSE (4.8553), MAE (3.9991), and MAPE (20.1575%). Theil's U statistic of 0.1181 (U<1) of MCDFN indicates its superiority over the naive forecasting approach, and a 10-fold cross-validated statistical paired t-test with a p-value of 5% indicated no significant difference between MCDFN's predictions and actual values. We apply explainable AI techniques like ShapTime and Permutation Feature Importance to enhance interpretability. This research advances demand forecasting methodologies and offers practical guidelines for integrating MCDFN into supply chain systems, highlighting future research directions for scalability and user-friendly deployment.
Authors: Minhak Song, Kwangjun Ahn, Chulhee Yun
Abstract: Understanding the training dynamics of deep neural networks is challenging due to their high-dimensional nature and intricate loss landscapes. Recent studies have revealed that, along the training trajectory, the gradient approximately aligns with a low-rank top eigenspace of the training loss Hessian, referred to as the dominant subspace. Given this alignment, this paper explores whether neural networks can be trained within the dominant subspace, which, if feasible, could lead to more efficient training methods. Our primary observation is that when the SGD update is projected onto the dominant subspace, the training loss does not decrease further. This suggests that the observed alignment between the gradient and the dominant subspace is spurious. Surprisingly, projecting out the dominant subspace proves to be just as effective as the original update, despite removing the majority of the original update component. We observe similar behavior across practical setups, including the large learning rate regime (also known as Edge of Stability), Sharpness-Aware Minimization, momentum, and adaptive optimizers. We discuss the main causes and implications of this spurious alignment, shedding light on the dynamics of neural network training.
Authors: Th\'eo Vincent, Fabian Wahren, Jan Peters, Boris Belousov, Carlo D'Eramo
Abstract: Deep Reinforcement Learning (RL) is well known for being highly sensitive to hyperparameters, requiring practitioners substantial efforts to optimize them for the problem at hand. This also limits the applicability of RL in real-world scenarios. In recent years, the field of automated Reinforcement Learning (AutoRL) has grown in popularity by trying to address this issue. However, these approaches typically hinge on additional samples to select well-performing hyperparameters, hindering sample-efficiency and practicality. Furthermore, most AutoRL methods are heavily based on already existing AutoML methods, which were originally developed neglecting the additional challenges inherent to RL due to its non-stationarities. In this work, we propose a new approach for AutoRL, called Adaptive $Q$-Network (AdaQN), that is tailored to RL to take into account the non-stationarity of the optimization procedure without requiring additional samples. AdaQN learns several $Q$-functions, each one trained with different hyperparameters, which are updated online using the $Q$-function with the smallest approximation error as a shared target. Our selection scheme simultaneously handles different hyperparameters while coping with the non-stationarity induced by the RL optimization procedure and being orthogonal to any critic-based RL algorithm. We demonstrate that AdaQN is theoretically sound and empirically validate it in MuJoCo control problems and Atari $2600$ games, showing benefits in sample-efficiency, overall performance, robustness to stochasticity and training stability.
Authors: Nikola Zubi\'c, Federico Sold\'a, Aurelio Sulser, Davide Scaramuzza
Abstract: Despite their successes, deep learning models struggle with tasks requiring complex reasoning and function composition. We present a theoretical and empirical investigation into the limitations of Structured State Space Models (SSMs) and Transformers in such tasks. We prove that one-layer SSMs cannot efficiently perform function composition over large domains without impractically large state sizes, and even with Chain-of-Thought prompting, they require a number of steps that scale unfavorably with the complexity of the function composition. Also, the language of a finite-precision SSM is within the class of regular languages. Our experiments corroborate these theoretical findings. Evaluating models on tasks including various function composition settings, multi-digit multiplication, dynamic programming, and Einstein's puzzle, we find significant performance degradation even with advanced prompting techniques. Models often resort to shortcuts, leading to compounding errors. These findings highlight fundamental barriers within current deep learning architectures rooted in their computational capacities. We underscore the need for innovative solutions to transcend these constraints and achieve reliable multi-step reasoning and compositional task-solving, which is critical for advancing toward general artificial intelligence.
Authors: Adam Fisch, Jacob Eisenstein, Vicky Zayats, Alekh Agarwal, Ahmad Beirami, Chirag Nagpal, Pete Shaw, Jonathan Berant
Abstract: Language model (LM) post-training (or alignment) involves maximizing a reward function that is derived from preference annotations. Direct Preference Optimization (DPO) is a popular offline alignment method that trains a policy directly on preference data without the need to train a reward model or apply reinforcement learning. However, the empirical evidence suggests that DPO typically assigns implicit rewards that overfit, and trend towards infinite magnitude. This frequently leads to degenerate policies, sometimes causing even the probabilities of the preferred generations to go to zero. In this work, we analyze this phenomenon and use distillation to get a better proxy for the true preference distribution over generation pairs: we train the LM such that its induced implicit reward, i.e., the scaled log-likelihood ratio of the model to the reference model, matches an explicit reward model trained on the preference data. Moreover, to account for uncertainty in the reward model we are distilling from, we optimize against a family of reward models that, as a whole, is likely to include at least one reasonable proxy for the preference distribution. Our results show that distilling from such a family of reward models leads to improved robustness to distribution shift in preference annotations, while preserving the simple supervised nature of DPO.
Authors: Linlin Yu, Bowen Yang, Tianhao Wang, Kangshuo Li, Feng Chen
Abstract: The fusion of raw sensor data to create a Bird's Eye View (BEV) representation is critical for autonomous vehicle planning and control. Despite the growing interest in using deep learning models for BEV semantic segmentation, anticipating segmentation errors and enhancing the explainability of these models remain underexplored. This paper introduces a comprehensive benchmark for predictive uncertainty quantification in BEV segmentation, evaluating multiple uncertainty quantification methods across three popular datasets with three representative network architectures. Our study focuses on the effectiveness of quantified uncertainty in detecting misclassified and out-of-distribution (OOD) pixels while also improving model calibration. Through empirical analysis, we uncover challenges in existing uncertainty quantification methods and demonstrate the potential of evidential deep learning techniques, which capture both aleatoric and epistemic uncertainty. To address these challenges, we propose a novel loss function, Uncertainty-Focal-Cross-Entropy (UFCE), specifically designed for highly imbalanced data, along with a simple uncertainty-scaling regularization term that improves both uncertainty quantification and model calibration for BEV segmentation.
Authors: Wenjing Chang, Kay Liu, Philip S. Yu, Jianjun Yu
Abstract: Graph anomaly detection (GAD) is increasingly crucial in various applications, ranging from financial fraud detection to fake news detection. However, current GAD methods largely overlook the fairness problem, which might result in discriminatory decisions skewed toward certain demographic groups defined on sensitive attributes (e.g., gender, religion, ethnicity, etc.). This greatly limits the applicability of these methods in real-world scenarios in light of societal and ethical restrictions. To address this critical gap, we make the first attempt to integrate fairness with utility in GAD decision-making. Specifically, we devise a novel DisEntangle-based FairnEss-aware aNomaly Detection framework on the attributed graph, named DEFEND. DEFEND first introduces disentanglement in GNNs to capture informative yet sensitive-irrelevant node representations, effectively reducing societal bias inherent in graph representation learning. Besides, to alleviate discriminatory bias in evaluating anomalous nodes, DEFEND adopts a reconstruction-based anomaly detection, which concentrates solely on node attributes without incorporating any graph structure. Additionally, given the inherent association between input and sensitive attributes, DEFEND constrains the correlation between the reconstruction error and the predicted sensitive attributes. Our empirical evaluations on real-world datasets reveal that DEFEND performs effectively in GAD and significantly enhances fairness compared to state-of-the-art baselines. To foster reproducibility, our code is available at https://github.com/AhaChang/DEFEND.
Authors: Yutao Sun, Mingshuai Chen, Kangjia Zhao, Jintao Chen
Abstract: Artificial intelligence is rapidly encroaching on the field of service regulation. This work-in-progress article presents the design principles behind HORAE, a unified specification language to model multimodal regulation rules across a diverse set of domains. We show how HORAE facilitates an intelligent service regulation pipeline by further exploiting a fine-tuned large language model named HORAE that automates the HORAE modeling process, thereby yielding an end-to-end framework for fully automated intelligent service regulation.
Authors: Ziyue Qiao, Junren Xiao, Qingqiang Sun, Meng Xiao, Xiao Luo, Hui Xiong
Abstract: This paper addresses the challenge of incremental learning in growing graphs with increasingly complex tasks. The goal is to continuously train a graph model to handle new tasks while retaining proficiency in previous tasks via memory replay. Existing methods usually overlook the importance of memory diversity, limiting in selecting high-quality memory from previous tasks and remembering broad previous knowledge within the scarce memory on graphs. To address that, we introduce a novel holistic Diversified Memory Selection and Generation (DMSG) framework for incremental learning in graphs, which first introduces a buffer selection strategy that considers both intra-class and inter-class diversities, employing an efficient greedy algorithm for sampling representative training nodes from graphs into memory buffers after learning each new task. Then, to adequately rememorize the knowledge preserved in the memory buffer when learning new tasks, a diversified memory generation replay method is introduced. This method utilizes a variational layer to generate the distribution of buffer node embeddings and sample synthesized ones for replaying. Furthermore, an adversarial variational embedding learning method and a reconstruction-based decoder are proposed to maintain the integrity and consolidate the generalization of the synthesized node embeddings, respectively. Extensive experimental results on publicly accessible datasets demonstrate the superiority of \method{} over state-of-the-art methods.
Authors: Alexander Nikulin, Ilya Zisman, Alexey Zemtsov, Vladislav Kurenkov
Abstract: Following the success of the in-context learning paradigm in large-scale language and computer vision models, the recently emerging field of in-context reinforcement learning is experiencing a rapid growth. However, its development has been held back by the lack of challenging benchmarks, as all the experiments have been carried out in simple environments and on small-scale datasets. We present XLand-100B, a large-scale dataset for in-context reinforcement learning based on the XLand-MiniGrid environment, as a first step to alleviate this problem. It contains complete learning histories for nearly $30,000$ different tasks, covering $100$B transitions and 2.5B episodes. It took 50,000 GPU hours to collect the dataset, which is beyond the reach of most academic labs. Along with the dataset, we provide the utilities to reproduce or expand it even further. We also benchmark common in-context RL baselines and show that they struggle to generalize to novel and diverse tasks. With this substantial effort, we aim to democratize research in the rapidly growing field of in-context reinforcement learning and provide a solid foundation for further scaling.
Authors: Jiawen Qin, Haonan Yuan, Qingyun Sun, Lyujin Xu, Jiaqi Yuan, Pengfeng Huang, Zhaonan Wang, Xingcheng Fu, Hao Peng, Jianxin Li, Philip S. Yu
Abstract: Deep graph learning has gained grand popularity over the past years due to its versatility and success in representing graph data across a wide range of domains. However, the pervasive issue of imbalanced graph data distributions, where certain parts exhibit disproportionally abundant data while others remain sparse, undermines the efficacy of conventional graph learning algorithms, leading to biased outcomes. To address this challenge, Imbalanced Graph Learning (IGL) has garnered substantial attention, enabling more balanced data distributions and better task performance. Despite the proliferation of IGL algorithms, the absence of consistent experimental protocols and fair performance comparisons pose a significant barrier to comprehending advancements in this field. To bridge this gap, we introduce IGL-Bench, a foundational comprehensive benchmark for imbalanced graph learning, embarking on 16 diverse graph datasets and 24 distinct IGL algorithms with uniform data processing and splitting strategies. Specifically, IGL-Bench systematically investigates state-of-the-art IGL algorithms in terms of effectiveness, robustness, and efficiency on node-level and graph-level tasks, with the scope of class-imbalance and topology-imbalance. Extensive experiments demonstrate the potential benefits of IGL algorithms on various imbalanced conditions, offering insights and opportunities in the IGL field. Further, we have developed an open-sourced and unified package to facilitate reproducible evaluation and inspire further innovative research, which is available at https://github.com/RingBDStack/IGL-Bench.
Authors: Shuyu Dong, Mich\`ele Sebag, Kento Uemura, Akito Fujii, Shuang Chang, Yusuke Koyanagi, Koji Maruhashi
Abstract: Causal learning tackles the computationally demanding task of estimating causal graphs. This paper introduces a new divide-and-conquer approach for causal graph learning, called DCILP. In the divide phase, the Markov blanket MB($X_i$) of each variable $X_i$ is identified, and causal learning subproblems associated with each MB($X_i$) are independently addressed in parallel. This approach benefits from a more favorable ratio between the number of data samples and the number of variables considered. In counterpart, it can be adversely affected by the presence of hidden confounders, as variables external to MB($X_i$) might influence those within it. The reconciliation of the local causal graphs generated during the divide phase is a challenging combinatorial optimization problem, especially in large-scale applications. The main novelty of DCILP is an original formulation of this reconciliation as an integer linear programming (ILP) problem, which can be delegated and efficiently handled by an ILP solver. Through experiments on medium to large scale graphs, and comparisons with state-of-the-art methods, DCILP demonstrates significant improvements in terms of computational complexity, while preserving the learning accuracy on real-world problem and suffering at most a slight loss of accuracy on synthetic problems.
Authors: Runzhe Wu, Ayush Sekhari, Akshay Krishnamurthy, Wen Sun
Abstract: We study computationally and statistically efficient Reinforcement Learning algorithms for the linear Bellman Complete setting. This setting uses linear function approximation to capture value functions and unifies existing models like linear Markov Decision Processes (MDP) and Linear Quadratic Regulators (LQR). While it is known from the prior works that this setting is statistically tractable, it remained open whether a computationally efficient algorithm exists. Our work provides a computationally efficient algorithm for the linear Bellman complete setting that works for MDPs with large action spaces, random initial states, and random rewards but relies on the underlying dynamics to be deterministic. Our approach is based on randomization: we inject random noise into least squares regression problems to perform optimistic value iteration. Our key technical contribution is to carefully design the noise to only act in the null space of the training data to ensure optimism while circumventing a subtle error amplification issue.
Authors: Taiqiang Wu, Jiahao Wang, Zhe Zhao, Ngai Wong
Abstract: In this paper, we introduce a subspace-inspired Low-Rank Adaptation (LoRA) method, which is computationally efficient, easy to implement, and readily applicable to large language, multimodal, and diffusion models. Initially, we equivalently decompose the weights of LoRA into two subspaces, and find that simply mixing them can enhance performance. To study such a phenomenon, we revisit it through a fine-grained subspace lens, showing that such modification is equivalent to employing a fixed mixer to fuse the subspaces. To be more flexible, we jointly learn the mixer with the original LoRA weights, and term the method Mixture-of-Subspaces LoRA (MoSLoRA). MoSLoRA consistently outperforms LoRA on tasks in different modalities, including commonsense reasoning, visual instruction tuning, and subject-driven text-to-image generation, demonstrating its effectiveness and robustness. Codes are available at https://github.com/wutaiqiang/MoSLoRA.
Authors: Yuan Chen, Dongbin Xiu
Abstract: We present a numerical method for learning unknown nonautonomous stochastic dynamical system, i.e., stochastic system subject to time dependent excitation or control signals. Our basic assumption is that the governing equations for the stochastic system are unavailable. However, short bursts of input/output (I/O) data consisting of certain known excitation signals and their corresponding system responses are available. When a sufficient amount of such I/O data are available, our method is capable of learning the unknown dynamics and producing an accurate predictive model for the stochastic responses of the system subject to arbitrary excitation signals not in the training data. Our method has two key components: (1) a local approximation of the training I/O data to transfer the learning into a parameterized form; and (2) a generative model to approximate the underlying unknown stochastic flow map in distribution. After presenting the method in detail, we present a comprehensive set of numerical examples to demonstrate the performance of the proposed method, especially for long-term system predictions.
Authors: Lorenzo Basile, Santiago Acevedo, Luca Bortolussi, Fabio Anselmi, Alex Rodriguez
Abstract: To gain insight into the mechanisms behind machine learning methods, it is crucial to establish connections among the features describing data points. However, these correlations often exhibit a high-dimensional and strongly nonlinear nature, which makes them challenging to detect using standard methods. This paper exploits the entanglement between intrinsic dimensionality and correlation to propose a metric that quantifies the (potentially nonlinear) correlation between high-dimensional manifolds. We first validate our method on synthetic data in controlled environments, showcasing its advantages and drawbacks compared to existing techniques. Subsequently, we extend our analysis to large-scale applications in neural network representations. Specifically, we focus on latent representations of multimodal data, uncovering clear correlations between paired visual and textual embeddings, whereas existing methods struggle significantly in detecting similarity. Our results indicate the presence of highly nonlinear correlation patterns between latent manifolds.
Authors: Somnath Basu Roy Chowdhury, Krzysztof Choromanski, Arijit Sehanobish, Avinava Dubey, Snigdha Chaturvedi
Abstract: Machine unlearning is the process of efficiently removing the influence of a training data instance from a trained machine learning model without retraining it from scratch. A popular subclass of unlearning approaches is exact machine unlearning, which focuses on techniques that explicitly guarantee the removal of the influence of a data instance from a model. Exact unlearning approaches use a machine learning model in which individual components are trained on disjoint subsets of the data. During deletion, exact unlearning approaches only retrain the affected components rather than the entire model. While existing approaches reduce retraining costs, it can still be expensive for an organization to retrain a model component as it requires halting a system in production, which leads to service failure and adversely impacts customers. To address these challenges, we introduce an exact unlearning framework -- Sequence-aware Sharded Sliced Training (S3T), which is designed to enhance the deletion capabilities of an exact unlearning system while minimizing the impact on model's performance. At the core of S3T, we utilize a lightweight parameter-efficient fine-tuning approach that enables parameter isolation by sequentially training layers with disjoint data slices. This enables efficient unlearning by simply deactivating the layers affected by data deletion. Furthermore, to reduce the retraining cost and improve model performance, we train the model on multiple data sequences, which allows S3T to handle an increased number of deletion requests. Both theoretically and empirically, we demonstrate that S3T attains superior deletion capabilities and enhanced performance compared to baselines across a wide range of settings.
Authors: Susan Athey, Herman Brunborg, Tianyu Du, Ayush Kanodia, Keyon Vafa
Abstract: Vafa et al. (2024) introduced a transformer-based econometric model, CAREER, that predicts a worker's next job as a function of career history (an "occupation model"). CAREER was initially estimated ("pre-trained") using a large, unrepresentative resume dataset, which served as a "foundation model," and parameter estimation was continued ("fine-tuned") using data from a representative survey. CAREER had better predictive performance than benchmarks. This paper considers an alternative where the resume-based foundation model is replaced by a large language model (LLM). We convert tabular data from the survey into text files that resemble resumes and fine-tune the LLMs using these text files with the objective to predict the next token (word). The resulting fine-tuned LLM is used as an input to an occupation model. Its predictive performance surpasses all prior models. We demonstrate the value of fine-tuning and further show that by adding more career data from a different population, fine-tuning smaller LLMs surpasses the performance of fine-tuning larger models.
Authors: Justin Xu, Jack Gallifant, Alistair E. W. Johnson, Matthew B. A. McDermott
Abstract: Reproducibility remains a significant challenge in machine learning (ML) for healthcare. Datasets, model pipelines, and even task or cohort definitions are often private in this field, leading to a significant barrier in sharing, iterating, and understanding ML results on electronic health record (EHR) datasets. We address a significant part of this problem by introducing the Automatic Cohort Extraction System (ACES) for event-stream data. This library is designed to simultaneously simplify the development of tasks and cohorts for ML in healthcare and also enable their reproduction, both at an exact level for single datasets and at a conceptual level across datasets. To accomplish this, ACES provides: (1) a highly intuitive and expressive domain-specific configuration language for defining both dataset-specific concepts and dataset-agnostic inclusion or exclusion criteria, and (2) a pipeline to automatically extract patient records that meet these defined criteria from real-world data. ACES can be automatically applied to any dataset in either the Medical Event Data Standard (MEDS) or Event Stream GPT (ESGPT) formats, or to *any* dataset in which the necessary task-specific predicates can be extracted in an event-stream form. ACES has the potential to significantly lower the barrier to entry for defining ML tasks in representation learning, redefine the way researchers interact with EHR datasets, and significantly improve the state of reproducibility for ML studies using this modality. ACES is available at: https://github.com/justin13601/aces.
Authors: Yuheng Zhang, Dian Yu, Baolin Peng, Linfeng Song, Ye Tian, Mingyue Huo, Nan Jiang, Haitao Mi, Dong Yu
Abstract: Reinforcement Learning with Human Feedback (RLHF) has achieved great success in aligning large language models (LLMs) with human preferences. Prevalent RLHF approaches are reward-based, following the Bradley-Terry (BT) model assumption, which may not fully capture the complexity of human preferences. In this paper, we explore RLHF under a general preference framework and approach it from a game-theoretic perspective. Specifically, we formulate the problem as a two-player game and propose a novel online algorithm, iterative Nash policy optimization (INPO). The key idea is to let the policy play against itself via no-regret learning, thereby approximating the Nash policy. Unlike previous methods, INPO bypasses the need for estimating the expected win rate for individual responses, which typically incurs high computational or annotation costs. Instead, we introduce a new loss objective that is directly minimized over a preference dataset. We provide theoretical analysis for our approach and demonstrate its effectiveness through experiments on various representative benchmarks. With an LLaMA-3-8B-based SFT model, INPO achieves a 42.6% length-controlled win rate on AlpacaEval 2.0 and a 37.8% win rate on Arena-Hard, showing substantial improvement over the state-of-the-art online RLHF algorithms.
Authors: Han-Jia Ye, Huai-Hong Yin, De-Chuan Zhan, Wei-Lun Chao
Abstract: The widespread enthusiasm for deep learning has recently expanded into the domain of tabular data. Recognizing that the advancement in deep tabular methods is often inspired by classical methods, e.g., integration of nearest neighbors into neural networks, we investigate whether these classical methods can be revitalized with modern techniques. We revisit a differentiable version of $K$-nearest neighbors (KNN) -- Neighbourhood Components Analysis (NCA) -- originally designed to learn a linear projection to capture semantic similarities between instances, and seek to gradually add modern deep learning techniques on top. Surprisingly, our implementation of NCA using SGD and without dimensionality reduction already achieves decent performance on tabular data, in contrast to the results of using existing toolboxes like scikit-learn. Further equipping NCA with deep representations and additional training stochasticity significantly enhances its capability, being on par with the leading tree-based method CatBoost and outperforming existing deep tabular models in both classification and regression tasks on 300 datasets. We conclude our paper by analyzing the factors behind these improvements, including loss functions, prediction strategies, and deep architectures. The code is available at https://github.com/qile2000/LAMDA-TALENT.
Authors: Yuxuan Wu, Ziyu Wang, Bhiksha Raj, Gus Xia
Abstract: We contribute an unsupervised method that effectively learns disentangled content and style representations from sequences of observations. Unlike most disentanglement algorithms that rely on domain-specific labels or knowledge, our method is based on the insight of domain-general statistical differences between content and style -- content varies more among different fragments within a sample but maintains an invariant vocabulary across data samples, whereas style remains relatively invariant within a sample but exhibits more significant variation across different samples. We integrate such inductive bias into an encoder-decoder architecture and name our method after V3 (variance-versus-invariance). Experimental results show that V3 generalizes across multiple domains and modalities, successfully learning disentangled content and style representations, such as pitch and timbre from music audio, digit and color from images of hand-written digits, and action and character appearance from simple animations. V3 demonstrates strong disentanglement performance compared to existing unsupervised methods, along with superior out-of-distribution generalization under few-shot adaptation compared to supervised counterparts. Lastly, symbolic-level interpretability emerges in the learned content codebook, forging a near one-to-one alignment between machine representation and human knowledge.
Authors: Yi-Chen Li, Fuxiang Zhang, Wenjie Qiu, Lei Yuan, Chengxing Jia, Zongzhang Zhang, Yang Yu, Bo An
Abstract: Large Language Models (LLMs), trained on a large amount of corpus, have demonstrated remarkable abilities. However, it may not be sufficient to directly apply open-source LLMs like Llama to certain real-world scenarios, since most of them are trained for \emph{general} purposes. Thus, the demands for customizing publicly available LLMs emerge, but are currently under-studied. In this work, we consider customizing pre-trained LLMs with new human preferences. Specifically, the LLM should not only meet the new preference but also preserve its original capabilities after customization. Drawing inspiration from the observation that human preference can be expressed as a reward model, we propose to cast LLM customization as optimizing the sum of two reward functions, one of which (denoted as $r_1$) was used to pre-train the LLM while the other (denoted as $r_2$) characterizes the new human preference. The obstacle here is that both reward functions are unknown, making the application of modern reinforcement learning methods infeasible. Thanks to the residual Q-learning framework, we can restore the customized LLM with the pre-trained LLM and the \emph{residual Q-function} without the reward function $r_1$. Moreover, we find that for a fixed pre-trained LLM, the reward function $r_2$ can be derived from the residual Q-function, enabling us to directly learn the residual Q-function from the new human preference data upon the Bradley-Terry model. We name our method Q-Adapter as it introduces an adapter module to approximate the residual Q-function for customizing the pre-trained LLM towards the new preference. Experiments based on the Llama-3.1 model on the DSP dataset and HH-RLHF dataset illustrate the superior effectiveness of Q-Adapter on both retaining existing knowledge and learning new preferences. Code is available at https://github.com/mansicer/Q-Adapter.
Authors: Jiawei Xu, Rui Yang, Shuang Qiu, Feng Luo, Meng Fang, Baoxiang Wang, Lei Han
Abstract: Learning policy from offline datasets through offline reinforcement learning (RL) holds promise for scaling data-driven decision-making while avoiding unsafe and costly online interactions. However, real-world data collected from sensors or humans often contains noise and errors, posing a significant challenge for existing offline RL methods, particularly when the real-world data is limited. Our study reveals that prior research focusing on adapting predominant offline RL methods based on temporal difference learning still falls short under data corruption when the dataset is limited. In contrast, we discover that vanilla sequence modeling methods, such as Decision Transformer, exhibit robustness against data corruption, even without specialized modifications. To unlock the full potential of sequence modeling, we propose Robust Decision Rransformer (RDT) by incorporating three simple yet effective robust techniques: embedding dropout to improve the model's robustness against erroneous inputs, Gaussian weighted learning to mitigate the effects of corrupted labels, and iterative data correction to eliminate corrupted data from the source. Extensive experiments on MuJoCo, Kitchen, and Adroit tasks demonstrate RDT's superior performance under various data corruption scenarios compared to prior methods. Furthermore, RDT exhibits remarkable robustness in a more challenging setting that combines training-time data corruption with test-time observation perturbations. These results highlight the potential of sequence modeling for learning from noisy or corrupted offline datasets, thereby promoting the reliable application of offline RL in real-world scenarios. Our code is available at https://github.com/jiawei415/RobustDecisionTransformer.
URLs: https://github.com/jiawei415/RobustDecisionTransformer.
Authors: Kai Ruan, Yilong Xu, Ze-Feng Gao, Yike Guo, Hao Sun, Ji-Rong Wen, Yang Liu
Abstract: Symbolic regression plays a crucial role in modern scientific research thanks to its capability of discovering concise and interpretable mathematical expressions from data. A grand challenge lies in the arduous search for parsimonious and generalizable mathematical formulas, in an infinite search space, while intending to fit the training data. Existing algorithms have faced a critical bottleneck of accuracy and efficiency over a decade when handling problems of complexity, which essentially hinders the pace of applying symbolic regression for scientific exploration across interdisciplinary domains. To this end, we introduce a parallel combinatorial tree search (PCTS) model to efficiently distill generic mathematical expressions from limited data. Through a series of extensive experiments, we demonstrate the superior accuracy and efficiency of PCTS for equation discovery, which greatly outperforms the state-of-the-art baseline models on over 200 synthetic and experimental datasets (e.g., lifting its performance by up to 99% accuracy improvement and one-order of magnitude speed up). PCTS represents a key advance in accurate and efficient data-driven discovery of symbolic, interpretable models (e.g., underlying physical laws) and marks a pivotal transition towards scalable symbolic learning.
Authors: Xingrun Xing, Boyan Gao, Zheng Zhang, David A. Clifton, Shitao Xiao, Li Du, Guoqi Li, Jiajun Zhang
Abstract: Recent advancements in large language models (LLMs) with billions of parameters have improved performance in various applications, but their inference processes demand significant energy and computational resources. In contrast, the human brain, with approximately 86 billion neurons, is much more energy-efficient than LLMs with similar parameters. Inspired by this, we redesign 7$\sim$70 billion parameter LLMs using bio-plausible spiking mechanisms, emulating the efficient behavior of the human brain. We propose the first spiking large language model, SpikeLLM. Coupled with the proposed model, two essential approaches are proposed to improve spike training efficiency: Generalized Integrate-and-Fire (GIF) neurons to compress spike length from $T$ to $\frac{T}{L} \log_2 L$ bits, and an Optimal Brain Spiking framework to divide outlier channels and allocate different $T$ for GIF neurons, which further compresses spike length to approximate $log_2T$ bits. The necessity of spike-driven LLM is proved by comparison with quantized LLMs with similar operations. In the OmniQuant pipeline, SpikeLLM reduces 11.01% WikiText2 perplexity and improves 2.55% accuracy of common scene reasoning on a LLAMA-7B W4A4 model. In the GPTQ pipeline, SpikeLLM achieves direct additive in linear layers, significantly exceeding PB-LLMs.
Authors: Tongzhou Liao, Barnab\'as P\'oczos
Abstract: Graph Neural Networks (GNNs) have become important tools for machine learning on graph-structured data. In this paper, we explore the synergistic combination of graph encoding, graph rewiring, and graph attention, by introducing Graph Attention with Stochastic Structures (GRASS), a novel GNN architecture. GRASS utilizes relative random walk probabilities (RRWP) encoding and a novel decomposed variant (D-RRWP) to efficiently capture structural information. It rewires the input graph by superimposing a random regular graph to enhance long-range information propagation. It also employs a novel additive attention mechanism tailored for graph-structured data. Our empirical evaluations demonstrate that GRASS achieves state-of-the-art performance on multiple benchmark datasets, including a 20.3% reduction in mean absolute error on the ZINC dataset.
Authors: Chongyang Gao, Lixu Wang, Kaize Ding, Chenkai Weng, Xiao Wang, Qi Zhu
Abstract: While large language models have demonstrated impressive performance across various domains and tasks, their security issues have become increasingly severe. Machine unlearning has emerged as a representative approach for model safety and security by removing the influence of undesired data on the target model. However, these methods do not sufficiently consider that unlearning requests in real-world scenarios are continuously emerging, especially in the context of LLMs, which may lead to accumulated model utility loss that eventually becomes unacceptable. Moreover, existing LLM unlearning methods often ignore previous data access limitations due to privacy concerns and copyright protection. Without previous data, the utility preservation during unlearning is much harder. To overcome these challenges, we propose the OOO framework that includes an Orthogonal low-rank adapter (LoRA) for continually unlearning requested data and an Out-Of-Distribution (OOD) detector to measure the similarity between input and unlearning data. The orthogonal LoRA achieves parameter disentanglement among continual unlearning requests. The OOD detector is trained with a novel contrastive entropy loss and utilizes a glocal-aware scoring mechanism. During inference, our OOO framework can decide whether and to what extent to load the unlearning LoRA based on the OOD detector's predicted similarity between the input and the unlearned knowledge. Notably, OOO's effectiveness does not rely on any retained data. We conducted extensive experiments on OOO and state-of-the-art LLM unlearning methods across three tasks and seven datasets. The results indicate that OOO consistently achieves the best unlearning effectiveness and utility preservation, especially when facing continuous unlearning requests. The source codes can be found at https://github.com/GCYZSL/O3-LLM-UNLEARNING.
Authors: Haohong Lin, Wenhao Ding, Jian Chen, Laixi Shi, Jiacheng Zhu, Bo Li, Ding Zhao
Abstract: Offline model-based reinforcement learning (MBRL) enhances data efficiency by utilizing pre-collected datasets to learn models and policies, especially in scenarios where exploration is costly or infeasible. Nevertheless, its performance often suffers from the objective mismatch between model and policy learning, resulting in inferior performance despite accurate model predictions. This paper first identifies the primary source of this mismatch comes from the underlying confounders present in offline data for MBRL. Subsequently, we introduce \textbf{B}ilin\textbf{E}ar \textbf{CAUS}al r\textbf{E}presentation~(BECAUSE), an algorithm to capture causal representation for both states and actions to reduce the influence of the distribution shift, thus mitigating the objective mismatch problem. Comprehensive evaluations on 18 tasks that vary in data quality and environment context demonstrate the superior performance of BECAUSE over existing offline RL algorithms. We show the generalizability and robustness of BECAUSE under fewer samples or larger numbers of confounders. Additionally, we offer theoretical analysis of BECAUSE to prove its error bound and sample efficiency when integrating causal representation into offline MBRL.
Authors: Pantelis Vafidis, Aman Bhargava, Antonio Rangel
Abstract: Intelligent perception and interaction with the world hinges on internal representations that capture its underlying structure (''disentangled'' or ''abstract'' representations). Disentangled representations serve as world models, isolating latent factors of variation in the world along approximately orthogonal directions, thus facilitating feature-based generalization. We provide experimental and theoretical results guaranteeing the emergence of disentangled representations in agents that optimally solve multi-task evidence accumulation classification tasks, canonical in the neuroscience literature. The key conceptual finding is that, by producing accurate multi-task classification estimates, a system implicitly represents a set of coordinates specifying a disentangled representation of the underlying latent state of the data it receives. The theory provides conditions for the emergence of these representations in terms of noise, number of tasks, and evidence accumulation time. We experimentally validate these predictions in RNNs trained to multi-task, which learn disentangled representations in the form of continuous attractors, leading to zero-shot out-of-distribution (OOD) generalization in predicting latent factors. We demonstrate the robustness of our framework across autoregressive architectures, decision boundary geometries and in tasks requiring classification confidence estimation. We find that transformers are particularly suited for disentangling representations, which might explain their unique world understanding abilities. Overall, our framework establishes a formal link between competence at multiple tasks and the formation of disentangled, interpretable world models in both biological and artificial systems, and helps explain why ANNs often arrive at human-interpretable concepts, and how they both may acquire exceptional zero-shot generalization capabilities.
Authors: Janez Bo\v{z}i\v{c}, Am\^andio R. Faustino, Boris Radovi\v{c}, Marco Canini, Veljko Pejovi\'c
Abstract: Progressing beyond centralized AI is of paramount importance, yet, distributed AI solutions, in particular various federated learning (FL) algorithms, are often not comprehensively assessed, which prevents the research community from identifying the most promising approaches and practitioners from being convinced that a certain solution is deployment-ready. The largest hurdle towards FL algorithm evaluation is the difficulty of conducting real-world experiments over a variety of FL client devices and different platforms, with different datasets and data distribution, all while assessing various dimensions of algorithm performance, such as inference accuracy, energy consumption, and time to convergence, to name a few. In this paper, we present CoLExT, a real-world testbed for FL research. CoLExT is designed to streamline experimentation with custom FL algorithms in a rich testbed configuration space, with a large number of heterogeneous edge devices, ranging from single-board computers to smartphones, and provides real-time collection and visualization of a variety of metrics through automatic instrumentation. According to our evaluation, porting FL algorithms to CoLExT requires minimal involvement from the developer, and the instrumentation introduces minimal resource usage overhead. Furthermore, through an initial investigation involving popular FL algorithms running on CoLExT, we reveal previously unknown trade-offs, inefficiencies, and programming bugs.
Authors: Alexey Skrynnik, Anton Andreychuk, Anatolii Borzilov, Alexander Chernyavskiy, Konstantin Yakovlev, Aleksandr Panov
Abstract: Multi-agent reinforcement learning (MARL) has recently excelled in solving challenging cooperative and competitive multi-agent problems in various environments, typically involving a small number of agents and full observability. Moreover, a range of crucial robotics-related tasks, such as multi-robot pathfinding, which have traditionally been approached with classical non-learnable methods (e.g., heuristic search), are now being suggested for solution using learning-based or hybrid methods. However, in this domain, it remains difficult, if not impossible, to conduct a fair comparison between classical, learning-based, and hybrid approaches due to the lack of a unified framework that supports both learning and evaluation. To address this, we introduce POGEMA, a comprehensive set of tools that includes a fast environment for learning, a problem instance generator, a collection of predefined problem instances, a visualization toolkit, and a benchmarking tool for automated evaluation. We also introduce and define an evaluation protocol that specifies a range of domain-related metrics, computed based on primary evaluation indicators (such as success rate and path length), enabling a fair multi-fold comparison. The results of this comparison, which involves a variety of state-of-the-art MARL, search-based, and hybrid methods, are presented.
Authors: Chenwei Zhang, Anne Condon, Khanh Dao Duc
Abstract: Generating synthetic cryogenic electron microscopy 3D density maps from molecular structures has potential important applications in structural biology. Yet existing simulation-based methods cannot mimic all the complex features present in experimental maps, such as secondary structure elements. As an alternative, we propose struc2mapGAN, a novel data-driven method that employs a generative adversarial network to produce improved experimental-like density maps from molecular structures. More specifically, struc2mapGAN uses a nested U-Net architecture as the generator, with an additional L1 loss term and further processing of raw training experimental maps to enhance learning efficiency. While struc2mapGAN can promptly generate maps after training, we demonstrate that it outperforms existing simulation-based methods for a wide array of tested maps and across various evaluation metrics.
Authors: Muhammad Arbab Arshad, Talukder Zaki Jubery, Tirtho Roy, Rim Nassiri, Asheesh K. Singh, Arti Singh, Chinmay Hegde, Baskar Ganapathysubramanian, Aditya Balu, Adarsh Krishnamurthy, Soumik Sarkar
Abstract: As Vision Language Models (VLMs) become increasingly accessible to farmers and agricultural experts, there is a growing need to evaluate their potential in specialized tasks. We present AgEval, a comprehensive benchmark for assessing VLMs' capabilities in plant stress phenotyping, offering a solution to the challenge of limited annotated data in agriculture. Our study explores how general-purpose VLMs can be leveraged for domain-specific tasks with only a few annotated examples, providing insights into their behavior and adaptability. AgEval encompasses 12 diverse plant stress phenotyping tasks, evaluating zero-shot and few-shot in-context learning performance of state-of-the-art models including Claude, GPT, Gemini, and LLaVA. Our results demonstrate VLMs' rapid adaptability to specialized tasks, with the best-performing model showing an increase in F1 scores from 46.24% to 73.37% in 8-shot identification. To quantify performance disparities across classes, we introduce metrics such as the coefficient of variation (CV), revealing that VLMs' training impacts classes differently, with CV ranging from 26.02% to 58.03%. We also find that strategic example selection enhances model reliability, with exact category examples improving F1 scores by 15.38% on average. AgEval establishes a framework for assessing VLMs in agricultural applications, offering valuable benchmarks for future evaluations. Our findings suggest that VLMs, with minimal few-shot examples, show promise as a viable alternative to traditional specialized models in plant stress phenotyping, while also highlighting areas for further refinement. Results and benchmark details are available at: https://github.com/arbab-ml/AgEval
Authors: Ali Kashefi
Abstract: Kolmogorov-Arnold Networks (KANs) have emerged as a promising alternative to traditional Multilayer Perceptrons (MLPs) in deep learning. KANs have already been integrated into various architectures, such as convolutional neural networks, graph neural networks, and transformers, and their potential has been assessed for predicting physical quantities. However, the combination of KANs with point-cloud-based neural networks (e.g., PointNet) for computational physics has not yet been explored. To address this, we present Kolmogorov-Arnold PointNet (KA-PointNet) as a novel supervised deep learning framework for the prediction of incompressible steady-state fluid flow fields in irregular domains, where the predicted fields are a function of the geometry of the domains. In KA-PointNet, we implement shared KANs in the segmentation branch of the PointNet architecture. We utilize Jacobi polynomials to construct shared KANs. As a benchmark test case, we consider incompressible laminar steady-state flow over a cylinder, where the geometry of its cross-section varies over the data set. We investigate the performance of Jacobi polynomials with different degrees as well as special cases of Jacobi polynomials such as Legendre polynomials, Chebyshev polynomials of the first and second kinds, and Gegenbauer polynomials, in terms of the computational cost of training and accuracy of prediction of the test set. Additionally, we compare the performance of PointNet with shared KANs (i.e., KA-PointNet) and PointNet with shared MLPs. It is observed that when the number of trainable parameters is approximately equal, PointNet with shared KANs (i.e., KA-PointNet) outperforms PointNet with shared MLPs. Moreover, KA-PointNet predicts the pressure and velocity distributions along the surface of cylinders more accurately, resulting in more precise computations of lift and drag.
Authors: Xinghao Dong, Chuanqi Chen, Jin-Long Wu
Abstract: Closure models are widely used in simulating complex multiscale dynamical systems such as turbulence and the earth system, for which direct numerical simulation that resolves all scales is often too expensive. For those systems without a clear scale separation, deterministic and local closure models often lack enough generalization capability, which limits their performance in many real-world applications. In this work, we propose a data-driven modeling framework for constructing stochastic and non-local closure models via conditional diffusion model and neural operator. Specifically, the Fourier neural operator is incorporated into a score-based diffusion model, which serves as a data-driven stochastic closure model for complex dynamical systems governed by partial differential equations (PDEs). We also demonstrate how accelerated sampling methods can improve the efficiency of the data-driven stochastic closure model. The results show that the proposed methodology provides a systematic approach via generative machine learning techniques to construct data-driven stochastic closure models for multiscale dynamical systems with continuous spatiotemporal fields.
Authors: Rafa{\l} Karczewski, Samuel Kaski, Markus Heinonen, Vikas Garg
Abstract: Several generative models with elaborate training and sampling procedures have been proposed to accelerate structure-based drug design (SBDD); however, their empirical performance turns out to be suboptimal. We seek to better understand this phenomenon from both theoretical and empirical perspectives. Since most of these models apply graph neural networks (GNNs), one may suspect that they inherit the representational limitations of GNNs. We analyze this aspect, establishing the first such results for protein-ligand complexes. A plausible counterview may attribute the underperformance of these models to their excessive parameterizations, inducing expressivity at the expense of generalization. We investigate this possibility with a simple metric-aware approach that learns an economical surrogate for affinity to infer an unlabelled molecular graph and optimizes for labels conditioned on this graph and molecular properties. The resulting model achieves state-of-the-art results using 100x fewer trainable parameters and affords up to 1000x speedup. Collectively, our findings underscore the need to reassess and redirect the existing paradigm and efforts for SBDD. Code is available at https://github.com/rafalkarczewski/SimpleSBDD.
Authors: Valdemar \v{S}v\'abensk\'y, Kristi\'an Tk\'a\v{c}ik, Aubrey Birdwell, Richard Weiss, Ryan S. Baker, Pavel \v{C}eleda, Jan Vykopal, Jens Mache, Ankur Chattopadhyay
Abstract: This full paper in the research track evaluates the usage of data logged from cybersecurity exercises in order to predict students who are potentially at risk of performing poorly. Hands-on exercises are essential for learning since they enable students to practice their skills. In cybersecurity, hands-on exercises are often complex and require knowledge of many topics. Therefore, students may miss solutions due to gaps in their knowledge and become frustrated, which impedes their learning. Targeted aid by the instructor helps, but since the instructor's time is limited, efficient ways to detect struggling students are needed. This paper develops automated tools to predict when a student is having difficulty. We formed a dataset with the actions of 313 students from two countries and two learning environments: KYPO CRP and EDURange. These data are used in machine learning algorithms to predict the success of students in exercises deployed in these environments. After extracting features from the data, we trained and cross-validated eight classifiers for predicting the exercise outcome and evaluated their predictive power. The contribution of this paper is comparing two approaches to feature engineering, modeling, and classification performance on data from two learning environments. Using the features from either learning environment, we were able to detect and distinguish between successful and struggling students. A decision tree classifier achieved the highest balanced accuracy and sensitivity with data from both learning environments. The results show that activity data from cybersecurity exercises are suitable for predicting student success. In a potential application, such models can aid instructors in detecting struggling students and providing targeted help. We publish data and code for building these models so that others can adopt or adapt them.
Authors: Chi-Heng Lin, Shangqian Gao, James Seale Smith, Abhishek Patel, Shikhar Tuli, Yilin Shen, Hongxia Jin, Yen-Chang Hsu
Abstract: Large Language Models (LLMs) have reshaped the landscape of artificial intelligence by demonstrating exceptional performance across various tasks. However, substantial computational requirements make their deployment challenging on devices with limited resources. Recently, compression methods using low-rank matrix techniques have shown promise, yet these often lead to degraded accuracy or introduce significant overhead in parameters and inference latency. This paper introduces \textbf{Mo}dular \textbf{De}composition (MoDeGPT), a novel structured compression framework that does not need recovery fine-tuning while resolving the above drawbacks. MoDeGPT partitions the Transformer block into modules comprised of matrix pairs and reduces the hidden dimensions via reconstructing the module-level outputs. MoDeGPT is developed based on a theoretical framework that utilizes three well-established matrix decomposition algorithms -- Nystr\"om approximation, CR decomposition, and SVD -- and applies them to our redefined transformer modules. Our comprehensive experiments show MoDeGPT, without backward propagation, matches or surpasses previous structured compression methods that rely on gradient information, and saves 98% of compute costs on compressing a 13B model. On \textsc{Llama}-2/3 and OPT models, MoDeGPT maintains 90-95% zero-shot performance with 25-30% compression rates. Moreover, the compression can be done on a single GPU within a few hours and increases the inference throughput by up to 46%.
Authors: Dominic Phillips, Flaviu Cipcigan
Abstract: Generative Flow Networks (GFlowNets) are a class of generative models that sample objects in proportion to a specified reward function through a learned policy. They can be trained either on-policy or off-policy, needing a balance between exploration and exploitation for fast convergence to a target distribution. While exploration strategies for discrete GFlowNets have been studied, exploration in the continuous case remains to be investigated, despite the potential for novel exploration algorithms due to the local connectedness of continuous domains. Here, we introduce Adapted Metadynamics, a variant of metadynamics that can be applied to arbitrary black-box reward functions on continuous domains. We use Adapted Metadynamics as an exploration strategy for continuous GFlowNets. We show several continuous domains where the resulting algorithm, MetaGFN, accelerates convergence to the target distribution and discovers more distant reward modes than previous off-policy exploration strategies used for GFlowNets.
Authors: Stephen Zhang, Vardan Papyan
Abstract: The recent paradigm shift to large-scale foundation models has brought about a new era for deep learning that, while has found great success in practice, has also been plagued by prohibitively expensive costs in terms of high memory consumption and compute. To mitigate these issues, there has been a concerted effort in post-hoc neural network pruning techniques that do not require costly retraining. Despite the considerable progress being made, existing methods often exhibit a steady drop in model performance as the compression increases. In this paper, we present a novel approach to compressing large transformers, coined OATS, that utilizes the second moment information in the input embeddings to decompose the model weights into a sum of sparse and low-rank matrices. Without any retraining, OATS achieves state-of-the-art performance when compressing models by up to $60\%$ on large language models such as Llama-3 and Phi-3 and vision transformers such as ViT and DINOv2 while delivering up to $1.37\times$ the CPU acceleration versus a model that was comparably pruned.
Authors: Th\'eo Bourdais, Houman Owhadi
Abstract: Whether deterministic or stochastic, models can be viewed as functions designed to approximate a specific quantity of interest. We introduce Minimal Empirical Variance Aggregation (MEVA), a data-driven framework that integrates predictions from various models, enhancing overall accuracy by leveraging the individual strengths of each. This non-intrusive, model-agnostic approach treats the contributing models as black boxes and accommodates outputs from diverse methodologies, including machine learning algorithms and traditional numerical solvers. We advocate for a point-wise linear aggregation process and consider two methods for optimizing this aggregate: Minimal Error Aggregation (MEA), which minimizes the prediction error, and Minimal Variance Aggregation (MVA), which focuses on reducing variance. We prove a theorem showing that MVA can be more robustly estimated from data than MEA, making MEVA superior to Minimal Empirical Error Aggregation (MEEA). Unlike MEEA, which interpolates target values directly, MEVA formulates aggregation as an error estimation problem, which can be performed using any backbone learning paradigm. We demonstrate the versatility and effectiveness of our framework across various applications, including data science and partial differential equations, illustrating its ability to significantly enhance both robustness and accuracy.
Authors: Qining Zhang, Lei Ying
Abstract: Reward inference (learning a reward model from human preferences) is a critical intermediate step in the Reinforcement Learning from Human Feedback (RLHF) pipeline for fine-tuning Large Language Models (LLMs). In practice, RLHF faces fundamental challenges such as distribution shift, reward model overfitting, and problem misspecification. An alternative approach is direct policy optimization without reward inference, such as Direct Preference Optimization (DPO), which provides a much simpler pipeline and has shown empirical success in LLM applications. However, DPO utilizes the closed-form expression between the optimal policy and the reward function, which is only suitable under the bandit setting or deterministic MDPs. This paper develops two RLHF algorithms without reward inference for general RL problems beyond bandits and deterministic MDPs, and general preference models beyond the Bradley-Terry model. The key idea is to estimate the local value function difference from human preferences and then approximate the policy gradient with a zeroth-order gradient approximator. For both algorithms, we establish polynomial convergence rates in terms of the number of policy gradient iterations, the number of trajectory samples, and human preference queries per iteration. Numerical experiments in stochastic environments validate the performance of our proposed algorithms, outperforming popular RLHF baselines such as DPO and PPO. Our paper shows there exist provably efficient methods to solve general RLHF problems without reward inference.
Authors: Francesco Mori, Stefano Sarao Mannelli, Francesca Mignacco
Abstract: Artificial neural networks often struggle with catastrophic forgetting when learning multiple tasks sequentially, as training on new tasks degrades the performance on previously learned tasks. Recent theoretical work has addressed this issue by analysing learning curves in synthetic frameworks under predefined training protocols. However, these protocols relied on heuristics and lacked a solid theoretical foundation assessing their optimality. In this paper, we fill this gap by combining exact equations for training dynamics, derived using statistical physics techniques, with optimal control methods. We apply this approach to teacher-student models for continual learning and multi-task problems, obtaining a theory for task-selection protocols maximising performance while minimising forgetting. Our theoretical analysis offers non-trivial yet interpretable strategies for mitigating catastrophic forgetting, shedding light on how optimal learning protocols modulate established effects, such as the influence of task similarity on forgetting. Finally, we validate our theoretical findings with experiments on real-world data.
Authors: Charles Marx, Volodymyr Kuleshov, Stefano Ermon
Abstract: Real-world data streams can change unpredictably due to distribution shifts, feedback loops and adversarial actors, which challenges the validity of forecasts. We present a forecasting framework ensuring valid uncertainty estimates regardless of how data evolves. Leveraging the concept of Blackwell approachability from game theory, we introduce a forecasting framework that guarantees calibrated uncertainties for outcomes in any compact space (e.g., classification or bounded regression). We extend this framework to recalibrate existing forecasters, guaranteeing calibration without sacrificing predictive performance. We implement both general-purpose gradient-based algorithms and algorithms optimized for popular special cases of our framework. Empirically, our algorithms improve calibration and downstream decision-making for energy systems.
Authors: Jie Cheng, Ruixi Qiao, Yingwei Ma, Binhua Li, Gang Xiong, Qinghai Miao, Yongbin Li, Yisheng Lv
Abstract: A significant aspiration of offline reinforcement learning (RL) is to develop a generalist agent with high capabilities from large and heterogeneous datasets. However, prior approaches that scale offline RL either rely heavily on expert trajectories or struggle to generalize to diverse unseen tasks. Inspired by the excellent generalization of world model in conditional video generation, we explore the potential of image observation-based world model for scaling offline RL and enhancing generalization on novel tasks. In this paper, we introduce JOWA: Jointly-Optimized World-Action model, an offline model-based RL agent pretrained on multiple Atari games with 6 billion tokens data to learn general-purpose representation and decision-making ability. Our method jointly optimizes a world-action model through a shared transformer backbone, which stabilize temporal difference learning with large models during pretraining. Moreover, we propose a provably efficient and parallelizable planning algorithm to compensate for the Q-value estimation error and thus search out better policies. Experimental results indicate that our largest agent, with 150 million parameters, achieves 78.9% human-level performance on pretrained games using only 10% subsampled offline data, outperforming existing state-of-the-art large-scale offline RL baselines by 31.6% on averange. Furthermore, JOWA scales favorably with model capacity and can sample-efficiently transfer to novel games using only 5k offline fine-tuning data (approximately 4 trajectories) per game, demonstrating superior generalization. We will release codes and model weights at https://github.com/CJReinforce/JOWA
Authors: Liangzu Peng, Juan Elenter, Joshua Agterberg, Alejandro Ribeiro, Ren\'e Vidal
Abstract: The goal of continual learning (CL) is to train a model that can solve multiple tasks presented sequentially. Recent CL approaches have achieved strong performance by leveraging large pre-trained models that generalize well to downstream tasks. However, such methods lack theoretical guarantees, making them prone to unexpected failures. Conversely, principled CL approaches often fail to achieve competitive performance. In this work, we aim to bridge this gap between theory and practice by designing a simple CL method that is theoretically sound and highly performant. Specifically, we lift pre-trained features into a higher dimensional space and formulate an over-parametrized minimum-norm least-squares problem. We find that the lifted features are highly ill-conditioned, potentially leading to large training errors (numerical instability) and increased generalization errors. We address these challenges by continually truncating the singular value decomposition (SVD) of the lifted features. Our approach, termed TSVD, is stable with respect to the choice of hyperparameters, can handle hundreds of tasks, and outperforms state-of-the-art CL methods on multiple datasets. Importantly, our method satisfies a recurrence relation throughout its continual learning process, which allows us to prove it maintains small training and generalization errors by appropriately truncating a fraction of SVD factors. This results in a stable continual learning method with strong empirical performance and theoretical guarantees. Code available: https://github.com/liangzu/tsvd.
Authors: Vahid Shahverdi, Giovanni Luca Marchetti, Kathl\'en Kohn
Abstract: We study convolutional neural networks with monomial activation functions. Specifically, we prove that their parameterization map is regular and is an isomorphism almost everywhere, up to rescaling the filters. By leveraging on tools from algebraic geometry, we explore the geometric properties of the image in function space of this map - typically referred to as neuromanifold. In particular, we compute the dimension and the degree of the neuromanifold, which measure the expressivity of the model, and describe its singularities. Moreover, for a generic large dataset, we derive an explicit formula that quantifies the number of critical points arising in the optimization of a regression loss.
Authors: Bocheng Zeng, Qi Wang, Mengtao Yan, Yang Liu, Ruizhi Chengze, Yi Zhang, Hongsheng Liu, Zidong Wang, Hao Sun
Abstract: Solving partial differential equations (PDEs) serves as a cornerstone for modeling complex dynamical systems. Recent progresses have demonstrated grand benefits of data-driven neural-based models for predicting spatiotemporal dynamics (e.g., tremendous speedup gain compared with classical numerical methods). However, most existing neural models rely on rich training data, have limited extrapolation and generalization abilities, and suffer to produce precise or reliable physical prediction under intricate conditions (e.g., irregular mesh or geometry, complex boundary conditions, diverse PDE parameters, etc.). To this end, we propose a new graph learning approach, namely, Physics-encoded Message Passing Graph Network (PhyMPGN), to model spatiotemporal PDE systems on irregular meshes given small training datasets. Specifically, we incorporate a GNN into a numerical integrator to approximate the temporal marching of spatiotemporal dynamics for a given PDE system. Considering that many physical phenomena are governed by diffusion processes, we further design a learnable Laplace block, which encodes the discrete Laplace-Beltrami operator, to aid and guide the GNN learning in a physically feasible solution space. A boundary condition padding strategy is also designed to improve the model convergence and accuracy. Extensive experiments demonstrate that PhyMPGN is capable of accurately predicting various types of spatiotemporal dynamics on coarse unstructured meshes, consistently achieves the state-of-the-art results, and outperforms other baselines with considerable gains.
Authors: Guoxia Wang, Jinle Zeng, Xiyuan Xiao, Siming Wu, Jiabin Yang, Lujing Zheng, Zeyu Chen, Jiang Bian, Dianhai Yu, Haifeng Wang
Abstract: The computational and memory demands of vanilla attention scale quadratically with the sequence length $N$, posing significant challenges for processing long sequences in Transformer models. FlashAttention alleviates these challenges by eliminating the $O(N^2)$ memory dependency and reducing attention latency through IO-aware memory optimizations. However, its native support for certain attention mask types is limited, and it does not inherently accommodate more complex masking requirements. Previous approaches resort to using dense masks with $O(N^2)$ memory complexity, leading to inefficiencies. In this paper, we propose FlashMask, an extension of FlashAttention that introduces a column-wise sparse representation of attention masks. This approach efficiently represents a wide range of mask types and facilitates the development of optimized kernel implementations. By adopting this novel representation, FlashMask achieves linear memory complexity $O(N)$, suitable for modeling long-context sequences. Moreover, this representation enables kernel optimizations that eliminate unnecessary computations by leveraging sparsity in the attention mask, without sacrificing computational accuracy, resulting in higher computational efficiency. We evaluate FlashMask's performance in fine-tuning and alignment training of LLMs such as SFT, LoRA, DPO, and RM. FlashMask achieves significant throughput improvements, with end-to-end speedups ranging from 1.65x to 3.22x compared to existing FlashAttention dense method. Additionally, our kernel-level comparisons demonstrate that FlashMask surpasses the latest counterpart, FlexAttention, by 12.1% to 60.7% in terms of kernel TFLOPs/s, achieving 37.8% to 62.3% of the theoretical maximum FLOPs/s on the A100 GPU. The code is open-sourced on PaddlePaddle and integrated into PaddleNLP, supporting models with over 100 billion parameters for contexts up to 128K tokens.
Authors: Emanuele Zappala
Abstract: We present an algorithm for learning operators between Banach spaces, based on the use of Leray-Schauder mappings to learn a finite-dimensional approximation of compact subspaces. We show that the resulting method is a universal approximator of (possibly nonlinear) operators. We demonstrate the efficiency of the approach on two benchmark datasets showing it achieves results comparable to state of the art models.
Authors: Shengyu Feng, Xiang Kong, Shuang Ma, Aonan Zhang, Dong Yin, Chong Wang, Ruoming Pang, Yiming Yang
Abstract: Augmenting the multi-step reasoning abilities of Large Language Models (LLMs) has been a persistent challenge. Recently, verification has shown promise in improving solution consistency by evaluating generated outputs. However, current verification approaches suffer from sampling inefficiencies, requiring a large number of samples to achieve satisfactory performance. Additionally, training an effective verifier often depends on extensive process supervision, which is costly to acquire. In this paper, we address these limitations by introducing a novel verification method based on Twisted Sequential Monte Carlo (TSMC). TSMC sequentially refines its sampling effort to focus exploration on promising candidates, resulting in more efficient generation of high-quality solutions. We apply TSMC to LLMs by estimating the expected future rewards at partial solutions. This approach results in a more straightforward training target that eliminates the need for step-wise human annotations. We empirically demonstrate the advantages of our method across multiple math benchmarks, and also validate our theoretical analysis of both our approach and existing verification methods.
Authors: Minh Le, Chau Nguyen, Huy Nguyen, Quyen Tran, Trung Le, Nhat Ho
Abstract: Prompt-based techniques, such as prompt-tuning and prefix-tuning, have gained prominence for their efficiency in fine-tuning large pre-trained models. Despite their widespread adoption, the theoretical foundations of these methods remain limited. For instance, in prefix-tuning, we observe that a key factor in achieving performance parity with full fine-tuning lies in the reparameterization strategy. However, the theoretical principles underpinning the effectiveness of this approach have yet to be thoroughly examined. Our study demonstrates that reparameterization is not merely an engineering trick but is grounded in deep theoretical foundations. Specifically, we show that the reparameterization strategy implicitly encodes a shared structure between prefix key and value vectors. Building on recent insights into the connection between prefix-tuning and mixture of experts models, we further illustrate that this shared structure significantly improves sample efficiency in parameter estimation compared to non-shared alternatives. The effectiveness of prefix-tuning across diverse tasks is empirically confirmed to be enhanced by the shared structure, through extensive experiments in both visual and language domains. Additionally, we uncover similar structural benefits in prompt-tuning, offering new perspectives on its success. Our findings provide theoretical and empirical contributions, advancing the understanding of prompt-based methods and their underlying mechanisms.
Authors: Hyunwoo Lee, Hayoung Choi, Hyunju Kim
Abstract: As a neural network's depth increases, it can improve generalization performance. However, training deep networks is challenging due to gradient and signal propagation issues. To address these challenges, extensive theoretical research and various methods have been introduced. Despite these advances, effective weight initialization methods for tanh neural networks remain insufficiently investigated. This paper presents a novel weight initialization method for neural networks with tanh activation function. Based on an analysis of the fixed points of the function $\tanh(ax)$, the proposed method aims to determine values of $a$ that mitigate activation saturation. A series of experiments on various classification datasets and physics-informed neural networks demonstrates that the proposed method outperforms Xavier initialization methods~(with or without normalization) in terms of robustness across different network sizes, data efficiency, and convergence speed. Code is available at https://github.com/1HyunwooLee/Tanh-Init
Authors: Tianchi Xie, Jiangning Zhu, Guozu Ma, Minzhi Lin, Wei Chen, Weikai Yang, Shixia Liu
Abstract: Sample selection improves the efficiency and effectiveness of machine learning models by providing informative and representative samples. Typically, samples can be modeled as a sample graph, where nodes are samples and edges represent their similarities. Most existing methods are based on local information, such as the training difficulty of samples, thereby overlooking global information, such as connectivity patterns. This oversight can result in suboptimal selection because global information is crucial for ensuring that the selected samples well represent the structural properties of the graph. To address this issue, we employ structural entropy to quantify global information and losslessly decompose it from the whole graph to individual nodes using the Shapley value. Based on the decomposition, we present $\textbf{S}$tructural-$\textbf{E}$ntropy-based sample $\textbf{S}$election ($\textbf{SES}$), a method that integrates both global and local information to select informative and representative samples. SES begins by constructing a $k$NN-graph among samples based on their similarities. It then measures sample importance by combining structural entropy (global metric) with training difficulty (local metric). Finally, SES applies importance-biased blue noise sampling to select a set of diverse and representative samples. Comprehensive experiments on three learning scenarios -- supervised learning, active learning, and continual learning -- clearly demonstrate the effectiveness of our method.
Authors: Rub\'en Ballester, Ernst R\"oell, Daniel B\=in Schmid, Mathieu Alain, Sergio Escalera, Carles Casacuberta, Bastian Rieck
Abstract: The rising interest in leveraging higher-order interactions present in complex systems has led to a surge in more expressive models exploiting higher-order structures in the data, especially in topological deep learning (TDL), which designs neural networks on higher-order domains such as simplicial complexes. However, progress in this field is hindered by the scarcity of datasets for benchmarking these architectures. To address this gap, we introduce MANTRA, the first large-scale, diverse, and intrinsically higher-order dataset for benchmarking higher-order models, comprising over 43,000 and 250,000 triangulations of surfaces and three-dimensional manifolds, respectively. With MANTRA, we assess several graph- and simplicial complex-based models on three topological classification tasks. We demonstrate that while simplicial complex-based neural networks generally outperform their graph-based counterparts in capturing simple topological invariants, they also struggle, suggesting a rethink of TDL. Thus, MANTRA serves as a benchmark for assessing and advancing topological methods, leading the way for more effective higher-order models.
Authors: Rayhan Zirvi, Bahareh Tolooshams, Anima Anandkumar
Abstract: Recent advancements in diffusion models have been effective in learning data priors for solving inverse problems. They leverage diffusion sampling steps for inducing a data prior while using a measurement guidance gradient at each step to impose data consistency. For general inverse problems, approximations are needed when an unconditionally trained diffusion model is used since the measurement likelihood is intractable, leading to inaccurate posterior sampling. In other words, due to their approximations, these methods fail to preserve the generation process on the data manifold defined by the diffusion prior, leading to artifacts in applications such as image restoration. To enhance the performance and robustness of diffusion models in solving inverse problems, we propose Diffusion State-Guided Projected Gradient (DiffStateGrad), which projects the measurement gradient onto a subspace that is a low-rank approximation of an intermediate state of the diffusion process. DiffStateGrad, as a module, can be added to a wide range of diffusion-based inverse solvers to improve the preservation of the diffusion process on the prior manifold and filter out artifact-inducing components. We highlight that DiffStateGrad improves the robustness of diffusion models in terms of the choice of measurement guidance step size and noise while improving the worst-case performance. Finally, we demonstrate that DiffStateGrad improves upon the state-of-the-art on linear and nonlinear image restoration inverse problems. Our code is available at https://github.com/neuraloperator/DiffStateGrad.
Authors: Yinuo Ren, Haoxuan Chen, Grant M. Rotskoff, Lexing Ying
Abstract: Discrete diffusion models have gained increasing attention for their ability to model complex distributions with tractable sampling and inference. However, the error analysis for discrete diffusion models remains less well-understood. In this work, we propose a comprehensive framework for the error analysis of discrete diffusion models based on L\'evy-type stochastic integrals. By generalizing the Poisson random measure to that with a time-independent and state-dependent intensity, we rigorously establish a stochastic integral formulation of discrete diffusion models and provide the corresponding change of measure theorems that are intriguingly analogous to It\^o integrals and Girsanov's theorem for their continuous counterparts. Our framework unifies and strengthens the current theoretical results on discrete diffusion models and obtains the first error bound for the $\tau$-leaping scheme in KL divergence. With error sources clearly identified, our analysis gives new insight into the mathematical properties of discrete diffusion models and offers guidance for the design of efficient and accurate algorithms for real-world discrete diffusion model applications.
Authors: Sijin Chen, Omar Hagrass, Jason M. Klusowski
Abstract: Decoding strategies play a pivotal role in text generation for modern language models, yet a puzzling gap divides theory and practice. Surprisingly, strategies that should intuitively be optimal, such as Maximum a Posteriori (MAP), often perform poorly in practice. Meanwhile, popular heuristic approaches like Top-$k$ and Nucleus sampling, which employ truncation and normalization of the conditional next-token probabilities, have achieved great empirical success but lack theoretical justifications. In this paper, we propose Decoding Game, a comprehensive theoretical framework which reimagines text generation as a two-player zero-sum game between Strategist, who seeks to produce text credible in the true distribution, and Nature, who distorts the true distribution adversarially. After discussing the decomposibility of multi-step generation, we derive the optimal strategy in closed form for one-step Decoding Game. It is shown that the adversarial Nature imposes an implicit regularization on likelihood maximization, and truncation-normalization methods are first-order approximations to the optimal strategy under this regularization. Additionally, by generalizing the objective and parameters of Decoding Game, near-optimal strategies encompass diverse methods such as greedy search, temperature scaling, and hybrids thereof. Numerical experiments are conducted to complement our theoretical analysis.
Authors: Alexander Atanasov, Alexandru Meterez, James B. Simon, Cengiz Pehlevan
Abstract: We consider neural networks (NNs) where the final layer is down-scaled by a fixed hyperparameter $\gamma$. Recent work has identified $\gamma$ as controlling the strength of feature learning. As $\gamma$ increases, network evolution changes from "lazy" kernel dynamics to "rich" feature-learning dynamics, with a host of associated benefits including improved performance on common tasks. In this work, we conduct a thorough empirical investigation of the effect of scaling $\gamma$ across a variety of models and datasets in the online training setting. We first examine the interaction of $\gamma$ with the learning rate $\eta$, identifying several scaling regimes in the $\gamma$-$\eta$ plane which we explain theoretically using a simple model. We find that the optimal learning rate $\eta^*$ scales non-trivially with $\gamma$. In particular, $\eta^* \propto \gamma^2$ when $\gamma \ll 1$ and $\eta^* \propto \gamma^{2/L}$ when $\gamma \gg 1$ for a feed-forward network of depth $L$. Using this optimal learning rate scaling, we proceed with an empirical study of the under-explored "ultra-rich" $\gamma \gg 1$ regime. We find that networks in this regime display characteristic loss curves, starting with a long plateau followed by a drop-off, sometimes followed by one or more additional staircase steps. We find networks of different large $\gamma$ values optimize along similar trajectories up to a reparameterization of time. We further find that optimal online performance is often found at large $\gamma$ and could be missed if this hyperparameter is not tuned. Our findings indicate that analytical study of the large-$\gamma$ limit may yield useful insights into the dynamics of representation learning in performant models.
Authors: Yong Liu, Guo Qin, Xiangdong Huang, Jianmin Wang, Mingsheng Long
Abstract: We present Timer-XL, a causal Transformer for unified time series forecasting. To uniformly predict multidimensional time series, we generalize next token prediction, predominantly adopted for 1D token sequences, to multivariate next token prediction. The paradigm formulates various forecasting tasks as a long-context prediction problem. We opt for decoder-only Transformers that capture causal dependencies from varying-length contexts for unified forecasting, making predictions on non-stationary univariate time series, multivariate series with complicated dynamics and correlations, as well as covariate-informed contexts that include exogenous variables. Technically, we propose a universal TimeAttention to capture fine-grained intra- and inter-series dependencies of flattened time series tokens (patches), which is further enhanced by deft position embedding for temporal causality and variable equivalence. Timer-XL achieves state-of-the-art performance across task-specific forecasting benchmarks through a unified approach. Based on large-scale pre-training, Timer-XL achieves state-of-the-art zero-shot performance, making it a promising architecture for pre-trained time series models. Code is available at this repository: https://github.com/thuml/Timer-XL.
Authors: Haokun Chen, Hang Li, Yao Zhang, Jinhe Bi, Gengyuan Zhang, Yueqi Zhang, Philip Torr, Jindong Gu, Denis Krompass, Volker Tresp
Abstract: One-Shot Federated Learning (OSFL), a special decentralized machine learning paradigm, has recently gained significant attention. OSFL requires only a single round of client data or model upload, which reduces communication costs and mitigates privacy threats compared to traditional FL. Despite these promising prospects, existing methods face challenges due to client data heterogeneity and limited data quantity when applied to real-world OSFL systems. Recently, Latent Diffusion Models (LDM) have shown remarkable advancements in synthesizing high-quality images through pretraining on large-scale datasets, thereby presenting a potential solution to overcome these issues. However, directly applying pretrained LDM to heterogeneous OSFL results in significant distribution shifts in synthetic data, leading to performance degradation in classification models trained on such data. This issue is particularly pronounced in rare domains, such as medical imaging, which are underrepresented in LDM's pretraining data. To address this challenge, we propose Federated Bi-Level Personalization (FedBiP), which personalizes the pretrained LDM at both instance-level and concept-level. Hereby, FedBiP synthesizes images following the client's local data distribution without compromising the privacy regulations. FedBiP is also the first approach to simultaneously address feature space heterogeneity and client data scarcity in OSFL. Our method is validated through extensive experiments on three OSFL benchmarks with feature space heterogeneity, as well as on challenging medical and satellite image datasets with label heterogeneity. The results demonstrate the effectiveness of FedBiP, which substantially outperforms other OSFL methods.
Authors: Bingrui Li, Wei Huang, Andi Han, Zhanpeng Zhou, Taiji Suzuki, Jun Zhu, Jianfei Chen
Abstract: The Adam optimizer is widely used for transformer optimization in practice, which makes understanding the underlying optimization mechanisms an important problem. However, due to the Adam's complexity, theoretical analysis of how it optimizes transformers remains a challenging task. Fortunately, Sign Gradient Descent (SignGD) serves as an effective surrogate for Adam. Despite its simplicity, theoretical understanding of how SignGD optimizes transformers still lags behind. In this work, we study how SignGD optimizes a two-layer transformer -- consisting of a softmax attention layer with trainable query-key parameterization followed by a linear layer -- on a linearly separable noisy dataset. We identify four stages in the training dynamics, each exhibiting intriguing behaviors. Based on the training dynamics, we prove the fast convergence but poor generalization of the learned transformer on the noisy dataset. We also show that Adam behaves similarly to SignGD in terms of both optimization and generalization in this setting. Additionally, we find that the poor generalization of SignGD is not solely due to data noise, suggesting that both SignGD and Adam requires high-quality data for real-world tasks. Finally, experiments on synthetic and real-world datasets empirically support our theoretical results.
Authors: Leo Zhang, Kianoosh Ashouritaklimi, Yee Whye Teh, Rob Cornish
Abstract: We propose SymDiff, a method for constructing equivariant diffusion models using the framework of stochastic symmetrisation. SymDiff resembles a learned data augmentation that is deployed at sampling time, and is lightweight, computationally efficient, and easy to implement on top of arbitrary off-the-shelf models. In contrast to previous work, SymDiff typically does not require any neural network components that are intrinsically equivariant, avoiding the need for complex parameterisations or the use of higher-order geometric features. Instead, our method can leverage highly scalable modern architectures as drop-in replacements for these more constrained alternatives. We show that this additional flexibility yields significant empirical benefit for $\mathrm{E}(3)$-equivariant molecular generation. To the best of our knowledge, this is the first application of symmetrisation to generative modelling, suggesting its potential in this domain more generally.
Authors: Nikita Balagansky, Ian Maksimov, Daniil Gavrilov
Abstract: Understanding how features evolve across layers in deep neural networks is a fundamental challenge in mechanistic interpretability, particularly due to polysemanticity and feature superposition. While Sparse Autoencoders (SAEs) have been used to extract interpretable features from individual layers, aligning these features across layers has remained an open problem. In this paper, we introduce SAE Match, a novel, data-free method for aligning SAE features across different layers of a neural network. Our approach involves matching features by minimizing the mean squared error between the folded parameters of SAEs, a technique that incorporates activation thresholds into the encoder and decoder weights to account for differences in feature scales. Through extensive experiments on the Gemma 2 language model, we demonstrate that our method effectively captures feature evolution across layers, improving feature matching quality. We also show that features persist over several layers and that our approach can approximate hidden states across layers. Our work advances the understanding of feature dynamics in neural networks and provides a new tool for mechanistic interpretability studies.
Authors: Katharine Daly, Hubert Eichner, Peter Kairouz, H. Brendan McMahan, Daniel Ramage, Zheng Xu
Abstract: Federated Learning (FL) is a machine learning technique that enables multiple entities to collaboratively learn a shared model without exchanging their local data. Over the past decade, FL systems have achieved substantial progress, scaling to millions of devices across various learning domains while offering meaningful differential privacy (DP) guarantees. Production systems from organizations like Google, Apple, and Meta demonstrate the real-world applicability of FL. However, key challenges remain, including verifying server-side DP guarantees and coordinating training across heterogeneous devices, limiting broader adoption. Additionally, emerging trends such as large (multi-modal) models and blurred lines between training, inference, and personalization challenge traditional FL frameworks. In response, we propose a redefined FL framework that prioritizes privacy principles rather than rigid definitions. We also chart a path forward by leveraging trusted execution environments and open-source ecosystems to address these challenges and facilitate future advancements in FL.
Authors: Yan Scholten, Stephan G\"unnemann
Abstract: Conformal prediction provides model-agnostic and distribution-free uncertainty quantification through prediction sets that are guaranteed to include the ground truth with any user-specified probability. Yet, conformal prediction is not reliable under poisoning attacks where adversaries manipulate both training and calibration data, which can significantly alter prediction sets in practice. As a solution, we propose reliable prediction sets (RPS): the first efficient method for constructing conformal prediction sets with provable reliability guarantees under poisoning. To ensure reliability under training poisoning, we introduce smoothed score functions that reliably aggregate predictions of classifiers trained on distinct partitions of the training data. To ensure reliability under calibration poisoning, we construct multiple prediction sets, each calibrated on distinct subsets of the calibration data. We then aggregate them into a majority prediction set, which includes a class only if it appears in a majority of the individual sets. Both proposed aggregations mitigate the influence of datapoints in the training and calibration data on the final prediction set. We experimentally validate our approach on image classification tasks, achieving strong reliability while maintaining utility and preserving coverage on clean data. Overall, our approach represents an important step towards more trustworthy uncertainty quantification in the presence of data poisoning.
Authors: Binghui Li, Zhixuan Pan, Kaifeng Lyu, Jian Li
Abstract: In this work, we investigate a particular implicit bias in gradient descent training, which we term "Feature Averaging," and argue that it is one of the principal factors contributing to the non-robustness of deep neural networks. We show that, even when multiple discriminative features are present in the input data, neural networks trained by gradient descent tend to rely on an average (or a certain combination) of these features for classification, rather than distinguishing and leveraging each feature individually. Specifically, we provide a detailed theoretical analysis of the training dynamics of two-layer ReLU networks on a binary classification task, where the data distribution consists of multiple clusters with mutually orthogonal centers. We rigorously prove that gradient descent biases the network towards feature averaging, where the weights of each hidden neuron represent an average of the cluster centers (each corresponding to a distinct feature), thereby making the network vulnerable to input perturbations aligned with the negative direction of the averaged features. On the positive side, we demonstrate that this vulnerability can be mitigated through more granular supervision. In particular, we prove that a two-layer ReLU network can achieve optimal robustness when trained to classify individual features rather than merely the original binary classes. Finally, we validate our theoretical findings with experiments on synthetic datasets, MNIST, and CIFAR-10, and confirm the prevalence of feature averaging and its impact on adversarial robustness. We hope these theoretical and empirical insights deepen the understanding of how gradient descent shapes feature learning and adversarial robustness, and how more detailed supervision can enhance robustness.
Authors: Cheng Lu, Yang Song
Abstract: Consistency models (CMs) are a powerful class of diffusion-based generative models optimized for fast sampling. Most existing CMs are trained using discretized timesteps, which introduce additional hyperparameters and are prone to discretization errors. While continuous-time formulations can mitigate these issues, their success has been limited by training instability. To address this, we propose a simplified theoretical framework that unifies previous parameterizations of diffusion models and CMs, identifying the root causes of instability. Based on this analysis, we introduce key improvements in diffusion process parameterization, network architecture, and training objectives. These changes enable us to train continuous-time CMs at an unprecedented scale, reaching 1.5B parameters on ImageNet 512x512. Our proposed training algorithm, using only two sampling steps, achieves FID scores of 2.06 on CIFAR-10, 1.48 on ImageNet 64x64, and 1.88 on ImageNet 512x512, narrowing the gap in FID scores with the best existing diffusion models to within 10%.
Authors: Alan T. L. Bacellar, Zachary Susskind, Mauricio Breternitz Jr., Eugene John, Lizy K. John, Priscila M. V. Lima, Felipe M. G. Fran\c{c}a
Abstract: We introduce the Differentiable Weightless Neural Network (DWN), a model based on interconnected lookup tables. Training of DWNs is enabled by a novel Extended Finite Difference technique for approximate differentiation of binary values. We propose Learnable Mapping, Learnable Reduction, and Spectral Regularization to further improve the accuracy and efficiency of these models. We evaluate DWNs in three edge computing contexts: (1) an FPGA-based hardware accelerator, where they demonstrate superior latency, throughput, energy efficiency, and model area compared to state-of-the-art solutions, (2) a low-power microcontroller, where they achieve preferable accuracy to XGBoost while subject to stringent memory constraints, and (3) ultra-low-cost chips, where they consistently outperform small models in both accuracy and projected hardware area. DWNs also compare favorably against leading approaches for tabular datasets, with higher average rank. Overall, our work positions DWNs as a pioneering solution for edge-compatible high-throughput neural networks.
Authors: Bo Chen, Xiaoyu Li, Yingyu Liang, Zhenmei Shi, Zhao Song
Abstract: In-context learning has been recognized as a key factor in the success of Large Language Models (LLMs). It refers to the model's ability to learn patterns on the fly from provided in-context examples in the prompt during inference. Previous studies have demonstrated that the Transformer architecture used in LLMs can implement a single-step gradient descent update by processing in-context examples in a single forward pass. Recent work has further shown that, during in-context learning, a looped Transformer can implement multi-step gradient descent updates in forward passes. However, their theoretical results require an exponential number of in-context examples, $n = \exp(\Omega(T))$, where $T$ is the number of loops or passes, to achieve a reasonably low error. In this paper, we study linear looped Transformers in-context learning on linear vector generation tasks. We show that linear looped Transformers can implement multi-step gradient descent efficiently for in-context learning. Our results demonstrate that as long as the input data has a constant condition number, e.g., $n = O(d)$, the linear looped Transformers can achieve a small error by multi-step gradient descent during in-context learning. Furthermore, our preliminary experiments validate our theoretical analysis. Our findings reveal that the Transformer architecture possesses a stronger in-context learning capability than previously understood, offering new insights into the mechanisms behind LLMs and potentially guiding the better design of efficient inference algorithms for LLMs.
Authors: Rong-Xi Tan, Ke Xue, Shen-Huan Lyu, Haopu Shang, Yao Wang, Yaoyuan Wang, Sheng Fu, Chao Qian
Abstract: Offline model-based optimization (MBO) aims to identify a design that maximizes a black-box function using only a fixed, pre-collected dataset of designs and their corresponding scores. A common approach in offline MBO is to train a regression-based surrogate model by minimizing mean squared error (MSE) and then find the best design within this surrogate model by different optimizers (e.g., gradient ascent). However, a critical challenge is the risk of out-of-distribution errors, i.e., the surrogate model may typically overestimate the scores and mislead the optimizers into suboptimal regions. Prior works have attempted to address this issue in various ways, such as using regularization techniques and ensemble learning to enhance the robustness of the model, but it still remains. In this paper, we argue that regression models trained with MSE are not well-aligned with the primary goal of offline MBO, which is to select promising designs rather than to predict their scores precisely. Notably, if a surrogate model can maintain the order of candidate designs based on their relative score relationships, it can produce the best designs even without precise predictions. To validate it, we conduct experiments to compare the relationship between the quality of the final designs and MSE, finding that the correlation is really very weak. In contrast, a metric that measures order-maintaining quality shows a significantly stronger correlation. Based on this observation, we propose learning a ranking-based model that leverages learning to rank techniques to prioritize promising designs based on their relative scores. We show that the generalization error on ranking loss can be well bounded. Empirical results across diverse tasks demonstrate the superior performance of our proposed ranking-based models than twenty existing methods.
Authors: Sajad Movahedi, Antonio Orvieto, Seyed-Mohsen Moosavi-Dezfooli
Abstract: In this paper, we propose the $\textit{geometric invariance hypothesis (GIH)}$, which argues that the input space curvature of a neural network remains invariant under transformation in certain architecture-dependent directions during training. We investigate a simple, non-linear binary classification problem residing on a plane in a high dimensional space and observe that$\unicode{x2014}$unlike MPLs$\unicode{x2014}$ResNets fail to generalize depending on the orientation of the plane. Motivated by this example, we define a neural network's $\textbf{average geometry}$ and $\textbf{average geometry evolution}$ as compact $\textit{architecture-dependent}$ summaries of the model's input-output geometry and its evolution during training. By investigating the average geometry evolution at initialization, we discover that the geometry of a neural network evolves according to the data covariance projected onto its average geometry. This means that the geometry only changes in a subset of the input space when the average geometry is low-rank, such as in ResNets. This causes an architecture-dependent invariance property in the input space curvature, which we dub GIH. Finally, we present extensive experimental results to observe the consequences of GIH and how it relates to generalization in neural networks.
Authors: Zihao Zhou, Yang Liu, Xianghong Xu, Qian Li
Abstract: Temporal graph clustering is a complex task that involves discovering meaningful structures in dynamic graphs where relationships and entities change over time. Existing methods typically require centralized data collection, which poses significant privacy and communication challenges. In this work, we introduce a novel Federated Temporal Graph Clustering (FTGC) framework that enables decentralized training of graph neural networks (GNNs) across multiple clients, ensuring data privacy throughout the process. Our approach incorporates a temporal aggregation mechanism to effectively capture the evolution of graph structures over time and a federated optimization strategy to collaboratively learn high-quality clustering representations. By preserving data privacy and reducing communication overhead, our framework achieves competitive performance on temporal graph datasets, making it a promising solution for privacy-sensitive, real-world applications involving dynamic data.
Authors: Zhaoyang Wang, Weilei He, Zhiyuan Liang, Xuchao Zhang, Chetan Bansal, Ying Wei, Weitong Zhang, Huaxiu Yao
Abstract: Recent self-rewarding large language models (LLM) have successfully applied LLM-as-a-Judge to iteratively improve the alignment performance without the need of human annotations for preference data. These methods commonly utilize the same LLM to act as both the policy model (which generates responses) and the reward model (which scores and ranks those responses). The ranked responses are then used as preference pairs to train the LLM via direct alignment technologies (e.g. DPO). However, it is noteworthy that throughout this process, there is no guarantee of accuracy in the rewarding and ranking, which is critical for ensuring accurate rewards and high-quality preference data. Empirical results from relatively small LLMs (e.g., 7B parameters) also indicate that improvements from self-rewarding may diminish after several iterations in certain situations, which we hypothesize is due to accumulated bias in the reward system. This bias can lead to unreliable preference data for training the LLM. To address this issue, we first formulate and analyze the generalized iterative preference fine-tuning framework for self-rewarding language model. We then introduce the regularization to this generalized framework to mitigate the overconfident preference labeling in the self-rewarding process. Based on this theoretical insight, we propose a Consistency Regularized sElf-rewarding lAnguage Model (CREAM) that leverages the consistency of rewards across different iterations to regularize the self-rewarding training, helping the model to learn from more reliable preference data. With this explicit regularization, our empirical results demonstrate the superiority of CREAM in improving both reward consistency and alignment performance. The code is publicly available at https://github.com/Raibows/CREAM.
Authors: Peng Xia, Kangyu Zhu, Haoran Li, Tianze Wang, Weijia Shi, Sheng Wang, Linjun Zhang, James Zou, Huaxiu Yao
Abstract: Artificial Intelligence (AI) has demonstrated significant potential in healthcare, particularly in disease diagnosis and treatment planning. Recent progress in Medical Large Vision-Language Models (Med-LVLMs) has opened up new possibilities for interactive diagnostic tools. However, these models often suffer from factual hallucination, which can lead to incorrect diagnoses. Fine-tuning and retrieval-augmented generation (RAG) have emerged as methods to address these issues. However, the amount of high-quality data and distribution shifts between training data and deployment data limit the application of fine-tuning methods. Although RAG is lightweight and effective, existing RAG-based approaches are not sufficiently general to different medical domains and can potentially cause misalignment issues, both between modalities and between the model and the ground truth. In this paper, we propose a versatile multimodal RAG system, MMed-RAG, designed to enhance the factuality of Med-LVLMs. Our approach introduces a domain-aware retrieval mechanism, an adaptive retrieved contexts selection method, and a provable RAG-based preference fine-tuning strategy. These innovations make the RAG process sufficiently general and reliable, significantly improving alignment when introducing retrieved contexts. Experimental results across five medical datasets (involving radiology, ophthalmology, pathology) on medical VQA and report generation demonstrate that MMed-RAG can achieve an average improvement of 43.8% in the factual accuracy of Med-LVLMs. Our data and code are available in https://github.com/richard-peng-xia/MMed-RAG.
Authors: Runzhe Wu, Yiding Chen, Gokul Swamy, Kiant\'e Brantley, Wen Sun
Abstract: Adversarial Imitation Learning is traditionally framed as a two-player zero-sum game between a learner and an adversarially chosen cost function, and can therefore be thought of as the sequential generalization of a Generative Adversarial Network (GAN). However, in recent years, diffusion models have emerged as a non-adversarial alternative to GANs that merely require training a score function via regression, yet produce generations of higher quality. In response, we investigate how to lift insights from diffusion modeling to the sequential setting. We propose diffusing states and performing score-matching along diffused states to measure the discrepancy between the expert's and learner's states. Thus, our approach only requires training score functions to predict noises via standard regression, making it significantly easier and more stable to train than adversarial methods. Theoretically, we prove first- and second-order instance-dependent bounds with linear scaling in the horizon, proving that our approach avoids the compounding errors that stymie offline approaches to imitation learning. Empirically, we show our approach outperforms both GAN-style imitation learning baselines and discriminator-free imitation learning baselines across various continuous control problems, including complex tasks like controlling humanoids to walk, sit, crawl, and navigate through obstacles.
Authors: Seong Ho Pahng, Sahand Hormoz
Abstract: Graph Neural Networks (GNNs) traditionally employ a message-passing mechanism that resembles diffusion over undirected graphs, which often leads to homogenization of node features and reduced discriminative power in tasks such as node classification. Our key insight for addressing this limitation is to assign fuzzy edge directions -- that can vary continuously from node $i$ pointing to node $j$ to vice versa -- to the edges of a graph so that features can preferentially flow in one direction between nodes to enable long-range information transmission across the graph. We also introduce a novel complex-valued Laplacian for directed graphs with fuzzy edges where the real and imaginary parts represent information flow in opposite directions. Using this Laplacian, we propose a general framework, called Continuous Edge Direction (CoED) GNN, for learning on graphs with fuzzy edges and prove its expressivity limits using a generalization of the Weisfeiler-Leman (WL) graph isomorphism test for directed graphs with fuzzy edges. Our architecture aggregates neighbor features scaled by the learned edge directions and processes the aggregated messages from in-neighbors and out-neighbors separately alongside the self-features of the nodes. Since continuous edge directions are differentiable, they can be learned jointly with the GNN weights via gradient-based optimization. CoED GNN is particularly well-suited for graph ensemble data where the graph structure remains fixed but multiple realizations of node features are available, such as in gene regulatory networks, web connectivity graphs, and power grids. We demonstrate through extensive experiments on both synthetic and real graph ensemble datasets that learning continuous edge directions significantly improves performance both for undirected and directed graphs compared with existing methods.
Authors: Timofei Gritsaev, Nikita Morozov, Sergey Samsonov, Daniil Tiapkin
Abstract: Generative Flow Networks (GFlowNets) are a family of generative models that learn to sample objects with probabilities proportional to a given reward function. The key concept behind GFlowNets is the use of two stochastic policies: a forward policy, which incrementally constructs compositional objects, and a backward policy, which sequentially deconstructs them. Recent results show a close relationship between GFlowNet training and entropy-regularized reinforcement learning (RL) problems with a particular reward design. However, this connection applies only in the setting of a fixed backward policy, which might be a significant limitation. As a remedy to this problem, we introduce a simple backward policy optimization algorithm that involves direct maximization of the value function in an entropy-regularized Markov Decision Process (MDP) over intermediate rewards. We provide an extensive experimental evaluation of the proposed approach across various benchmarks in combination with both RL and GFlowNet algorithms and demonstrate its faster convergence and mode discovery in complex environments.
Authors: Shiyu Wang, Jiawei Li, Xiaoming Shi, Zhou Ye, Baichuan Mo, Wenze Lin, Shengtong Ju, Zhixuan Chu, Ming Jin
Abstract: Time series analysis plays a critical role in numerous applications, supporting tasks such as forecasting, classification, anomaly detection, and imputation. In this work, we present the time series pattern machine (TSPM), a model designed to excel in a broad range of time series tasks through powerful representation and pattern extraction capabilities. Traditional time series models often struggle to capture universal patterns, limiting their effectiveness across diverse tasks. To address this, we define multiple scales in the time domain and various resolutions in the frequency domain, employing various mixing strategies to extract intricate, task-adaptive time series patterns. Specifically, we introduce a general-purpose TSPM that processes multi-scale time series using (1) multi-resolution time imaging (MRTI), (2) time image decomposition (TID), (3) multi-scale mixing (MCM), and (4) multi-resolution mixing (MRM) to extract comprehensive temporal patterns. MRTI transforms multi-scale time series into multi-resolution time images, capturing patterns across both temporal and frequency domains. TID leverages dual-axis attention to extract seasonal and trend patterns, while MCM hierarchically aggregates these patterns across scales. MRM adaptively integrates all representations across resolutions. This method achieves state-of-the-art performance across 8 time series analytical tasks, consistently surpassing both general-purpose and task-specific models. Our work marks a promising step toward the next generation of TSPMs, paving the way for further advancements in time series analysis.
Authors: Thomas Robert, Mher Safaryan, Ionut-Vlad Modoranu, Dan Alistarh
Abstract: We introduce LDAdam, a memory-efficient optimizer for training large models, that performs adaptive optimization steps within lower dimensional subspaces, while consistently exploring the full parameter space during training. This strategy keeps the optimizer's memory footprint to a fraction of the model size. LDAdam relies on a new projection-aware update rule for the optimizer states that allows for transitioning between subspaces, i.e., estimation of the statistics of the projected gradients. To mitigate the errors due to low-rank projection, LDAdam integrates a new generalized error feedback mechanism, which explicitly accounts for both gradient and optimizer state compression. We prove the convergence of LDAdam under standard assumptions, and show that LDAdam allows for accurate and efficient fine-tuning and pre-training of language models. Code is available at https://github.com/IST-DASLab/LDAdam
Authors: Xiang Cheng, Lawrence Carin, Suvrit Sra
Abstract: We show theoretically and empirically that the linear Transformer, when applied to graph data, can implement algorithms that solve canonical problems such as electric flow and eigenvector decomposition. The Transformer has access to information on the input graph only via the graph's incidence matrix. We present explicit weight configurations for implementing each algorithm, and we bound the constructed Transformers' errors by the errors of the underlying algorithms. Our theoretical findings are corroborated by experiments on synthetic data. Additionally, on a real-world molecular regression task, we observe that the linear Transformer is capable of learning a more effective positional encoding than the default one based on Laplacian eigenvectors. Our work is an initial step towards elucidating the inner-workings of the Transformer for graph data. Code is available at https://github.com/chengxiang/LinearGraphTransformer
Authors: Samy Jelassi, Clara Mohri, David Brandfonbrener, Alex Gu, Nikhil Vyas, Nikhil Anand, David Alvarez-Melis, Yuanzhi Li, Sham M. Kakade, Eran Malach
Abstract: The Mixture-of-Experts (MoE) architecture enables a significant increase in the total number of model parameters with minimal computational overhead. However, it is not clear what performance tradeoffs, if any, exist between MoEs and standard dense transformers. In this paper, we show that as we increase the number of experts (while fixing the number of active parameters), the memorization performance consistently increases while the reasoning capabilities saturate. We begin by analyzing the theoretical limitations of MoEs at reasoning. We prove that there exist graph problems that cannot be solved by any number of experts of a certain width; however, the same task can be easily solved by a dense model with a slightly larger width. On the other hand, we find that on memory-intensive tasks, MoEs can effectively leverage a small number of active parameters with a large number of experts to memorize the data. We empirically validate these findings on synthetic graph problems and memory-intensive closed book retrieval tasks. Lastly, we pre-train a series of MoEs and dense transformers and evaluate them on commonly used benchmarks in math and natural language. We find that increasing the number of experts helps solve knowledge-intensive tasks, but fails to yield the same benefits for reasoning tasks.
Authors: Ihor Neporozhnii, Julien Roy, Emmanuel Bengio, Jason Hartford
Abstract: In drug discovery, highly automated high-throughput laboratories are used to screen a large number of compounds in search of effective drugs. These experiments are expensive, so one might hope to reduce their cost by only experimenting on a subset of the compounds, and predicting the outcomes of the remaining experiments. In this work, we model this scenario as a sequential subset selection problem: we aim to select the smallest set of candidates in order to achieve some desired level of accuracy for the system as a whole. Our key observation is that, if there is heterogeneity in the difficulty of the prediction problem across the input space, selectively obtaining the labels for the hardest examples in the acquisition pool will leave only the relatively easy examples to remain in the inference set, leading to better overall system performance. We call this mechanism inference set design, and propose the use of a confidence-based active learning solution to prune out these challenging examples. Our algorithm includes an explicit stopping criterion that interrupts the acquisition loop when it is sufficiently confident that the system has reached the target performance. Our empirical studies on image and molecular datasets, as well as a real-world large-scale biological assay, show that active learning for inference set design leads to significant reduction in experimental cost while retaining high system performance.
Authors: Vivek F. Farias, Adam D. Jozefiak
Abstract: Plasticity Loss is an increasingly important phenomenon that refers to the empirical observation that as a neural network is continually trained on a sequence of changing tasks, its ability to adapt to a new task diminishes over time. We introduce Self-Normalized Resets (SNR), a simple adaptive algorithm that mitigates plasticity loss by resetting a neuron's weights when evidence suggests its firing rate has effectively dropped to zero. Across a battery of continual learning problems and network architectures, we demonstrate that SNR consistently attains superior performance compared to its competitor algorithms. We also demonstrate that SNR is robust to its sole hyperparameter, its rejection percentile threshold, while competitor algorithms show significant sensitivity. SNR's threshold-based reset mechanism is motivated by a simple hypothesis test that we derive. Seen through the lens of this hypothesis test, competing reset proposals yield suboptimal error rates in correctly detecting inactive neurons, potentially explaining our experimental observations. We also conduct a theoretical investigation of the optimization landscape for the problem of learning a single ReLU. We show that even when initialized adversarially, an idealized version of SNR learns the target ReLU, while regularization-based approaches can fail to learn.
Authors: Guneet S. Dhillon, Xingjian Shi, Yee Whye Teh, Alex Smola
Abstract: Supervised fine-tuning (SFT) and alignment of large language models (LLMs) are key steps in providing a good user experience. However, the concept of an appropriate alignment is inherently application-dependent, and current methods often rely on heuristic choices to drive optimization. In this work, we formulate SFT and alignment as a constrained optimization problem: the LLM is fine-tuned on a task while being required to meet application-specific requirements, without resorting to heuristics. To solve this, we propose Lagrange Large Language Models (L3Ms), which employ logarithmic barriers to enforce the constraints. This approach allows for the customization of L3Ms across diverse applications while avoiding heuristic-driven processes. We experimentally demonstrate the versatility and efficacy of L3Ms in achieving tailored alignments for various applications.
Authors: Chun-Wei Kong, Luca Laurenti, Jay McMahon, Morteza Lahijanian
Abstract: Stochastic differential equations are commonly used to describe the evolution of stochastic processes. The state uncertainty of such processes is best represented by the probability density function (PDF), whose evolution is governed by the Fokker-Planck partial differential equation (FP-PDE). However, it is generally infeasible to solve the FP-PDE in closed form. In this work, we show that physics-informed neural networks (PINNs) can be trained to approximate the solution PDF. Our main contribution is the analysis of PINN approximation error: we develop a theoretical framework to construct tight error bounds using PINNs. In addition, we derive a practical error bound that can be efficiently constructed with standard training methods. We discuss that this error-bound framework generalizes to approximate solutions of other linear PDEs. Empirical results on nonlinear, high-dimensional, and chaotic systems validate the correctness of our error bounds while demonstrating the scalability of PINNs and their significant computational speedup in obtaining accurate PDF solutions compared to the Monte Carlo approach.
Authors: Michael Matthews, Michael Beukman, Chris Lu, Jakob Foerster
Abstract: While large models trained with self-supervised learning on offline datasets have shown remarkable capabilities in text and image domains, achieving the same generalisation for agents that act in sequential decision problems remains an open challenge. In this work, we take a step towards this goal by procedurally generating tens of millions of 2D physics-based tasks and using these to train a general reinforcement learning (RL) agent for physical control. To this end, we introduce Kinetix: an open-ended space of physics-based RL environments that can represent tasks ranging from robotic locomotion and grasping to video games and classic RL environments, all within a unified framework. Kinetix makes use of our novel hardware-accelerated physics engine Jax2D that allows us to cheaply simulate billions of environment steps during training. Our trained agent exhibits strong physical reasoning capabilities in 2D space, being able to zero-shot solve unseen human-designed environments. Furthermore, fine-tuning this general agent on tasks of interest shows significantly stronger performance than training an RL agent *tabula rasa*. This includes solving some environments that standard RL training completely fails at. We believe this demonstrates the feasibility of large scale, mixed-quality pre-training for online RL and we hope that Kinetix will serve as a useful framework to investigate this further.
Authors: Lukas Miklautz, Timo Klein, Kevin Sidak, Collin Leiber, Thomas Lang, Andrii Shkabrii, Sebastian Tschiatschek, Claudia Plant
Abstract: This work investigates an important phenomenon in centroid-based deep clustering (DC) algorithms: Performance quickly saturates after a period of rapid early gains. Practitioners commonly address early saturation with periodic reclustering, which we demonstrate to be insufficient to address performance plateaus. We call this phenomenon the "reclustering barrier" and empirically show when the reclustering barrier occurs, what its underlying mechanisms are, and how it is possible to Break the Reclustering Barrier with our algorithm BRB. BRB avoids early over-commitment to initial clusterings and enables continuous adaptation to reinitialized clustering targets while remaining conceptually simple. Applying our algorithm to widely-used centroid-based DC algorithms, we show that (1) BRB consistently improves performance across a wide range of clustering benchmarks, (2) BRB enables training from scratch, and (3) BRB performs competitively against state-of-the-art DC algorithms when combined with a contrastive loss. We release our code and pre-trained models at https://github.com/Probabilistic-and-Interactive-ML/breaking-the-reclustering-barrier .
URLs: https://github.com/Probabilistic-and-Interactive-ML/breaking-the-reclustering-barrier
Authors: Mufei Li, Viraj Shitole, Eli Chien, Changhai Man, Zhaodong Wang, Srinivas Sridharan, Ying Zhang, Tushar Krishna, Pan Li
Abstract: Directed acyclic graphs (DAGs) serve as crucial data representations in domains such as hardware synthesis and compiler/program optimization for computing systems. DAG generative models facilitate the creation of synthetic DAGs, which can be used for benchmarking computing systems while preserving intellectual property. However, generating realistic DAGs is challenging due to their inherent directional and logical dependencies. This paper introduces LayerDAG, an autoregressive diffusion model, to address these challenges. LayerDAG decouples the strong node dependencies into manageable units that can be processed sequentially. By interpreting the partial order of nodes as a sequence of bipartite graphs, LayerDAG leverages autoregressive generation to model directional dependencies and employs diffusion models to capture logical dependencies within each bipartite graph. Comparative analyses demonstrate that LayerDAG outperforms existing DAG generative models in both expressiveness and generalization, particularly for generating large-scale DAGs with up to 400 nodes-a critical scenario for system benchmarking. Extensive experiments on both synthetic and real-world flow graphs from various computing platforms show that LayerDAG generates valid DAGs with superior statistical properties and benchmarking performance. The synthetic DAGs generated by LayerDAG enhance the training of ML-based surrogate models, resulting in improved accuracy in predicting performance metrics of real-world DAGs across diverse computing platforms.
Authors: Manuel Gloeckler, Shoji Toyota, Kenji Fukumizu, Jakob H. Macke
Abstract: Amortized simulation-based inference (SBI) methods train neural networks on simulated data to perform Bayesian inference. While this strategy avoids the need for tractable likelihoods, it often requires a large number of simulations and has been challenging to scale to time series data. Scientific simulators frequently emulate real-world dynamics through thousands of single-state transitions over time. We propose an SBI approach that can exploit such Markovian simulators by locally identifying parameters consistent with individual state transitions. We then compose these local results to obtain a posterior over parameters that align with the entire time series observation. We focus on applying this approach to neural posterior score estimation but also show how it can be applied, e.g., to neural likelihood (ratio) estimation. We demonstrate that our approach is more simulation-efficient than directly estimating the global posterior on several synthetic benchmark tasks and simulators used in ecology and epidemiology. Finally, we validate scalability and simulation efficiency of our approach by applying it to a high-dimensional Kolmogorov flow simulator with around one million data dimensions.
Authors: Sagar Shrestha, Xiao Fu
Abstract: Understanding identifiability of latent content and style variables from unaligned multi-domain data is essential for tasks such as domain translation and data generation. Existing works on content-style identification were often developed under somewhat stringent conditions, e.g., that all latent components are mutually independent and that the dimensions of the content and style variables are known. We introduce a new analytical framework via cross-domain \textit{latent distribution matching} (LDM), which establishes content-style identifiability under substantially more relaxed conditions. Specifically, we show that restrictive assumptions such as component-wise independence of the latent variables can be removed. Most notably, we prove that prior knowledge of the content and style dimensions is not necessary for ensuring identifiability, if sparsity constraints are properly imposed onto the learned latent representations. Bypassing the knowledge of the exact latent dimension has been a longstanding aspiration in unsupervised representation learning -- our analysis is the first to underpin its theoretical and practical viability. On the implementation side, we recast the LDM formulation into a regularized multi-domain GAN loss with coupled latent variables. We show that the reformulation is equivalent to LDM under mild conditions -- yet requiring considerably less computational resource. Experiments corroborate with our theoretical claims.
Authors: Thea Br\"usch, Kristoffer K. Wickstr{\o}m, Mikkel N. Schmidt, Robert Jenssen, Tommy S. Alstr{\o}m
Abstract: State-of-the-art methods for explaining predictions from time series involve learning an instance-wise saliency mask for each time step; however, many types of time series are difficult to interpret in the time domain, due to the inherently complex nature of the data. Instead, we propose to view time series explainability as saliency maps over interpretable parts, leaning on established signal processing methodology on signal decomposition. Specifically, we propose a new method called FLEXtime that uses a bank of bandpass filters to split the time series into frequency bands. Then, we learn the combination of these bands that optimally explains the model's prediction. Our extensive evaluation shows that, on average, FLEXtime outperforms state-of-the-art explainability methods across a range of datasets. FLEXtime fills an important gap in the current time series explainability methodology and is a valuable tool for a wide range of time series such as EEG and audio. Code will be made available at https://github.com/theabrusch/FLEXtime.
Authors: Pascal Janetzky, Tobias Schlagenhauf, Stefan Feuerriegel
Abstract: A common challenge in continual learning (CL) is catastrophic forgetting, where the performance on old tasks drops after new, additional tasks are learned. In this paper, we propose a novel framework called ReCL to slow down forgetting in CL. Our framework exploits an implicit bias of gradient-based neural networks due to which these converge to margin maximization points. Such convergence points allow us to reconstruct old data from previous tasks, which we then combine with the current training data. Our framework is flexible and can be applied on top of existing, state-of-the-art CL methods. We further demonstrate the performance gain from our framework across a large series of experiments, including two challenging CL scenarios (class incremental and domain incremental learning), different datasets (MNIST, CIFAR10, TinyImagenet), and different network architectures. Across all experiments, we find large performance gains through ReCL. To the best of our knowledge, our framework is the first to address catastrophic forgetting by leveraging models in CL as their own memory buffers.
Authors: Xiang Li, Jianpeng Qi, Zhongying Zhao, Guanjie Zheng, Lei Cao, Junyu Dong, Yanwei Yu
Abstract: Graph anomaly detection (GAD) is a critical task in graph machine learning, with the primary objective of identifying anomalous nodes that deviate significantly from the majority. This task is widely applied in various real-world scenarios, including fraud detection and social network analysis. However, existing GAD methods still face two major challenges: (1) They are often limited to detecting anomalies in single-type interaction graphs and struggle with multiple interaction types in multiplex heterogeneous graphs. (2) In unsupervised scenarios, selecting appropriate anomaly score thresholds remains a significant challenge for accurate anomaly detection. To address the above challenges, we propose a novel Unsupervised Multiplex Graph Anomaly Detection method, named UMGAD. We first learn multi-relational correlations among nodes in multiplex heterogeneous graphs and capture anomaly information during node attribute and structure reconstruction through graph-masked autoencoder (GMAE). Then, to further extract abnormal information, we generate attribute-level and subgraph-level augmented-view graphs respectively, and perform attribute and structure reconstruction through GMAE. Finally, we learn to optimize node attributes and structural features through contrastive learning between original-view and augmented-view graphs to improve the model's ability to capture anomalies. Meanwhile, we also propose a new anomaly score threshold selection strategy, which allows the model to be independent of ground truth information in real unsupervised scenarios. Extensive experiments on four datasets show that our UMGAD significantly outperforms state-of-the-art methods, achieving average improvements of 13.48% in AUC and 11.68% in Macro-F1 across all datasets.
Authors: Guangkun Nie, Gongzheng Tang, Shenda Hong
Abstract: Imbalanced data distributions are prevalent in real-world scenarios, posing significant challenges in both imbalanced classification and imbalanced regression tasks. They often cause deep learning models to overfit in areas of high sample density (many-shot regions) while underperforming in areas of low sample density (few-shot regions). This characteristic restricts the utility of deep learning models in various sectors, notably healthcare, where areas with few-shot data hold greater clinical relevance. While recent studies have shown the benefits of incorporating distribution information in imbalanced classification tasks, such strategies are rarely explored in imbalanced regression. In this paper, we address this issue by introducing a novel loss function, termed Dist Loss, designed to minimize the distribution distance between the model's predictions and the target labels in a differentiable manner, effectively integrating distribution information into model training. Dist Loss enables deep learning models to regularize their output distribution during training, effectively enhancing their focus on few-shot regions. We have conducted extensive experiments across three datasets spanning computer vision and healthcare: IMDB-WIKI-DIR, AgeDB-DIR, and ECG-Ka-DIR. The results demonstrate that Dist Loss effectively mitigates the negative impact of imbalanced data distribution on model performance, achieving state-of-the-art results in sparse data regions. Furthermore, Dist Loss is easy to integrate, complementing existing methods.
Authors: Nawaf Alampara, Mara Schilling-Wilhelmi, Marti\~no R\'ios-Garc\'ia, Indrajeet Mandal, Pranav Khetarpal, Hargun Singh Grover, N. M. Anoop Krishnan, Kevin Maik Jablonka
Abstract: Recent advancements in artificial intelligence have sparked interest in scientific assistants that could support researchers across the full spectrum of scientific workflows, from literature review to experimental design and data analysis. A key capability for such systems is the ability to process and reason about scientific information in both visual and textual forms - from interpreting spectroscopic data to understanding laboratory setups. Here, we introduce MaCBench, a comprehensive benchmark for evaluating how vision-language models handle real-world chemistry and materials science tasks across three core aspects: data extraction, experimental understanding, and results interpretation. Through a systematic evaluation of leading models, we find that while these systems show promising capabilities in basic perception tasks - achieving near-perfect performance in equipment identification and standardized data extraction - they exhibit fundamental limitations in spatial reasoning, cross-modal information synthesis, and multi-step logical inference. Our insights have important implications beyond chemistry and materials science, suggesting that developing reliable multimodal AI scientific assistants may require advances in curating suitable training data and approaches to training those models.
Authors: Roozbeh Yousefzadeh, Xuenan Cao, Azim Ospanov
Abstract: Using AI to write formal proofs for mathematical problems is a challenging task that has seen some advancements in recent years. Automated systems such as Lean can verify the correctness of proofs written in formal language, yet writing the proofs in formal language can be challenging for humans and machines. The miniF2F benchmark has 20 IMO problems in its test set, yet formal proofs are available only for 6 of these problems (3 of which are only written by mathematicians). The model with best accuracy can only prove 2 of these 20 IMO problems, from 1950s and 60s, while its training set is a secret. In this work, we write complete, original formal proofs for the remaining IMO problems in Lean along with 3 extra problems from IMO 2022 and 2023. This effort expands the availability of proof currently in the public domain by creating 5,880 lines of Lean proof. The goal of the paper is to pave the way for developing AI models that can automatically write the formal proofs for all the IMO problems in miniF2F and beyond by providing an evaluation benchmark. In this pursuit, we devise a method to decompose the proofs of these problems into their building blocks, constructing a dataset of 1,329 lemmas with more than 40k lines of Lean code. These lemmas are not trivial, yet they are approachable, providing the opportunity to evaluate and diagnose the failures and successes of AI models. We evaluate the ability of the SOTA LLMs on our dataset and analyze their success and failure modes from different perspectives. Our dataset and code is available at: https://github.com/roozbeh-yz/IMO-Steps.
Authors: Mahalakshmi Sabanayagam, Lukas Gosch, Stephan G\"unnemann, Debarghya Ghoshdastidar
Abstract: Machine learning models are highly vulnerable to label flipping, i.e., the adversarial modification (poisoning) of training labels to compromise performance. Thus, deriving robustness certificates is important to guarantee that test predictions remain unaffected and to understand worst-case robustness behavior. However, for Graph Neural Networks (GNNs), the problem of certifying label flipping has so far been unsolved. We change this by introducing an exact certification method, deriving both sample-wise and collective certificates. Our method leverages the Neural Tangent Kernel (NTK) to capture the training dynamics of wide networks enabling us to reformulate the bilevel optimization problem representing label flipping into a Mixed-Integer Linear Program (MILP). We apply our method to certify a broad range of GNN architectures in node classification tasks. Thereby, concerning the worst-case robustness to label flipping: $(i)$ we establish hierarchies of GNNs on different benchmark graphs; $(ii)$ quantify the effect of architectural choices such as activations, depth and skip-connections; and surprisingly, $(iii)$ uncover a novel phenomenon of the robustness plateauing for intermediate perturbation budgets across all investigated datasets and architectures. While we focus on GNNs, our certificates are applicable to sufficiently wide NNs in general through their NTK. Thus, our work presents the first exact certificate to a poisoning attack ever derived for neural networks, which could be of independent interest. The code is available at https://github.com/saper0/qpcert.
Authors: Dongjae Jeon, Dueun Kim, Albert No
Abstract: In this paper, we introduce a geometric framework to analyze memorization in diffusion models through the sharpness of the log probability density. We mathematically justify a previously proposed score-difference-based memorization metric by demonstrating its effectiveness in quantifying sharpness. Additionally, we propose a novel memorization metric that captures sharpness at the initial stage of image generation in latent diffusion models, offering early insights into potential memorization. Leveraging this metric, we develop a mitigation strategy that optimizes the initial noise of the generation process using a sharpness-aware regularization term.
Authors: Namgyu Kang, Jaemin Oh, Youngjoon Hong, Eunbyung Park
Abstract: The numerical approximation of partial differential equations (PDEs) using neural networks has seen significant advancements through Physics-Informed Neural Networks (PINNs). Despite their straightforward optimization framework and flexibility in implementing various PDEs, PINNs often suffer from limited accuracy due to the spectral bias of Multi-Layer Perceptrons (MLPs), which struggle to effectively learn high-frequency and nonlinear components. Recently, parametric mesh representations in combination with neural networks have been investigated as a promising approach to eliminate the inductive bias of MLPs. However, they usually require high-resolution grids and a large number of collocation points to achieve high accuracy while avoiding overfitting. In addition, the fixed positions of the mesh parameters restrict their flexibility, making accurate approximation of complex PDEs challenging. To overcome these limitations, we propose Physics-Informed Gaussians (PIGs), which combine feature embeddings using Gaussian functions with a lightweight neural network. Our approach uses trainable parameters for the mean and variance of each Gaussian, allowing for dynamic adjustment of their positions and shapes during training. This adaptability enables our model to optimally approximate PDE solutions, unlike models with fixed parameter positions. Furthermore, the proposed approach maintains the same optimization framework used in PINNs, allowing us to benefit from their excellent properties. Experimental results show the competitive performance of our model across various PDEs, demonstrating its potential as a robust tool for solving complex PDEs. Our project page is available at https://namgyukang.github.io/Physics-Informed-Gaussians/
URLs: https://namgyukang.github.io/Physics-Informed-Gaussians/
Authors: Yekun Ke, Yingyu Liang, Zhenmei Shi, Zhao Song, Chiwun Yang
Abstract: The application of transformer-based models on time series forecasting (TSF) tasks has long been popular to study. However, many of these works fail to beat the simple linear residual model, and the theoretical understanding of this issue is still limited. In this work, we propose the first theoretical explanation of the inefficiency of transformers on TSF tasks. We attribute the mechanism behind it to {\bf Asymmetric Learning} in training attention networks. When the sign of the previous step is inconsistent with the sign of the current step in the next-step-prediction time series, attention fails to learn the residual features. This makes it difficult to generalize on out-of-distribution (OOD) data, especially on the sign-inconsistent next-step-prediction data, with the same representation pattern, whereas a linear residual network could easily accomplish it. We hope our theoretical insights provide important necessary conditions for designing the expressive and efficient transformer-based architecture for practitioners.
Authors: Yao Fu, Yinsicheng Jiang, Yeqi Huang, Ping Nie, Zhan Lu, Leyang Xue, Congjie He, Man-Kit Sit, Jilong Xue, Li Dong, Ziming Miao, Kai Zou, Edoardo Ponti, Luo Mai
Abstract: The Mixture-of-Experts (MoE) architecture is increasingly favored for scaling Large Language Models (LLMs). Its key feature, sparse activation, selectively activates only a subset of parameters (experts) per token, reducing memory bandwidth and compute FLOPs compared to dense models. To capitalize on this, MoE designers leverage heterogeneous compute and memory hardware to lower system costs. However, the interaction between model sparsity and hardware heterogeneity introduces trade-offs in Cost, Accuracy, and Performance (CAP). To address this, we introduce MoE-CAP, a benchmarking method for evaluating sparse MoE systems across these three dimensions. Its key innovation is a sparsity-aware CAP analysis model, the first to integrate cost, performance, and accuracy metrics into a single diagram while estimating the impact of sparsity on system performance. MoE-CAP helps practitioners optimize hardware provisioning for an MoE model-or vice versa. MoE-CAP supports various MoE models and provides more accurate metrics than existing methods.
Authors: Billy Joe Franks, Moshe Eliasof, Semih Cant\"urk, Guy Wolf, Carola-Bibiane Sch\"onlieb, Sophie Fellenz, Marius Kloft
Abstract: Recent advances in integrating positional and structural encodings (PSEs) into graph neural networks (GNNs) have significantly enhanced their performance across various graph learning tasks. However, the general applicability of these encodings and their potential to serve as foundational representations for graphs remain uncertain. This paper investigates the fine-tuning efficiency, scalability with sample size, and generalization capability of learnable PSEs across diverse graph datasets. Specifically, we evaluate their potential as universal pre-trained models that can be easily adapted to new tasks with minimal fine-tuning and limited data. Furthermore, we assess the expressivity of the learned representations, particularly, when used to augment downstream GNNs. We demonstrate through extensive benchmarking and empirical analysis that PSEs generally enhance downstream models. However, some datasets may require specific PSE-augmentations to achieve optimal performance. Nevertheless, our findings highlight their significant potential to become integral components of future graph foundation models. We provide new insights into the strengths and limitations of PSEs, contributing to the broader discourse on foundation models in graph learning.
Authors: Lingzhi Shen, Yunfei Long, Xiaohao Cai, Imran Razzak, Guanming Chen, Kang Liu, Shoaib Jameel
Abstract: Multimodal fake news detection often involves modelling heterogeneous data sources, such as vision and language. Existing detection methods typically rely on fusion effectiveness and cross-modal consistency to model the content, complicating understanding how each modality affects prediction accuracy. Additionally, these methods are primarily based on static feature modelling, making it difficult to adapt to the dynamic changes and relationships between different data modalities. This paper develops a significantly novel approach, GAMED, for multimodal modelling, which focuses on generating distinctive and discriminative features through modal decoupling to enhance cross-modal synergies, thereby optimizing overall performance in the detection process. GAMED leverages multiple parallel expert networks to refine features and pre-embed semantic knowledge to improve the experts' ability in information selection and viewpoint sharing. Subsequently, the feature distribution of each modality is adaptively adjusted based on the respective experts' opinions. GAMED also introduces a novel classification technique to dynamically manage contributions from different modalities, while improving the explainability of decisions. Experimental results on the Fakeddit and Yang datasets demonstrate that GAMED performs better than recently developed state-of-the-art models. The source code can be accessed at https://github.com/slz0925/GAMED.
Authors: Austin Cheng, Alston Lo, Kin Long Kelvin Lee, Santiago Miret, Al\'an Aspuru-Guzik
Abstract: Molecular structure elucidation is a fundamental step in understanding chemical phenomena, with applications in identifying molecules in natural products, lab syntheses, forensic samples, and the interstellar medium. We consider the task of predicting a molecule's all-atom 3D structure given only its molecular formula and moments of inertia, motivated by the ability of rotational spectroscopy to measure these moments. While existing generative models can conditionally sample 3D structures with approximately correct moments, this soft conditioning fails to leverage the many digits of precision afforded by experimental rotational spectroscopy. To address this, we first show that the space of $n$-atom point clouds with a fixed set of moments of inertia is embedded in the Stiefel manifold $\mathrm{St}(n, 4)$. We then propose Stiefel Flow Matching as a generative model for elucidating 3D structure under exact moment constraints. Additionally, we learn simpler and shorter flows by finding approximate solutions for equivariant optimal transport on the Stiefel manifold. Empirically, enforcing exact moment constraints allows Stiefel Flow Matching to achieve higher success rates and faster sampling than Euclidean diffusion models, even on high-dimensional manifolds corresponding to large molecules in the GEOM dataset.
Authors: Zheng Chen, Yasuko Matsubara, Yasushi Sakurai, Jimeng Sun
Abstract: Deep learning models have recently shown great success in classifying epileptic patients using EEG recordings. Unfortunately, classification-based methods lack a sound mechanism to detect the onset of seizure events. In this work, we propose a two-stage framework, SODor, that explicitly models seizure onset through a novel task formulation of subsequence clustering. Given an EEG sequence, the framework first learns a set of second-level embeddings with label supervision. It then employs model-based clustering to explicitly capture long-term temporal dependencies in EEG sequences and identify meaningful subsequences. Epochs within a subsequence share a common cluster assignment (normal or seizure), with cluster or state transitions representing successful onset detections. Extensive experiments on three datasets demonstrate that our method can correct misclassifications, achieving 5\%-11\% classification improvements over other baselines and accurately detecting seizure onsets.
Authors: Anish Dhir, Matthew Ashman, James Requeima, Mark van der Wilk
Abstract: Discovering a unique causal structure is difficult due to both inherent identifiability issues, and the consequences of finite data. As such, uncertainty over causal structures, such as those obtained from a Bayesian posterior, are often necessary for downstream tasks. Finding an accurate approximation to this posterior is challenging, due to the large number of possible causal graphs, as well as the difficulty in the subproblem of finding posteriors over the functional relationships of the causal edges. Recent works have used meta-learning to view the problem of estimating the maximum a-posteriori causal graph as supervised learning. Yet, these methods are limited when estimating the full posterior as they fail to encode key properties of the posterior, such as correlation between edges and permutation equivariance with respect to nodes. Further, these methods also cannot reliably sample from the posterior over causal structures. To address these limitations, we propose a Bayesian meta learning model that allows for sampling causal structures from the posterior and encodes these key properties. We compare our meta-Bayesian causal discovery against existing Bayesian causal discovery methods, demonstrating the advantages of directly learning a posterior over causal structure.
Authors: Marta Skreta, Lazar Atanackovic, Avishek Joey Bose, Alexander Tong, Kirill Neklyudov
Abstract: The Cambrian explosion of easily accessible pre-trained diffusion models suggests a demand for methods that combine multiple different pre-trained diffusion models without incurring the significant computational burden of re-training a larger combined model. In this paper, we cast the problem of combining multiple pre-trained diffusion models at the generation stage under a novel proposed framework termed superposition. Theoretically, we derive superposition from rigorous first principles stemming from the celebrated continuity equation and design two novel algorithms tailor-made for combining diffusion models in SuperDiff. SuperDiff leverages a new scalable It\^o density estimator for the log likelihood of the diffusion SDE which incurs no additional overhead compared to the well-known Hutchinson's estimator needed for divergence calculations. We demonstrate that SuperDiff is scalable to large pre-trained diffusion models as superposition is performed solely through composition during inference, and also enjoys painless implementation as it combines different pre-trained vector fields through an automated re-weighting scheme. Notably, we show that SuperDiff is efficient during inference time, and mimics traditional composition operators such as the logical OR and the logical AND. We empirically demonstrate the utility of using SuperDiff for generating more diverse images on CIFAR-10, more faithful prompt conditioned image editing using Stable Diffusion, as well as improved conditional molecule generation and unconditional de novo structure design of proteins. https://github.com/necludov/super-diffusion
Authors: Ioannis Bilionis, Ricardo C. Berrios, Luis Fernandez-Luque, Carlos Castillo
Abstract: Machine Learning (ML) algorithms are vital for supporting clinical decision-making in biomedical informatics. However, their predictive performance can vary across demographic groups, often due to the underrepresentation of historically marginalized populations in training datasets. The investigation reveals widespread sex- and age-related inequities in chronic disease datasets and their derived ML models. Thus, a novel analytical framework is introduced, combining systematic arbitrariness with traditional metrics like accuracy and data complexity. The analysis of data from over 25,000 individuals with chronic diseases revealed mild sex-related disparities, favoring predictive accuracy for males, and significant age-related differences, with better accuracy for younger patients. Notably, older patients showed inconsistent predictive accuracy across seven datasets, linked to higher data complexity and lower model performance. This highlights that representativeness in training data alone does not guarantee equitable outcomes, and model arbitrariness must be addressed before deploying models in clinical settings.
Authors: S. David Mis, Matti Lassas, Maarten V. de Hoop
Abstract: Many numerical algorithms in scientific computing -- particularly in areas like numerical linear algebra, PDE simulation, and inverse problems -- produce outputs that can be represented by semialgebraic functions; that is, the graph of the computed function can be described by finitely many polynomial equalities and inequalities. In this work, we introduce Semialgebraic Neural Networks (SANNs), a neural network architecture capable of representing any bounded semialgebraic function, and computing such functions up to the accuracy of a numerical ODE solver chosen by the programmer. Conceptually, we encode the graph of the learned function as the kernel of a piecewise polynomial selected from a class of functions whose roots can be evaluated using a particular homotopy continuation method. We show by construction that the SANN architecture is able to execute this continuation method, thus evaluating the learned semialgebraic function. Furthermore, the architecture can exactly represent even discontinuous semialgebraic functions by executing a continuation method on each connected component of the target function. Lastly, we provide example applications of these networks and show they can be trained with traditional deep-learning techniques.
Authors: Zheqi Lv, Keming Ye, Zishu Wei, Qi Tian, Shengyu Zhang, Wenqiao Zhang, Wenjie Wang, Kun Kuang, Tat-Seng Chua, Fei Wu
Abstract: Deep neural networks have become foundational to advancements in multiple domains, including recommendation systems, natural language processing, and so on. Despite their successes, these models often contain incompatible parameters that can be underutilized or detrimental to model performance, particularly when faced with specific, varying data distributions. Existing research excels in removing such parameters or merging the outputs of multiple different pretrained models. However, the former focuses on efficiency rather than performance, while the latter requires several times more computing and storage resources to support inference. In this paper, we set the goal to explicitly improve these incompatible parameters by leveraging the complementary strengths of different models, thereby directly enhancing the models without any additional parameters. Specifically, we propose Compatibility-aware Knowledge Integration (CKI), which consists of Parameter Compatibility Assessment and Parameter Splicing, which are used to evaluate the knowledge content of multiple models and integrate the knowledge into one model, respectively. The integrated model can be used directly for inference or for further fine-tuning. We conduct extensive experiments on various datasets for recommendation and language tasks, and the results show that Compatibility-aware Knowledge Integration can effectively optimize incompatible parameters under multiple tasks and settings to break through the training limit of the original model without increasing the inference cost.
Authors: David Millard, Arielle Carr, St\'ephane Gaudreault, Ali Baheri
Abstract: We present PEARL (Preconditioner Enhancement through Actor-critic Reinforcement Learning), a novel approach to learning matrix preconditioners. Existing preconditioners such as Jacobi, Incomplete LU, and Algebraic Multigrid methods offer problem-specific advantages but rely heavily on hyperparameter tuning. Recent advances have explored using deep neural networks to learn preconditioners, though challenges such as misbehaved objective functions and costly training procedures remain. PEARL introduces a reinforcement learning approach for learning preconditioners, specifically, a contextual bandit formulation. The framework utilizes an actor-critic model, where the actor generates the incomplete Cholesky decomposition of preconditioners, and the critic evaluates them based on reward-specific feedback. To further guide the training, we design a dual-objective function, combining updates from the critic and condition number. PEARL contributes a generalizable preconditioner learning method, dynamic sparsity exploration, and cosine schedulers for improved stability and exploratory power. We compare our approach to traditional and neural preconditioners, demonstrating improved flexibility and iterative solving speed.
Authors: Weiyu Chen, Xiaoyuan Zhang, Baijiong Lin, Xi Lin, Han Zhao, Qingfu Zhang, James T. Kwok
Abstract: Multi-objective optimization (MOO) in deep learning aims to simultaneously optimize multiple conflicting objectives, a challenge frequently encountered in areas like multi-task learning and multi-criteria learning. Recent advancements in gradient-based MOO methods have enabled the discovery of diverse types of solutions, ranging from a single balanced solution to finite or even infinite Pareto sets, tailored to user needs. These developments have broad applications across domains such as reinforcement learning, computer vision, recommendation systems, and large language models. This survey provides the first comprehensive review of gradient-based MOO in deep learning, covering algorithms, theories, and practical applications. By unifying various approaches and identifying critical challenges, it serves as a foundational resource for driving innovation in this evolving field. A comprehensive list of MOO algorithms in deep learning is available at https://github.com/Baijiong-Lin/Awesome-Multi-Objective-Deep-Learning.
URLs: https://github.com/Baijiong-Lin/Awesome-Multi-Objective-Deep-Learning.
Authors: Nachiket Kapure, Harsh Joshi, Parul Kumari, Rajeshwari Mistri, Manasi Mali
Abstract: The challenges in feature selection, particularly in balancing model accuracy, interpretability, and computational efficiency, remain a critical issue in advancing machine learning methodologies. To address these complexities, this study introduces a novel hybrid approach, the Forward Recursive Adaptive Model Extraction Technique (FRAME), which combines Forward Selection and Recursive Feature Elimination (RFE) to enhance feature selection across diverse datasets. By combining the exploratory capabilities of Forward Selection with the refinement strengths of RFE, FRAME systematically identifies optimal feature subsets, striking a harmonious trade-off between experimentation and precision. A comprehensive evaluation of FRAME is conducted against traditional methods such as SelectKBest and Lasso Regression, using high-dimensional, noisy, and heterogeneous datasets. The results demonstrate that FRAME consistently delivers superior predictive performance based on downstream machine learning evaluation metrics. It efficiently performs dimensionality reduction with strong model performance, thus being especially useful for applications that need interpretable and accurate predictions, e.g., biomedical diagnostics. This research emphasizes the need to evaluate feature selection techniques on diverse datasets to test their robustness and generalizability. The results indicate that FRAME has great potential for further development, especially by incorporating deep learning frameworks for adaptive and real-time feature selection in dynamic settings. By advancing feature selection methodologies, FRAME offers a practical and effective solution to improve machine learning applications across multiple domains.
Authors: Haowei Lin, Shanda Li, Haotian Ye, Yiming Yang, Stefano Ermon, Yitao Liang, Jianzhu Ma
Abstract: Given an unconditional generative model and a predictor for a target property (e.g., a classifier), the goal of training-free guidance is to generate samples with desirable target properties without additional training. As a highly efficient technique for steering generative models toward flexible outcomes, training-free guidance has gained increasing attention in diffusion models. However, existing methods only handle data in continuous spaces, while many scientific applications involve both continuous and discrete data (referred to as multimodality). Another emerging trend is the growing use of the simple and general flow matching framework in building generative foundation models, where guided generation remains under-explored. To address this, we introduce TFG-Flow, a novel training-free guidance method for multimodal generative flow. TFG-Flow addresses the curse-of-dimensionality while maintaining the property of unbiased sampling in guiding discrete variables. We validate TFG-Flow on four molecular design tasks and show that TFG-Flow has great potential in drug design by generating molecules with desired properties.
Authors: Minh Le, Anh Nguyen, Huy Nguyen, Chau Nguyen, Nhat Ho
Abstract: Visual Prompt Tuning (VPT) has recently emerged as a powerful method for adapting pre-trained vision models to downstream tasks. By introducing learnable prompt tokens as task-specific instructions, VPT effectively guides pre-trained transformer models with minimal overhead. Despite its empirical success, a comprehensive theoretical understanding of VPT remains an active area of research. Building on recent insights into the connection between mixture of experts and prompt-based approaches, we identify a key limitation in VPT: the restricted functional expressiveness in prompt formulation. To address this limitation, we propose Visual Adaptive Prompt Tuning (VAPT), a new generation of prompts that redefines prompts as adaptive functions of the input. Our theoretical analysis shows that this simple yet intuitive approach achieves optimal sample efficiency. Empirical results on VTAB-1K and FGVC further demonstrate VAPT's effectiveness, with performance gains of 7.34% and 1.04% over fully fine-tuning baselines, respectively. Notably, VAPT also surpasses VPT by a substantial margin while using fewer parameters. These results highlight both the effectiveness and efficiency of our method and pave the way for future research to explore the potential of adaptive prompts.
Authors: Yunchuan Guan, Yu Liu, Ke Zhou, Zhiqi Shen, Serge Belongie, Jenq-Neng Hwang, Lei Li
Abstract: Diffusion-based algorithms have emerged as promising techniques for weight generation, particularly in scenarios like multi-task learning that require frequent weight updates. However, existing solutions suffer from limited cross-task transferability. In addition, they only utilize optimal weights as training samples, ignoring the value of other weights in the optimization process. To address these issues, we propose Lt-Di, which integrates the diffusion algorithm with meta-learning to generate weights for unseen tasks. Furthermore, we extend the vanilla diffusion algorithm into a trajectory diffusion algorithm to utilize other weights along the optimization trajectory. Trajectory diffusion decomposes the entire diffusion chain into multiple shorter ones, improving training and inference efficiency. We analyze the convergence properties of the weight generation paradigm and improve convergence efficiency without additional time overhead. Our experiments demonstrate Lt-Di's higher accuracy while reducing computational overhead across various tasks, including zero-shot and few-shot learning, multi-domain generalization, and large-scale language model fine-tuning.Our code is released at https://anonymous.4open.science/r/Lt-Di-0E51.
Authors: Wei Yao, Wenkai Yang, Ziqiao Wang, Yankai Lin, Yong Liu
Abstract: Weak-to-strong generalization, where weakly supervised strong models outperform their weaker teachers, offers a promising approach to aligning superhuman models with human values. To deepen the understanding of this approach, we provide theoretical insights into its capabilities and limitations. First, in the classification setting, we establish upper and lower generalization error bounds for the strong model, identifying the primary limitations as stemming from the weak model's generalization error and the optimization objective itself. Additionally, we derive lower and upper bounds on the calibration error of the strong model. These theoretical bounds reveal two critical insights: (1) the weak model should demonstrate strong generalization performance and maintain well-calibrated predictions, and (2) the strong model's training process must strike a careful balance, as excessive optimization could undermine its generalization capability by over-relying on the weak supervision signals. Finally, in the regression setting, we extend the work of Charikar et al. (2024) to a loss function based on Kullback-Leibler (KL) divergence, offering guarantees that the strong student can outperform its weak teacher by at least the magnitude of their disagreement. We conduct sufficient experiments to validate our theory.
Authors: Kim Yong Tan, Yueming Lyu, Ivor Tsang, Yew-Soon Ong
Abstract: Guided diffusion-model generation is a promising direction for customizing the generation process of a pre-trained diffusion-model to address the specific downstream tasks. Existing guided diffusion models either rely on training of the guidance model with pre-collected datasets or require the objective functions to be differentiable. However, for most real-world tasks, the offline datasets are often unavailable, and their objective functions are often not differentiable, such as image generation with human preferences, molecular generation for drug discovery, and material design. Thus, we need an $\textbf{online}$ algorithm capable of collecting data during runtime and supporting a $\textbf{black-box}$ objective function. Moreover, the $\textbf{query efficiency}$ of the algorithm is also critical because the objective evaluation of the query is often expensive in the real-world scenarios. In this work, we propose a novel and simple algorithm, $\textbf{Fast Direct}$, for query-efficient online black-box target generation. Our Fast Direct builds a pseudo-target on the data manifold to update the noise sequence of the diffusion model with a universal direction, which is promising to perform query-efficient guided generation. Extensive experiments on twelve high-resolution ($\small {1024 \times 1024}$) image target generation tasks and six 3D-molecule target generation tasks show $\textbf{6}\times$ up to $\textbf{10}\times$ query efficiency improvement and $\textbf{11}\times$ up to $\textbf{44}\times$ query efficiency improvement, respectively. Our implementation is publicly available at: https://github.com/kimyong95/guide-stable-diffusion/tree/fast-direct
URLs: https://github.com/kimyong95/guide-stable-diffusion/tree/fast-direct
Authors: Ryotaro Kawata, Kazusato Oko, Atsushi Nitanda, Taiji Suzuki
Abstract: We introduce a novel alignment method for diffusion models from distribution optimization perspectives while providing rigorous convergence guarantees. We first formulate the problem as a generic regularized loss minimization over probability distributions and directly optimize the distribution using the Dual Averaging method. Next, we enable sampling from the learned distribution by approximating its score function via Doob's $h$-transform technique. The proposed framework is supported by rigorous convergence guarantees and an end-to-end bound on the sampling error, which imply that when the original distribution's score is known accurately, the complexity of sampling from shifted distributions is independent of isoperimetric conditions. This framework is broadly applicable to general distribution optimization problems, including alignment tasks in Reinforcement Learning with Human Feedback (RLHF), Direct Preference Optimization (DPO), and Kahneman-Tversky Optimization (KTO). We empirically validate its performance on synthetic and image datasets using the DPO objective.
Authors: Shahaf Bassan, Ron Eliav, Shlomit Gur
Abstract: *Minimal sufficient reasons* represent a prevalent form of explanation - the smallest subset of input features which, when held constant at their corresponding values, ensure that the prediction remains unchanged. Previous *post-hoc* methods attempt to obtain such explanations but face two main limitations: (1) Obtaining these subsets poses a computational challenge, leading most scalable methods to converge towards suboptimal, less meaningful subsets; (2) These methods heavily rely on sampling out-of-distribution input assignments, potentially resulting in counterintuitive behaviors. To tackle these limitations, we propose in this work a self-supervised training approach, which we term *sufficient subset training* (SST). Using SST, we train models to generate concise sufficient reasons for their predictions as an integral part of their output. Our results indicate that our framework produces succinct and faithful subsets substantially more efficiently than competing post-hoc methods, while maintaining comparable predictive performance.
Authors: Saaketh Narayan, Abhay Gupta, Mansheej Paul, Davis Blalock
Abstract: Large Language Model training with 8-bit floating point (FP8) formats promises significant efficiency improvements, but reduced numerical precision makes training challenging. It is currently possible to train in FP8 only if one is willing to tune various hyperparameters, reduce model scale, or accept the overhead of computing dynamic scale factors. We demonstrate simple, scalable FP8 training that requires no dynamic scaling factors or special hyperparameters, even at large model sizes. Our method, $\mu$nit Scaling ($\mu$S), also enables simple hyperparameter transfer across model widths, matched numerics across training and inference, and other desirable properties. $\mu$nit Scaling is straightforward to implement, consisting of a set of minimal interventions based on a first-principles analysis of common transformer operations. We validate our method by training models from 1B to 13B parameters, performing all hidden linear layer computations in FP8. We achieve quality equal to higher precision baselines while also training up to 33% faster.
Authors: Pranav Nair, Puranjay Datta, Jeff Dean, Prateek Jain, Aditya Kusupati
Abstract: Quantizing model weights is critical for reducing the communication and inference costs of large models. However, quantizing models -- especially to low precisions like int4 or int2 -- requires a trade-off in model quality; int2, in particular, is known to severely degrade model quality. Consequently, practitioners are often forced to maintain multiple models with different quantization levels or serve a single model that best satisfies the quality-latency trade-off. On the other hand, integer data types, such as int8, inherently possess a nested (Matryoshka) structure where smaller bit-width integers, like int4 or int2, are nested within the most significant bits. Leveraging this insight, in this paper, we propose Matryoshka Quantization (MatQuant), a novel multi-scale quantization technique that alleviates the aforementioned challenge. This technique allows us to train and maintain a single quantized model but serve it with the precision demanded by the deployment. Furthermore, leveraging MatQuant's co-training and co-distillation regularization, int2 precision models extracted by MatQuant outperform standard int2 quantization by up to to 4% and 7% with OmniQuant and QAT as base algorithms respectively. Finally, we demonstrate that by using an extra bit to represent outliers, a model with an effective precision of 2.05-bit gives an additional 6% improvement with OmniQuant as the base algorithm.
Authors: Cale Coffman, Lizhong Chen
Abstract: Kolmogorov-Arnold Networks (KAN) are a new class of neural network architecture representing a promising alternative to the Multilayer Perceptron (MLP), demonstrating improved expressiveness and interpretability. However, KANs suffer from slow training and inference speeds relative to MLPs due in part to the recursive nature of the underlying B-spline calculations. This issue is particularly apparent with respect to KANs utilizing high-degree B-splines, as the number of required non-parallelizable recursions is proportional to B-spline degree. We solve this issue by proposing MatrixKAN, a novel optimization that parallelizes B-spline calculations with matrix representation and operations, thus significantly improving effective computation time for models utilizing high-degree B-splines. In this paper, we demonstrate the superior scaling of MatrixKAN's computation time relative to B-spline degree. Further, our experiments demonstrate speedups of approximately 40x relative to KAN, with significant additional speedup potential for larger datasets or higher spline degrees.
Authors: Mingtian Zhang, Jiajun He, Wenlin Chen, Zijing Ou, Jos\'e Miguel Hern\'andez-Lobato, Bernhard Sch\"olkopf, David Barber
Abstract: Recent advances in one-step generative models typically follow a two-stage process: first training a teacher diffusion model and then distilling it into a one-step student model. This distillation process traditionally relies on both the teacher model's score function to compute the distillation loss and its weights for student initialization. In this paper, we explore whether one-step generative models can be trained directly without this distillation process. First, we show that the teacher's score function is not essential and propose a family of distillation methods that achieve competitive results without relying on score estimation. Next, we demonstrate that initialization from teacher weights is indispensable in successful training. Surprisingly, we find that this benefit is not due to improved ``input-output" mapping but rather the learned feature representations, which dominate distillation quality. Our findings provide a better understanding of the role of initialization in one-step model training and its impact on distillation quality.
Authors: Bishwajit Prasad Gond, Durga Prasad Mohapatra
Abstract: Malware classification in dynamic environments presents a significant challenge due to concept drift, where the statistical properties of malware data evolve over time, complicating detection efforts. To address this issue, we propose a deep learning framework enhanced with a genetic algorithm to improve malware classification accuracy and adaptability. Our approach incorporates mutation operations and fitness score evaluations within genetic algorithms to continuously refine the deep learning model, ensuring robustness against evolving malware threats. Experimental results demonstrate that this hybrid method significantly enhances classification performance and adaptability, outperforming traditional static models. Our proposed approach offers a promising solution for real-time malware classification in ever-changing cybersecurity landscapes.
Authors: Saul Santos, Ant\'onio Farinhas, Daniel C. McNamee, Andr\'e F. T. Martins
Abstract: Recent research has established a connection between modern Hopfield networks (HNs) and transformer attention heads, with guarantees of exponential storage capacity. However, these models still face challenges scaling storage efficiently. Inspired by psychological theories of continuous neural resource allocation in working memory, we propose an approach that compresses large discrete Hopfield memories into smaller, continuous-time memories. Leveraging continuous attention, our new energy function modifies the update rule of HNs, replacing the traditional softmax-based probability mass function with a probability density, over the continuous memory. This formulation aligns with modern perspectives on human executive function, offering a principled link between attractor dynamics in working memory and resource-efficient memory allocation. Our framework maintains competitive performance with HNs while leveraging a compressed memory, reducing computational costs across synthetic and video datasets.
Authors: Shenglong Zhou, Ouya Wang, Ziyan Luo, Yongxu Zhu, Geoffrey Ye Li
Abstract: The recent advancement of foundation models (FMs) has brought about a paradigm shift, revolutionizing various sectors worldwide. The popular optimizers used to train these models are stochastic gradient descent-based algorithms, which face inherent limitations, such as slow convergence and stringent assumptions for convergence. In particular, data heterogeneity arising from distributed settings poses significant challenges to their theoretical and numerical performance. This paper develops an algorithm, PISA ({P}reconditioned {I}nexact {S}tochastic {A}lternating Direction Method of Multipliers), which enables scalable parallel computing and supports various second-moment schemes. Grounded in rigorous theoretical guarantees, the algorithm converges under the sole assumption of Lipschitz continuity of the gradient, thereby removing the need for other conditions commonly imposed by stochastic methods. This capability enables PISA to tackle the challenge of data heterogeneity effectively. Comprehensive experimental evaluations for training or fine-tuning diverse FMs, including vision models, large language models, reinforcement learning models, generative adversarial networks, and recurrent neural networks, demonstrate its superior numerical performance compared to various state-of-the-art optimizers.
Authors: Bohan Lyu, Siqiao Huang, Zichen Liang
Abstract: Neural surrogate models have emerged as powerful and efficient tools in data mining. Meanwhile, large language models (LLMs) have demonstrated remarkable capabilities in code-related tasks. We investigate a novel application: using LLMs as surrogate models for code execution prediction. Given LLMs' unique ability to understand and process diverse programs, they present a promising direction for building general-purpose surrogate models. To systematically investigate this capability, we introduce SURGE, a comprehensive benchmark with $1160$ problems covering $8$ key aspects: multi-language programming tasks, competition-level programming problems, repository-level code analysis, high-cost scientific computing, time-complexity-intensive algorithms, buggy code analysis, programs dependent on specific compilers or execution environments, and formal mathematical proof verification. Through extensive empirical analysis of $21$ open-source and proprietary LLMs, we examine scaling laws, data efficiency, and predictive accuracy. Our findings reveal important insights about the feasibility of LLMs as efficient surrogates for computational processes, with implications for automated software testing, program analysis, and computational resource optimization in data mining applications. Code and dataset are released at https://github.com/Imbernoulli/SURGE.
Authors: Zhiyuan Zeng, Qinyuan Cheng, Zhangyue Yin, Yunhua Zhou, Xipeng Qiu
Abstract: The advent of test-time scaling in large language models (LLMs), exemplified by OpenAI's o1 series, has advanced reasoning capabilities by scaling computational resource allocation during inference. While successors like QwQ, Deepseek-R1 (R1) and LIMO replicate these advancements, whether these models truly possess test-time scaling capabilities remains underexplored. This study found that longer CoTs of these o1-like models do not consistently enhance accuracy; in fact, correct solutions are often shorter than incorrect ones for the same questions. Further investigation shows this phenomenon is closely related to models' self-revision capabilities - longer CoTs contain more self-revisions, which often lead to performance degradation. We then compare sequential and parallel scaling strategies on QwQ, R1 and LIMO, finding that parallel scaling achieves better coverage and scalability. Based on these insights, we propose Shortest Majority Vote, a method that combines parallel scaling strategies with CoT length characteristics, significantly improving models' test-time scalability compared to conventional majority voting approaches.
Authors: Jacob Fein-Ashley
Abstract: Diffusion kernels capture global dependencies. We present Linear Diffusion Networks (LDNs), a novel architecture that reinterprets sequential data processing as a unified diffusion process. Our model integrates adaptive diffusion modules with localized nonlinear updates and a diffusion-inspired attention mechanism. This design enables efficient global information propagation while preserving fine-grained temporal details. LDN overcomes the limitations of conventional recurrent and transformer models by allowing full parallelization across time steps and supporting robust multi-scale temporal representations. Experiments on benchmark sequence modeling tasks demonstrate that LDN delivers competitive performance across ImageNet and GLUE tasks.
Authors: Behraj Khan, Behroz Mirza, Nouman Durrani, Tahir Syed
Abstract: The continuous surge in data volume and velocity is often dealt with using data orchestration and distributed processing approaches, abstracting away the machine learning challenges that exist at the algorithmic level. With growing interest in automating the learning loop, training with data that arrive in a sequence rather than in the classical in-memory training data form will face a machine learning challenge because of evolving feature distributions across batches of training data biasing the cross-validation step (\cite{sugiyama2012machine}). This work takes a distributed density estimation angle to the problem where data are temporally distributed. It processes data in batches and allows a neural network to treat a batch as training data. The method accumulates knowledge about the data density via posterior probability absorption using the Fisher Information Matrix, which contains information about the local optimization gradients for the batch. This is then used as a regularizer for the loss in the following batch, and therefore the density estimate for the entire dataset constructively gets more robust to the non-iid distribution shift. This needs the presence of a pair of batches in memory at a time, so the space cost is not a function of the size of the complete, distributed dataset. We proposed a novel regularization-based approach Covariate Shift Correction $C^{2}A$ that leverages Fisher information and Kullback-Leibler divergence to adapt to both natural and sequential covariate shift caused by dataset fragmentation. $C^{2}A$ achieves $19\%$ accuracy at maximum against state-of-the-art methods.
Authors: Sangwoong Yoon, Himchan Hwang, Hyeokju Jeong, Dong Kyu Shin, Che-Sang Park, Sehee Kweon, Frank Chongwoo Park
Abstract: We propose the Value Gradient Sampler (VGS), a trainable sampler based on the interpretation of sampling as discrete-time sequential decision-making. VGS generates samples from a given unnormalized density (i.e., energy) by drifting and diffusing randomly initialized particles. In VGS, finding the optimal drift is equivalent to solving an optimal control problem where the cost is the upper bound of the KL divergence between the target density and the samples. We employ value-based dynamic programming to solve this optimal control problem, which gives the gradient of the value function as the optimal drift vector. The connection to sequential decision making allows VGS to leverage extensively studied techniques in reinforcement learning, making VGS a fast, adaptive, and accurate sampler that achieves competitive results in various sampling benchmarks. Furthermore, VGS can replace MCMC in contrastive divergence training of energy-based models. We demonstrate the effectiveness of VGS in training accurate energy-based models in industrial anomaly detection applications.
Authors: Puning Zhao, Chuan Ma, Li Shen, Shaowei Wang, Rongfei Fan
Abstract: Label differential privacy (DP) is designed for learning problems involving private labels and public features. While various methods have been proposed for learning under label DP, the theoretical limits remain largely unexplored. In this paper, we investigate the fundamental limits of learning with label DP in both local and central models for both classification and regression tasks, characterized by minimax convergence rates. We establish lower bounds by converting each task into a multiple hypothesis testing problem and bounding the test error. Additionally, we develop algorithms that yield matching upper bounds. Our results demonstrate that under label local DP (LDP), the risk has a significantly faster convergence rate than that under full LDP, i.e. protecting both features and labels, indicating the advantages of relaxing the DP definition to focus solely on labels. In contrast, under the label central DP (CDP), the risk is only reduced by a constant factor compared to full DP, indicating that the relaxation of CDP only has limited benefits on the performance.
Authors: Stefan K. Nielsen, Rachel S. Y. Teo, Laziz U. Abdullaev, Tan M. Nguyen
Abstract: Sparse Mixture-of-Experts (MoE) architectures have emerged as a promising approach to decoupling model capacity from computational cost. At the core of the MoE model is the router, which learns the underlying clustering structure of the input distribution in order to send input tokens to appropriate experts. However, latent clusters may be unidentifiable in high dimension, which causes slow convergence, susceptibility to data contamination, and overall degraded representations as the router is unable to perform appropriate token-expert matching. We examine the router through the lens of clustering optimization and derive optimal feature weights that maximally identify the latent clusters. We use these weights to compute the token-expert routing assignments in an adaptively transformed space that promotes well-separated clusters, which helps identify the best-matched expert for each token. In particular, for each expert cluster, we compute a set of weights that scales features according to whether that expert clusters tightly along that feature. We term this novel router the Adaptive Clustering (AC) router. Our AC router enables the MoE model to obtain three connected benefits: 1) faster convergence, 2) better robustness to data corruption, and 3) overall performance improvement, as experts are specialized in semantically distinct regions of the input space. We empirically demonstrate the advantages of our AC router over baseline routing methods when applied on a variety of MoE backbones for language modeling and image recognition tasks in both clean and corrupted settings.
Authors: Wenyue Hua, Tyler Wong, Sun Fei, Liangming Pan, Adam Jardine, William Yang Wang
Abstract: Large language models (LLMs) have shown remarkable improvements in reasoning and many existing benchmarks have been addressed by models such as o1 and o3 either fully or partially. However, a majority of these benchmarks emphasize deductive reasoning, including mathematical and coding tasks in which rules such as mathematical axioms or programming syntax are clearly defined, based on which LLMs can plan and apply these rules to arrive at a solution. In contrast, inductive reasoning, where one infers the underlying rules from observed data, remains less explored. Such inductive processes lie at the heart of scientific discovery, as they enable researchers to extract general principles from empirical observations. To assess whether LLMs possess this capacity, we introduce InductionBench, a new benchmark designed to evaluate the inductive reasoning ability of LLMs. Our experimental findings reveal that even the most advanced models available struggle to master the simplest complexity classes within the subregular hierarchy of functions, highlighting a notable deficiency in current LLMs' inductive reasoning capabilities. Coda and data are available https://github.com/Wenyueh/inductive_reasoning_benchmark.
URLs: https://github.com/Wenyueh/inductive_reasoning_benchmark.
Authors: Zhipeng Cheng, Xiaoyu Xia, Hong Wang, Minghui Liwang, Ning Chen, Xuwei Fan, Xianbin Wang
Abstract: Edge inference (EI) is a key solution to address the growing challenges of delayed response times, limited scalability, and privacy concerns in cloud-based Deep Neural Network (DNN) inference. However, deploying DNN models on resource-constrained edge devices faces more severe challenges, such as model storage limitations, dynamic service requests, and privacy risks. This paper proposes a novel framework for privacy-aware joint DNN model deployment and partition optimization to minimize long-term average inference delay under resource and privacy constraints. Specifically, the problem is formulated as a complex optimization problem considering model deployment, user-server association, and model partition strategies. To handle the NP-hardness and future uncertainties, a Lyapunov-based approach is introduced to transform the long-term optimization into a single-time-slot problem, ensuring system performance. Additionally, a coalition formation game model is proposed for edge server association, and a greedy-based algorithm is developed for model deployment within each coalition to efficiently solve the problem. Extensive simulations show that the proposed algorithms effectively reduce inference delay while satisfying privacy constraints, outperforming baseline approaches in various scenarios.
Authors: Guoqi Yu, Yaoming Li, Juncheng Wang, Xiaoyu Guo, Angelica I. Aviles-Rivero, Tong Yang, Shujun Wang
Abstract: Recent advancements have progressively incorporated frequency-based techniques into deep learning models, leading to notable improvements in accuracy and efficiency for time series analysis tasks. However, the Mid-Frequency Spectrum Gap in the real-world time series, where the energy is concentrated at the low-frequency region while the middle-frequency band is negligible, hinders the ability of existing deep learning models to extract the crucial frequency information. Additionally, the shared Key-Frequency in multivariate time series, where different time series share indistinguishable frequency patterns, is rarely exploited by existing literature. This work introduces a novel module, Adaptive Mid-Frequency Energy Optimizer, based on convolution and residual learning, to emphasize the significance of mid-frequency bands. We also propose an Energy-based Key-Frequency Picking Block to capture shared Key-Frequency, which achieves superior inter-series modeling performance with fewer parameters. A novel Key-Frequency Enhanced Training strategy is employed to further enhance Key-Frequency modeling, where spectral information from other channels is randomly introduced into each channel. Our approach advanced multivariate time series forecasting on the challenging Traffic, ECL, and Solar benchmarks, reducing MSE by 4%, 6%, and 5% compared to the previous SOTA iTransformer. Code is available at this GitHub Repository: https://github.com/Levi-Ackman/ReFocus.
Authors: Yichi Zhang, Yici Yan, Alex Schwing, Zhizhen Zhao
Abstract: We formulate a hierarchical rectified flow to model data distributions. It hierarchically couples multiple ordinary differential equations (ODEs) and defines a time-differentiable stochastic process that generates a data distribution from a known source distribution. Each ODE resembles the ODE that is solved in a classic rectified flow, but differs in its domain, i.e., location, velocity, acceleration, etc. Unlike the classic rectified flow formulation, which formulates a single ODE in the location domain and only captures the expected velocity field (sufficient to capture a multi-modal data distribution), the hierarchical rectified flow formulation models the multi-modal random velocity field, acceleration field, etc., in their entirety. This more faithful modeling of the random velocity field enables integration paths to intersect when the underlying ODE is solved during data generation. Intersecting paths in turn lead to integration trajectories that are more straight than those obtained in the classic rectified flow formulation, where integration paths cannot intersect. This leads to modeling of data distributions with fewer neural function evaluations. We empirically verify this on synthetic 1D and 2D data as well as MNIST, CIFAR-10, and ImageNet-32 data. Our code is available at: https://riccizz.github.io/HRF/.
Authors: Dung V. Nguyen, Minh H. Nguyen, Luc Q. Nguyen, Rachel S. Y. Teo, Tan M. Nguyen, Linh Duy Tran
Abstract: Existing methods for merging experts during model training and fine-tuning predominantly rely on Euclidean geometry, which assumes a flat parameter space. This assumption can limit the model's generalization ability, especially during the pre-training phase, where the parameter manifold might exhibit more complex curvature. Curvature-aware merging methods typically require additional information and computational resources to approximate the Fisher Information Matrix, adding memory overhead. In this paper, we introduce CAMEx (Curvature-Aware Merging of Experts), a novel expert merging protocol that incorporates natural gradients to account for the non-Euclidean curvature of the parameter manifold. By leveraging natural gradients, CAMEx adapts more effectively to the structure of the parameter space, improving alignment between model updates and the manifold's geometry. This approach enhances both pre-training and fine-tuning, resulting in better optimization trajectories and improved generalization without the substantial memory overhead typically associated with curvature-aware methods. Our contributions are threefold: (1) CAMEx significantly outperforms traditional Euclidean-based expert merging techniques across various natural language processing tasks, leading to enhanced performance during pre-training and fine-tuning; (2) we introduce a dynamic merging architecture that optimizes resource utilization, achieving high performance while reducing computational costs, facilitating efficient scaling of large language models; and (3) we provide both theoretical and empirical evidence to demonstrate the efficiency of our proposed method. The code is publicly available at: https://github.com/kpup1710/CAMEx.
Authors: Weilin Chen, Ruichu Cai, Junjie Wan, Zeqin Yang, Jos\'e Miguel Hern\'andez-Lobato
Abstract: Long-term causal inference has drawn increasing attention in many scientific domains. Existing methods mainly focus on estimating average long-term causal effects by combining long-term observational data and short-term experimental data. However, it is still understudied how to robustly and effectively estimate heterogeneous long-term causal effects, significantly limiting practical applications. In this paper, we propose several two-stage style nonparametric estimators for heterogeneous long-term causal effect estimation, including propensity-based, regression-based, and multiple robust estimators. We conduct a comprehensive theoretical analysis of their asymptotic properties under mild assumptions, with the ultimate goal of building a better understanding of the conditions under which some estimators can be expected to perform better. Extensive experiments across several semi-synthetic and real-world datasets validate the theoretical results and demonstrate the effectiveness of the proposed estimators.
Authors: Li Ju, Xingyi Yang, Qi Li, Xinchao Wang
Abstract: Graph neural networks (GNNs) are conventionally trained on a per-domain, per-task basis. It creates a significant barrier in transferring the acquired knowledge to different, heterogeneous data setups. This paper introduces GraphBridge, a novel framework to enable knowledge transfer across disparate tasks and domains in GNNs, circumventing the need for modifications to task configurations or graph structures. Specifically, GraphBridge allows for the augmentation of any pre-trained GNN with prediction heads and a bridging network that connects the input to the output layer. This architecture not only preserves the intrinsic knowledge of the original model but also supports outputs of arbitrary dimensions. To mitigate the negative transfer problem, GraphBridge merges the source model with a concurrently trained model, thereby reducing the source bias when applied to the target domain. Our method is thoroughly evaluated across diverse transfer learning scenarios, including Graph2Graph, Node2Node, Graph2Node, and graph2point-cloud. Empirical validation, conducted over 16 datasets representative of these scenarios, confirms the framework's capacity for task- and domain-agnostic transfer learning within graph-like data, marking a significant advancement in the field of GNNs. Code is available at https://github.com/jujulili888/GraphBridge.
Authors: Jan Wehner, Sahar Abdelnabi, Daniel Tan, David Krueger, Mario Fritz
Abstract: Representation Engineering (RepE) is a novel paradigm for controlling the behavior of LLMs. Unlike traditional approaches that modify inputs or fine-tune the model, RepE directly manipulates the model's internal representations. As a result, it may offer more effective, interpretable, data-efficient, and flexible control over models' behavior. We present the first comprehensive survey of RepE for LLMs, reviewing the rapidly growing literature to address key questions: What RepE methods exist and how do they differ? For what concepts and problems has RepE been applied? What are the strengths and weaknesses of RepE compared to other methods? To answer these, we propose a unified framework describing RepE as a pipeline comprising representation identification, operationalization, and control. We posit that while RepE methods offer significant potential, challenges remain, including managing multiple concepts, ensuring reliability, and preserving models' performance. Towards improving RepE, we identify opportunities for experimental and methodological improvements and construct a guide for best practices.
Authors: Aristotelis Ballas, Christos Diou
Abstract: Domain Generalization (DG) research has gained considerable traction as of late, since the ability to generalize to unseen data distributions is a requirement that eludes even state-of-the-art training algorithms. In this paper we observe that the initial iterations of model training play a key role in domain generalization effectiveness, since the loss landscape may be significantly different across the training and test distributions, contrary to the case of i.i.d. data. Conflicts between gradients of the loss components of each domain lead the optimization procedure to undesirable local minima that do not capture the domain-invariant features of the target classes. We propose alleviating domain conflicts in model optimization, by iteratively annealing the parameters of a model in the early stages of training and searching for points where gradients align between domains. By discovering a set of parameter values where gradients are updated towards the same direction for each data distribution present in the training set, the proposed Gradient-Guided Annealing (GGA) algorithm encourages models to seek out minima that exhibit improved robustness against domain shifts. The efficacy of GGA is evaluated on five widely accepted and challenging image classification domain generalization benchmarks, where its use alone is able to establish highly competitive or even state-of-the-art performance. Moreover, when combined with previously proposed domain-generalization algorithms it is able to consistently improve their effectiveness by significant margins.
Authors: Yongjia Lei, Haoyu Han, Ryan A. Rossi, Franck Dernoncourt, Nedim Lipka, Mahantesh M Halappanavar, Jiliang Tang, Yu Wang
Abstract: Text-rich Graph Knowledge Bases (TG-KBs) have become increasingly crucial for answering queries by providing textual and structural knowledge. However, current retrieval methods often retrieve these two types of knowledge in isolation without considering their mutual reinforcement and some hybrid methods even bypass structural retrieval entirely after neighboring aggregation. To fill in this gap, we propose a Mixture of Structural-and-Textual Retrieval (MoR) to retrieve these two types of knowledge via a Planning-Reasoning-Organizing framework. In the Planning stage, MoR generates textual planning graphs delineating the logic for answering queries. Following planning graphs, in the Reasoning stage, MoR interweaves structural traversal and textual matching to obtain candidates from TG-KBs. In the Organizing stage, MoR further reranks fetched candidates based on their structural trajectory. Extensive experiments demonstrate the superiority of MoR in harmonizing structural and textual retrieval with insights, including uneven retrieving performance across different query logics and the benefits of integrating structural trajectories for candidate reranking. Our code is available at https://github.com/Yoega/MoR.
Authors: Zhongyang Li, Ziyue Li, Tianyi Zhou
Abstract: In large multimodal models (LMMs), the perception of non-language modalities (e.g., visual representations) is usually not on par with the large language models (LLMs)' powerful reasoning capabilities, deterring LMMs' performance on challenging downstream tasks. This weakness has been recently mitigated by replacing the vision encoder with a mixture-of-experts (MoE), which provides rich, multi-granularity, and diverse representations required by diverse downstream tasks. The performance of multimodal MoE largely depends on its router, which reweights and mixes the representations of different experts for each input. However, we find that the end-to-end trained router does not always produce the optimal routing weights for every test sample. To bridge the gap, we propose a novel and efficient method "Re-Routing in Test-Time (R2-T2)" that locally optimizes the vector of routing weights in test-time by moving it toward those vectors of the correctly predicted samples in a neighborhood of the test sample. We propose three R2-T2 strategies with different optimization objectives and neighbor-search spaces. R2-T2 consistently and greatly improves state-of-the-art LMMs' performance on challenging benchmarks of diverse tasks, without training any base-model parameters.
Authors: Ruta Binkyte, Ivaxi Sheth, Zhijing Jin, Mohammad Havaei, Bernhard Sch\"olkopf, Mario Fritz
Abstract: Ensuring trustworthiness in machine learning (ML) systems is crucial as they become increasingly embedded in high-stakes domains. This paper advocates for integrating causal methods into machine learning to navigate the trade-offs among key principles of trustworthy ML, including fairness, privacy, robustness, accuracy, and explainability. While these objectives should ideally be satisfied simultaneously, they are often addressed in isolation, leading to conflicts and suboptimal solutions. Drawing on existing applications of causality in ML that successfully align goals such as fairness and accuracy or privacy and robustness, this paper argues that a causal approach is essential for balancing multiple competing objectives in both trustworthy ML and foundation models. Beyond highlighting these trade-offs, we examine how causality can be practically integrated into ML and foundation models, offering solutions to enhance their reliability and interpretability. Finally, we discuss the challenges, limitations, and opportunities in adopting causal frameworks, paving the way for more accountable and ethically sound AI systems.
Authors: Baiting Luo, Ava Pettet, Aron Laszka, Abhishek Dubey, Ayan Mukhopadhyay
Abstract: Sequential decision-making in high-dimensional continuous action spaces, particularly in stochastic environments, faces significant computational challenges. We explore this challenge in the traditional offline RL setting, where an agent must learn how to make decisions based on data collected through a stochastic behavior policy. We present Latent Macro Action Planner (L-MAP), which addresses this challenge by learning a set of temporally extended macro-actions through a state-conditional Vector Quantized Variational Autoencoder (VQ-VAE), effectively reducing action dimensionality. L-MAP employs a (separate) learned prior model that acts as a latent transition model and allows efficient sampling of plausible actions. During planning, our approach accounts for stochasticity in both the environment and the behavior policy by using Monte Carlo tree search (MCTS). In offline RL settings, including stochastic continuous control tasks, L-MAP efficiently searches over discrete latent actions to yield high expected returns. Empirical results demonstrate that L-MAP maintains low decision latency despite increased action dimensionality. Notably, across tasks ranging from continuous control with inherently stochastic dynamics to high-dimensional robotic hand manipulation, L-MAP significantly outperforms existing model-based methods and performs on-par with strong model-free actor-critic baselines, highlighting the effectiveness of the proposed approach in planning in complex and stochastic environments with high-dimensional action spaces.
Authors: Massimo La Morgia, Alessandro Mei, Alberto Maria Mongardini, Jie Wu
Abstract: Telegram is one of the most used instant messaging apps worldwide. Some of its success lies in providing high privacy protection and social network features like the channels -- virtual rooms in which only the admins can post and broadcast messages to all its subscribers. However, these same features contributed to the emergence of borderline activities and, as is common with Online Social Networks, the heavy presence of fake accounts. Telegram started to address these issues by introducing the verified and scam marks for the channels. Unfortunately, the problem is far from being solved. In this work, we perform a large-scale analysis of Telegram by collecting 35,382 different channels and over 130,000,000 messages. We study the channels that Telegram marks as verified or scam, highlighting analogies and differences. Then, we move to the unmarked channels. Here, we find some of the infamous activities also present on privacy-preserving services of the Dark Web, such as carding, sharing of illegal adult and copyright protected content. In addition, we identify and analyze two other types of channels: the clones and the fakes. Clones are channels that publish the exact content of another channel to gain subscribers and promote services. Instead, fakes are channels that attempt to impersonate celebrities or well-known services. Fakes are hard to identify even by the most advanced users. To detect the fake channels automatically, we propose a machine learning model that is able to identify them with an accuracy of 86%. Lastly, we study Sabmyk, a conspiracy theory that exploited fakes and clones to spread quickly on the platform reaching over 1,000,000 users.
Authors: Haoran You, Baopu Li, Zhanyi Sun, Xu Ouyang, Yingyan Celine Lin
Abstract: Neural architecture search (NAS) has demonstrated amazing success in searching for efficient deep neural networks (DNNs) from a given supernet. In parallel, the lottery ticket hypothesis has shown that DNNs contain small subnetworks that can be trained from scratch to achieve a comparable or higher accuracy than original DNNs. As such, it is currently a common practice to develop efficient DNNs via a pipeline of first search and then prune. Nevertheless, doing so often requires a search-train-prune-retrain process and thus prohibitive computational cost. In this paper, we discover for the first time that both efficient DNNs and their lottery subnetworks (i.e., lottery tickets) can be directly identified from a supernet, which we term as SuperTickets, via a two-in-one training scheme with jointly architecture searching and parameter pruning. Moreover, we develop a progressive and unified SuperTickets identification strategy that allows the connectivity of subnetworks to change during supernet training, achieving better accuracy and efficiency trade-offs than conventional sparse training. Finally, we evaluate whether such identified SuperTickets drawn from one task can transfer well to other tasks, validating their potential of handling multiple tasks simultaneously. Extensive experiments and ablation studies on three tasks and four benchmark datasets validate that our proposed SuperTickets achieve boosted accuracy and efficiency trade-offs than both typical NAS and pruning pipelines, regardless of having retraining or not. Codes and pretrained models are available at https://github.com/RICE-EIC/SuperTickets.
Authors: Khashayar Namdar, Matthias W. Wagner, Birgit B. Ertl-Wagner, Farzad Khalvati
Abstract: Background: As an important branch of machine learning pipelines in medical imaging, radiomics faces two major challenges namely reproducibility and accessibility. In this work, we introduce open-radiomics, a set of radiomics datasets along with a comprehensive radiomics pipeline based on our proposed technical protocol to improve the reproducibility of the results. Methods: We curated large-scale radiomics datasets based on three open-source datasets; BraTS 2020 for high-grade glioma (HGG) versus low-grade glioma (LGG) classification and survival analysis, BraTS 2023 for O6-methylguanine-DNA methyltransferase classification, and non-small cell lung cancer survival analysis from the Cancer Imaging Archive. Using BraTS 2020 Magnetic Resonance Imaging (MRI) dataset, we applied our protocol to 369 brain tumor patients (76 LGG, 293 HGG). Leveraging PyRadiomics for LGG vs. HGG classification, we generated 288 datasets from 4 MRI sequences, 3 binWidths, 6 normalization methods, and 4 tumor subregions. Random Forest classifiers were trained and validated (60%,20%,20%) across 100 different data splits (28,800 test results), evaluating Area Under the Receiver Operating Characteristic Curve (AUROC). Results: Unlike binWidth and image normalization, tumor subregion and imaging sequence significantly affected performance of the models. T1 contrast-enhanced sequence and the union of Necrotic and the non-enhancing tumor core subregions resulted in the highest AUROCs (average test AUROC 0.951, 95% confidence interval of (0.949, 0.952)). Although several settings and data splits (28 out of 28800) yielded test AUROC of 1, they were irreproducible. Conclusion: Our experiments demonstrate the sources of variability in radiomics pipelines (e.g., tumor subregion) can have a significant impact on the results, which may lead to superficial perfect performances that are irreproducible.
Authors: Yiming Li, Linghui Zhu, Xiaojun Jia, Yang Bai, Yong Jiang, Shu-Tao Xia, Xiaochun Cao, Kui Ren
Abstract: Currently, deep neural networks (DNNs) are widely adopted in different applications. Despite its commercial values, training a well-performing DNN is resource-consuming. Accordingly, the well-trained model is valuable intellectual property for its owner. However, recent studies revealed the threats of model stealing, where the adversaries can obtain a function-similar copy of the victim model, even when they can only query the model. In this paper, we propose an effective and harmless model ownership verification (MOVE) to defend against different types of model stealing simultaneously, without introducing new security risks. In general, we conduct the ownership verification by verifying whether a suspicious model contains the knowledge of defender-specified external features. Specifically, we embed the external features by modifying a few training samples with style transfer. We then train a meta-classifier to determine whether a model is stolen from the victim. This approach is inspired by the understanding that the stolen models should contain the knowledge of features learned by the victim model. In particular, \revision{we develop our MOVE method under both white-box and black-box settings and analyze its theoretical foundation to provide comprehensive model protection.} Extensive experiments on benchmark datasets verify the effectiveness of our method and its resistance to potential adaptive attacks. The codes for reproducing the main experiments of our method are available at https://github.com/THUYimingLi/MOVE.
Authors: Jack M. Buckingham, Sebastian Rojas Gonzalez, Juergen Branke
Abstract: Multi-objective Bayesian optimization aims to find the Pareto front of trade-offs between a set of expensive objectives while collecting as few samples as possible. In some cases, it is possible to evaluate the objectives separately, and a different latency or evaluation cost can be associated with each objective. This decoupling of the objectives presents an opportunity to learn the Pareto front faster by avoiding unnecessary, expensive evaluations. We propose a scalarization based knowledge gradient acquisition function which accounts for the different evaluation costs of the objectives. We prove asymptotic consistency of the estimator of the optimum for an arbitrary, D-dimensional, real compact search space and show empirically that the algorithm performs comparably with the state of the art and significantly outperforms versions which always evaluate both objectives.
Authors: Licong Lin, Koulik Khamaru, Martin J. Wainwright
Abstract: Many standard estimators, when applied to adaptively collected data, fail to be asymptotically normal, thereby complicating the construction of confidence intervals. We address this challenge in a semi-parametric context: estimating the parameter vector of a generalized linear regression model contaminated by a non-parametric nuisance component. We construct suitably weighted estimating equations that account for adaptivity in data collection, and provide conditions under which the associated estimates are asymptotically normal. Our results characterize the degree of "explorability" required for asymptotic normality to hold. For the simpler problem of estimating a linear functional, we provide similar guarantees under much weaker assumptions. We illustrate our general theory with concrete consequences for various problems, including standard linear bandits and sparse generalized bandits, and compare with other methods via simulation studies.
Authors: Vincent P. Grande, Michael T. Schaub
Abstract: We present Topological Point Cloud Clustering (TPCC), a new method to cluster points in an arbitrary point cloud based on their contribution to global topological features. TPCC synthesizes desirable features from spectral clustering and topological data analysis and is based on considering the spectral properties of a simplicial complex associated to the considered point cloud. As it is based on considering sparse eigenvector computations, TPCC is similarly easy to interpret and implement as spectral clustering. However, by focusing not just on a single matrix associated to a graph created from the point cloud data, but on a whole set of Hodge-Laplacians associated to an appropriately constructed simplicial complex, we can leverage a far richer set of topological features to characterize the data points within the point cloud and benefit from the relative robustness of topological techniques against noise. We test the performance of TPCC on both synthetic and real-world data and compare it with classical spectral clustering.
Authors: Shiyu Wang, Yinbo Sun, Xiaoming Shi, Shiyi Zhu, Lin-Tao Ma, James Zhang, Yifei Zheng, Jian Liu
Abstract: The rapid rise in cloud computing has resulted in an alarming increase in data centers' carbon emissions, which now accounts for >3% of global greenhouse gas emissions, necessitating immediate steps to combat their mounting strain on the global climate. An important focus of this effort is to improve resource utilization in order to save electricity usage. Our proposed Full Scaling Automation (FSA) mechanism is an effective method of dynamically adapting resources to accommodate changing workloads in large-scale cloud computing clusters, enabling the clusters in data centers to maintain their desired CPU utilization target and thus improve energy efficiency. FSA harnesses the power of deep representation learning to accurately predict the future workload of each service and automatically stabilize the corresponding target CPU usage level, unlike the previous autoscaling methods, such as Autopilot or FIRM, that need to adjust computing resources with statistical models and expert knowledge. Our approach achieves significant performance improvement compared to the existing work in real-world datasets. We also deployed FSA on large-scale cloud computing clusters in industrial data centers, and according to the certification of the China Environmental United Certification Center (CEC), a reduction of 947 tons of carbon dioxide, equivalent to a saving of 1538,000 kWh of electricity, was achieved during the Double 11 shopping festival of 2022, marking a critical step for our company's strategic goal towards carbon neutrality by 2030.
Authors: David Fleischer, David A. Stephens, Archer Yang
Abstract: We propose a computationally efficient alternative to generalized random forests arXiv:1610.01271 (GRFs) for estimating heterogeneous effects in large dimensions. While GRFs rely on a gradient-based splitting criterion, which is large dimensions is computationally expensive and unstable, our method introduces a fixed-point approximation that eliminates the need for Jacobian estimation. This gradient-free approach preserves GRFs theoretical guarantees of consistency and asymptotic normality while significantly improving computational efficiency. We demonstrate that our method achieves multiple times the speed over standard GRFs without compromising statistical accuracy. Experiments on both simulated and real-world data, validate our approach. Our findings suggest that the proposed method is a scalable alternative for localized effect estimation in machine learning and causal inference applications.
Authors: Satoshi Yamamori, Jun Morimoto
Abstract: In this study, we propose a multitask reinforcement learning algorithm for foundational policy acquisition to generate novel motor skills. \textcolor{\hcolor}{Learning the rich representation of the multitask policy is a challenge in dynamic movement generation tasks because the policy needs to cope with changes in goals or environments with different reward functions or physical parameters. Inspired by human sensorimotor adaptation mechanisms, we developed the learning pipeline to construct the encoder-decoder networks and network selection to facilitate foundational policy acquisition under multiple situations. First, we compared the proposed method with previous multitask reinforcement learning methods in the standard multi-locomotion tasks. The results showed that the proposed approach outperformed the baseline methods. Then, we applied the proposed method to the ball heading task using a monopod robot model to evaluate skill generation performance. The results showed that the proposed method was able to adapt to novel target positions or inexperienced ball restitution coefficients but to acquire a foundational policy network, originally learned for heading motion, which can generate an entirely new overhead kicking skill.
Authors: Rebecca M. Hurwitz, Georg Hahn
Abstract: Principal components computed via PCA (principal component analysis) are traditionally used to reduce dimensionality in genomic data or to correct for population stratification. In this paper, we explore the penalized eigenvalue problem (PEP) which reformulates the computation of the first eigenvector as an optimization problem and adds an $L_1$ penalty constraint to enforce sparseness of the solution. The contribution of our article is threefold. First, we extend PEP by applying smoothing to the original LASSO-type $L_1$ penalty. This allows one to compute analytical gradients which enable faster and more efficient minimization of the objective function associated with the optimization problem. Second, we demonstrate how higher order eigenvectors can be calculated with PEP using established results from singular value decomposition (SVD). Third, we present four experimental studies to demonstrate the usefulness of the smoothed penalized eigenvectors. Using data from the 1000 Genomes Project dataset, we empirically demonstrate that our proposed smoothed PEP allows one to increase numerical stability and obtain meaningful eigenvectors. We also employ the penalized eigenvector approach in two additional real data applications (computation of a polygenic risk score and clustering), demonstrating that exchanging the penalized eigenvectors for their smoothed counterparts can increase prediction accuracy in polygenic risk scores and enhance discernibility of clusterings. Moreover, we compare our proposed smoothed PEP to seven state-of-the-art algorithms for sparse PCA and evaluate the accuracy of the obtained eigenvectors, their support recovery, and their runtime.
Authors: Weidong Liu, Jiyuan Tu, Xi Chen, Yichen Zhang
Abstract: Reinforcement learning has emerged as one of the prominent topics attracting attention in modern statistical learning, with policy evaluation being a key component. Unlike the traditional machine learning literature on this topic, our work emphasizes statistical inference for the model parameters and value functions of reinforcement learning algorithms. While most existing analyses assume random rewards to follow standard distributions, we embrace the concept of robust statistics in reinforcement learning by simultaneously addressing issues of outlier contamination and heavy-tailed rewards within a unified framework. In this paper, we develop a fully online robust policy evaluation procedure, and establish the Bahadur-type representation of our estimator. Furthermore, we develop an online procedure to efficiently conduct statistical inference based on the asymptotic distribution. This paper connects robust statistics and statistical inference in reinforcement learning, offering a more versatile and reliable approach to online policy evaluation. Finally, we validate the efficacy of our algorithm through numerical experiments conducted in simulations and real-world reinforcement learning experiments.
Authors: Marcel Kollovieh, Lukas Gosch, Yan Scholten, Marten Lienen, Stephan G\"unnemann
Abstract: Most adversarial attacks and defenses focus on perturbations within small $\ell_p$-norm constraints. However, $\ell_p$ threat models cannot capture all relevant semantics-preserving perturbations, and hence, the scope of robustness evaluations is limited. In this work, we introduce Score-Based Adversarial Generation (ScoreAG), a novel framework that leverages the advancements in score-based generative models to generate unrestricted adversarial examples that overcome the limitations of $\ell_p$-norm constraints. Unlike traditional methods, ScoreAG maintains the core semantics of images while generating adversarial examples, either by transforming existing images or synthesizing new ones entirely from scratch. We further exploit the generative capability of ScoreAG to purify images, empirically enhancing the robustness of classifiers. Our extensive empirical evaluation demonstrates that ScoreAG improves upon the majority of state-of-the-art attacks and defenses across multiple benchmarks. This work highlights the importance of investigating adversarial examples bounded by semantics rather than $\ell_p$-norm constraints. ScoreAG represents an important step towards more encompassing robustness assessments.
Authors: Ruohao Guo, Xianghua Ying, Yaru Chen, Dantong Niu, Guangyao Li, Liao Qu, Yanyu Qi, Jinxing Zhou, Bowei Xing, Wenzhen Yue, Ji Shi, Qixun Wang, Peiliang Zhang, Buwen Liang
Abstract: In this paper, we propose a new multi-modal task, termed audio-visual instance segmentation (AVIS), which aims to simultaneously identify, segment and track individual sounding object instances in audible videos. To facilitate this research, we introduce a high-quality benchmark named AVISeg, containing over 90K instance masks from 26 semantic categories in 926 long videos. Additionally, we propose a strong baseline model for this task. Our model first localizes sound source within each frame, and condenses object-specific contexts into concise tokens. Then it builds long-range audio-visual dependencies between these tokens using window-based attention, and tracks sounding objects among the entire video sequences. Extensive experiments reveal that our method performs best on AVISeg, surpassing the existing methods from related tasks. We further conduct the evaluation on several multi-modal large models. Unfortunately, they exhibits subpar performance on instance-level sound source localization and temporal perception. We expect that AVIS will inspire the community towards a more comprehensive multi-modal understanding. Dataset and code is available at https://github.com/ruohaoguo/avis.
Authors: Dzung Pham, Shreyas Kulkarni, Amir Houmansadr
Abstract: Federated learning has emerged as a promising privacy-preserving solution for machine learning domains that rely on user interactions, particularly recommender systems and online learning to rank. While there has been substantial research on the privacy of traditional federated learning, little attention has been paid to the privacy properties of these interaction-based settings. In this work, we show that users face an elevated risk of having their private interactions reconstructed by the central server when the server can control the training features of the items that users interact with. We introduce RAIFLE, a novel optimization-based attack framework where the server actively manipulates the features of the items presented to users to increase the success rate of reconstruction. Our experiments with federated recommendation and online learning-to-rank scenarios demonstrate that RAIFLE is significantly more powerful than existing reconstruction attacks like gradient inversion, achieving high performance consistently in most settings. We discuss the pros and cons of several possible countermeasures to defend against RAIFLE in the context of interaction-based federated learning. Our code is open-sourced at https://github.com/dzungvpham/raifle.
Authors: Mohammad Junayed Hasan, M. R. C. Mahdy
Abstract: Quantum neural networks (QNNs), harnessing superposition and entanglement, have shown potential to surpass classical methods in complex learning tasks but remain limited by hardware constraints and noisy conditions. In this work, we present a novel framework for transferring knowledge from classical convolutional neural networks (CNNs) to QNNs via knowledge distillation, thereby reducing the need for resource intensive quantum training and error mitigation. We conduct extensive experiments using two parameterized quantum circuits (PQCs) with 4 and 8 qubits on MNIST, Fashion MNIST, and CIFAR10 datasets. The approach demonstrates consistent accuracy improvements attributed to distilled knowledge from larger classical networks. Through ablation studies, we systematically compare the effect of state of the art dimensionality reduction techniques fully connected layers, center cropping, principal component analysis, and pooling to compress high-dimensional image data prior to quantum encoding. Our findings reveal that fully connected layers retain the most salient features for QNN inference, thereby surpassing other down sampling approaches. Additionally, we examine state of the art data encoding methods (amplitude, angle, and qubit encoding) and identify amplitude encoding as the optimal strategy, yielding superior accuracy across all tested datasets and qubit configurations. Through computational analyses, we show that our distilled 4-qubit and 8-qubit QNNs achieve competitive performance while utilizing significantly fewer parameters than their classical counterparts. Our results establish a promising paradigm for bridging classical deep learning and emerging quantum computing, paving the way for more powerful, resource conscious models in quantum machine intelligence.
Authors: Lokesh Veeramacheneni (University of Bonn), Moritz Wolter (University of Bonn), Hildegard Kuehne (University of Tuebingen, MIT-IBM Watson AI Lab), Juergen Gall (University of Bonn, Lamarr Institute for Machine Learning,Artificial Intelligence)
Abstract: Modern metrics for generative learning like Fr\'echet Inception Distance (FID) and DINOv2-Fr\'echet Distance (FD-DINOv2) demonstrate impressive performance. However, they suffer from various shortcomings, like a bias towards specific generators and datasets. To address this problem, we propose the Fr\'echet Wavelet Distance (FWD) as a domain-agnostic metric based on the Wavelet Packet Transform ($W_p$). FWD provides a sight across a broad spectrum of frequencies in images with a high resolution, preserving both spatial and textural aspects. Specifically, we use $W_p$ to project generated and real images to the packet coefficient space. We then compute the Fr\'echet distance with the resultant coefficients to evaluate the quality of a generator. This metric is general-purpose and dataset-domain agnostic, as it does not rely on any pre-trained network, while being more interpretable due to its ability to compute Fr\'echet distance per packet, enhancing transparency. We conclude with an extensive evaluation of a wide variety of generators across various datasets that the proposed FWD can generalize and improve robustness to domain shifts and various corruptions compared to other metrics.
Authors: Binh M. Le, Jiwon Kim, Simon S. Woo, Kristen Moore, Alsharif Abuadbba, Shahroz Tariq
Abstract: Deepfakes have rapidly emerged as a serious threat to society due to their ease of creation and dissemination, triggering the accelerated development of detection technologies. However, many existing detectors rely on labgenerated datasets for validation, which may not prepare them for novel, real-world deepfakes. This paper extensively reviews and analyzes state-of-the-art deepfake detectors, evaluating them against several critical criteria. These criteria categorize detectors into 4 high-level groups and 13 finegrained sub-groups, aligned with a unified conceptual framework we propose. This classification offers practical insights into the factors affecting detector efficacy. We evaluate the generalizability of 16 leading detectors across comprehensive attack scenarios, including black-box, white-box, and graybox settings. Our systematized analysis and experiments provide a deeper understanding of deepfake detectors and their generalizability, paving the way for future research and the development of more proactive defenses against deepfakes.
Authors: Achref Jaziri, Sina Ditzel, Iuliia Pliushch, Visvanathan Ramesh
Abstract: Modern data-driven machine learning system designs exploit inductive biases in architectural structure, invariance and equivariance requirements, task-specific loss functions, and computational optimization tools. Previous works have illustrated that human-specified quasi-invariant filters can serve as a powerful inductive bias in the early layers of the encoder, enhancing robustness and transparency in learned classifiers. This paper explores this further within the context of representation learning with bio-inspired Hebbian learning rules. We propose a modular framework trained with a bio-inspired variant of contrastive predictive coding, comprising parallel encoders that leverage different invariant visual descriptors as inductive biases. We evaluate the representation learning capacity of our system in classification scenarios using diverse image datasets (GTSRB, STL10, CODEBRIM) and video datasets (UCF101). Our findings indicate that this form of inductive bias significantly improves the robustness of learned representations and narrows the performance gap between models using local Hebbian plasticity rules and those using backpropagation, while also achieving superior performance compared to non-decomposed encoders.
Authors: Yiling Xie, Xiaoming Huo
Abstract: Adversarial training has been proposed to protect machine learning models against adversarial attacks. This paper focuses on adversarial training under $\ell_\infty$-perturbation, which has recently attracted much research attention. The asymptotic behavior of the adversarial training estimator is investigated in the generalized linear model. The results imply that the asymptotic distribution of the adversarial training estimator under $\ell_\infty$-perturbation could put a positive probability mass at $0$ when the true parameter is $0$, providing a theoretical guarantee of the associated sparsity-recovery ability. Alternatively, a two-step procedure is proposed -- adaptive adversarial training, which could further improve the performance of adversarial training under $\ell_\infty$-perturbation. Specifically, the proposed procedure could achieve asymptotic variable-selection consistency and unbiasedness. Numerical experiments are conducted to show the sparsity-recovery ability of adversarial training under $\ell_\infty$-perturbation and to compare the empirical performance between classic adversarial training and adaptive adversarial training.
Authors: Sahil Gulania, Yuri Alexeev, Stephen K. Gray, Bo Peng, Niranjan Govind
Abstract: Quantum computing shows great potential, but errors pose a significant challenge. This study explores new strategies for mitigating quantum errors using artificial neural networks (ANN) and the Yang-Baxter equation (YBE). Unlike traditional error mitigation methods, which are computationally intensive, we investigate artificial error mitigation. We developed a novel method that combines ANN for noise mitigation combined with the YBE to generate noisy data. This approach effectively reduces noise in quantum simulations, enhancing the accuracy of the results. The YBE rigorously preserves quantum correlations and symmetries in spin chain simulations in certain classes of integrable lattice models, enabling effective compression of quantum circuits while retaining linear scalability with the number of qubits. This compression facilitates both full and partial implementations, allowing the generation of noisy quantum data on hardware alongside noiseless simulations using classical platforms. By introducing controlled noise through the YBE, we enhance the dataset for error mitigation. We train an ANN model on partial data from quantum simulations, demonstrating its effectiveness in mitigating errors in time-evolving quantum states, providing a scalable framework to enhance quantum computation fidelity, particularly in noisy intermediate-scale quantum (NISQ) systems. We demonstrate the efficacy of this approach by performing quantum time dynamics simulations using the Heisenberg XY Hamiltonian on real quantum devices.
Authors: Chinmay Mittal, Krishna Kartik, Mausam, Parag Singla
Abstract: Can the large language models (LLMs) solve challenging first-order combinatorial reasoning problems such as graph coloring, knapsack, and cryptarithmetic? By first-order, we mean these problems can be instantiated with potentially an infinite number of problem instances of varying sizes. They are also challenging being NP-hard and requiring several reasoning steps to reach a solution. While existing work has focused on coming up with datasets with hard benchmarks, there is limited work which exploits the first-order nature of the problem structure. To address this challenge, we present FCoReBench, a dataset of 40 such challenging problems, along with scripts to generate problem instances of varying sizes and automatically verify and generate their solutions. We first observe that LLMs, even when aided by symbolic solvers, perform rather poorly on our dataset, being unable to leverage the underlying structure of these problems. We specifically observe a drop in performance with increasing problem size. In response, we propose a new approach, SymPro-LM, which combines LLMs with both symbolic solvers and program interpreters, along with feedback from a few solved examples, to achieve huge performance gains. Our proposed approach is robust to changes in the problem size, and has the unique characteristic of not requiring any LLM call during inference time, unlike earlier approaches. As an additional experiment, we also demonstrate SymPro-LM's effectiveness on other logical reasoning benchmarks.
Authors: Bhisham Dev Verma, Rameshwar Pratap
Abstract: Locality sensitive hashing (LSH) is a fundamental algorithmic toolkit used by data scientists for approximate nearest neighbour search problems that have been used extensively in many large scale data processing applications such as near duplicate detection, nearest neighbour search, clustering, etc. In this work, we aim to propose faster and space efficient locality sensitive hash functions for Euclidean distance and cosine similarity for tensor data. Typically, the naive approach for obtaining LSH for tensor data involves first reshaping the tensor into vectors, followed by applying existing LSH methods for vector data $E2LSH$ and $SRP$. However, this approach becomes impractical for higher order tensors because the size of the reshaped vector becomes exponential in the order of the tensor. Consequently, the size of LSH parameters increases exponentially. To address this problem, we suggest two methods for LSH for Euclidean distance and cosine similarity, namely $CP-E2LSH$, $TT-E2LSH$, and $CP-SRP$, $TT-SRP$, respectively, building on $CP$ and tensor train $(TT)$ decompositions techniques. Our approaches are space efficient and can be efficiently applied to low rank $CP$ or $TT$ tensors. We provide a rigorous theoretical analysis of our proposal on their correctness and efficacy.
Authors: Niklas Muennighoff, Hongjin Su, Liang Wang, Nan Yang, Furu Wei, Tao Yu, Amanpreet Singh, Douwe Kiela
Abstract: All text-based language problems can be reduced to either generation or embedding. Current models only perform well at one or the other. We introduce generative representational instruction tuning (GRIT) whereby a large language model is trained to handle both generative and embedding tasks by distinguishing between them through instructions. Compared to other open models, our resulting GritLM 7B sets a new state of the art on the Massive Text Embedding Benchmark (MTEB) and outperforms all models up to its size on a range of generative tasks. By scaling up further, GritLM 8x7B outperforms all open generative language models that we tried while still being among the best embedding models. Notably, we find that GRIT matches training on only generative or embedding data, thus we can unify both at no performance loss. Among other benefits, the unification via GRIT speeds up Retrieval-Augmented Generation (RAG) by > 60% for long documents, by no longer requiring separate retrieval and generation models. Models, code, etc. are freely available at https://github.com/ContextualAI/gritlm.
Authors: Yibin Chen, Yifu Yuan, Zeyu Zhang, Yan Zheng, Jinyi Liu, Fei Ni, Jianye Hao, Hangyu Mao, Fuzheng Zhang
Abstract: Spreadsheets are ubiquitous across the World Wide Web, playing a critical role in enhancing work efficiency across various domains. Large language model (LLM) has been recently attempted for automatic spreadsheet manipulation but has not yet been investigated in complicated and realistic tasks where reasoning challenges exist (e.g., long horizon manipulation with multi-step reasoning and ambiguous requirements). To bridge the gap with the real-world requirements, we introduce SheetRM, a benchmark featuring long-horizon and multi-category tasks with reasoning-dependent manipulation caused by real-life challenges. To mitigate the above challenges, we further propose SheetAgent, a novel autonomous agent that utilizes the power of LLMs. SheetAgent consists of three collaborative modules: Planner, Informer, and Retriever, achieving both advanced reasoning and accurate manipulation over spreadsheets without human interaction through iterative task reasoning and reflection. Extensive experiments demonstrate that SheetAgent delivers 20--40\% pass rate improvements on multiple benchmarks over baselines, achieving enhanced precision in spreadsheet manipulation and demonstrating superior table reasoning abilities. More details and visualizations are available at the project website: https://sheetagent.github.io/. The datasets and source code are available at https://anonymous.4open.science/r/SheetAgent.
URLs: https://sheetagent.github.io/., https://anonymous.4open.science/r/SheetAgent.
Authors: Zhuang Liu, Kaiming He
Abstract: We revisit the "dataset classification" experiment suggested by Torralba & Efros (2011) a decade ago, in the new era with large-scale, diverse, and hopefully less biased datasets as well as more capable neural network architectures. Surprisingly, we observe that modern neural networks can achieve excellent accuracy in classifying which dataset an image is from: e.g., we report 84.7% accuracy on held-out validation data for the three-way classification problem consisting of the YFCC, CC, and DataComp datasets. Our further experiments show that such a dataset classifier could learn semantic features that are generalizable and transferable, which cannot be explained by memorization. We hope our discovery will inspire the community to rethink issues involving dataset bias.
Authors: Jingling Li, Zeyu Tang, Xiaoyu Liu, Peter Spirtes, Kun Zhang, Liu Leqi, Yang Liu
Abstract: Large language models (LLMs), despite their remarkable capabilities, are susceptible to generating biased and discriminatory responses. As LLMs increasingly influence high-stakes decision-making (e.g., hiring and healthcare), mitigating these biases becomes critical. In this work, we propose a causality-guided debiasing framework to tackle social biases, aiming to reduce the objectionable dependence between LLMs' decisions and the social information in the input. Our framework introduces a novel perspective to identify how social information can affect an LLM's decision through different causal pathways. Leveraging these causal insights, we outline principled prompting strategies that regulate these pathways through selection mechanisms. This framework not only unifies existing prompting-based debiasing techniques, but also opens up new directions for reducing bias by encouraging the model to prioritize fact-based reasoning over reliance on biased social cues. We validate our framework through extensive experiments on real-world datasets across multiple domains, demonstrating its effectiveness in debiasing LLM decisions, even with only black-box access to the model.
Authors: Lingdong Kong, Xiang Xu, Jun Cen, Wenwei Zhang, Liang Pan, Kai Chen, Ziwei Liu
Abstract: Safety-critical 3D scene understanding tasks necessitate not only accurate but also confident predictions from 3D perception models. This study introduces Calib3D, a pioneering effort to benchmark and scrutinize the reliability of 3D scene understanding models from an uncertainty estimation viewpoint. We comprehensively evaluate 28 state-of-the-art models across 10 diverse 3D datasets, uncovering insightful phenomena that cope with both the aleatoric and epistemic uncertainties in 3D scene understanding. We discover that despite achieving impressive levels of accuracy, existing models frequently fail to provide reliable uncertainty estimates -- a pitfall that critically undermines their applicability in safety-sensitive contexts. Through extensive analysis of key factors such as network capacity, LiDAR representations, rasterization resolutions, and 3D data augmentation techniques, we correlate these aspects directly with the model calibration efficacy. Furthermore, we introduce DeptS, a novel depth-aware scaling approach aimed at enhancing 3D model calibration. Extensive experiments across a wide range of configurations validate the superiority of our method. We hope this work could serve as a cornerstone for fostering reliable 3D scene understanding. Code and benchmark toolkit are publicly available.
Authors: Andrey Gromov, Kushal Tirumala, Hassan Shapourian, Paolo Glorioso, Daniel A. Roberts
Abstract: How is knowledge stored in an LLM's weights? We study this via layer pruning: if removing a certain layer does not affect model performance in common question-answering benchmarks, then the weights in that layer are not necessary for storing the knowledge needed to answer those questions. To find these unnecessary parameters, we identify the optimal block of layers to prune by considering similarity across layers; then, to "heal" the damage, we perform a small amount of finetuning. Surprisingly, with this method we find minimal degradation of performance until after a large fraction (up to half) of the layers are removed for some common open-weight models. From a scientific perspective, the robustness of these LLMs to the deletion of layers implies either that current pretraining methods are not properly leveraging the parameters in the deeper layers of the network or that the shallow layers play a critical role in storing knowledge. For our study, we use parameter-efficient finetuning (PEFT) methods, specifically quantization and Low Rank Adapters (QLoRA), such that each of our experiments can be performed on a single 40GB A100 GPU.
Authors: Ricard Puig, Marc Drudis, Supanut Thanasilp, Zo\"e Holmes
Abstract: The barren plateau phenomenon, characterized by loss gradients that vanish exponentially with system size, poses a challenge to scaling variational quantum algorithms. Here we explore the potential of warm starts, whereby one initializes closer to a solution in the hope of enjoying larger loss variances. Focusing on an iterative variational method for learning shorter-depth circuits for quantum real time evolution we conduct a case study to elucidate the potential and limitations of warm starts. We start by proving that the iterative variational algorithm will exhibit substantial (at worst vanishing polynomially in system size) gradients in a small region around the initializations at each time-step. Convexity guarantees for these regions are then established, suggesting trainability for polynomial size time-steps. However, our study highlights scenarios where a good minimum shifts outside the region with trainability guarantees. Our analysis leaves open the question whether such minima jumps necessitate optimization across barren plateau landscapes or whether there exist gradient flows, i.e., fertile valleys away from the plateau with substantial gradients, that allow for training. While our main focus is on this case study of variational quantum simulation, we end by discussing how our results work in other iterative settings.
Authors: Hubert Eichner, Daniel Ramage, Kallista Bonawitz, Dzmitry Huba, Tiziano Santoro, Brett McLarnon, Timon Van Overveldt, Nova Fallen, Peter Kairouz, Albert Cheu, Katharine Daly, Adria Gascon, Marco Gruteser, Brendan McMahan
Abstract: Federated Learning and Analytics (FLA) have seen widespread adoption by technology platforms for processing sensitive on-device data. However, basic FLA systems have privacy limitations: they do not necessarily require anonymization mechanisms like differential privacy (DP), and provide limited protections against a potentially malicious service provider. Adding DP to a basic FLA system currently requires either adding excessive noise to each device's updates, or assuming an honest service provider that correctly implements the mechanism and only uses the privatized outputs. Secure multiparty computation (SMPC) -based oblivious aggregations can limit the service provider's access to individual user updates and improve DP tradeoffs, but the tradeoffs are still suboptimal, and they suffer from scalability challenges and susceptibility to Sybil attacks. This paper introduces a novel system architecture that leverages trusted execution environments (TEEs) and open-sourcing to both ensure confidentiality of server-side computations and provide externally verifiable privacy properties, bolstering the robustness and trustworthiness of private federated computations.
Authors: Peiyang Song, Kaiyu Yang, Anima Anandkumar
Abstract: Neural theorem proving combines large language models (LLMs) with proof assistants such as Lean, where the correctness of formal proofs can be rigorously verified, leaving no room for hallucination. With existing neural theorem provers pretrained on a fixed collection of data and offering valuable suggestions at times, it is challenging for them to continually prove novel theorems in a fully autonomous mode, where human insights may be critical. In this paper, we explore LLMs as copilots that assist humans in proving theorems. We introduce Lean Copilot, an general framework for running LLM inference natively in Lean. It enables programmers to build various LLM-based proof automation tools that integrate seamlessly into the workflow of Lean users. Lean users can use our pretrained models or bring their own ones that run either locally (with or without GPUs) or on the cloud. Using Lean Copilot, we build LLM-based tools that suggest proof steps, complete proof goals, and select relevant premises. Experimental results on the Mathematics in Lean textbook demonstrate the effectiveness of our method compared to existing rule-based proof automation in Lean (aesop). When assisting humans, Lean Copilot requires only 2.08 manually-entered proof steps on average (3.86 required by aesop); when automating the theorem proving process, Lean Copilot automates 74.2% proof steps on average, 85% better than aesop (40.1%). We open source all code and artifacts under a permissive MIT license to facilitate further research.
Authors: Zerui Chen, Shizhe Chen, Etienne Arlaud, Ivan Laptev, Cordelia Schmid
Abstract: In this work, we aim to learn a unified vision-based policy for multi-fingered robot hands to manipulate a variety of objects in diverse poses. Though prior work has shown benefits of using human videos for policy learning, performance gains have been limited by the noise in estimated trajectories. Moreover, reliance on privileged object information such as ground-truth object states further limits the applicability in realistic scenarios. To address these limitations, we propose a new framework ViViDex to improve vision-based policy learning from human videos. It first uses reinforcement learning with trajectory guided rewards to train state-based policies for each video, obtaining both visually natural and physically plausible trajectories from the video. We then rollout successful episodes from state-based policies and train a unified visual policy without using any privileged information. We propose coordinate transformation to further enhance the visual point cloud representation, and compare behavior cloning and diffusion policy for the visual policy training. Experiments both in simulation and on the real robot demonstrate that ViViDex outperforms state-of-the-art approaches on three dexterous manipulation tasks.
Authors: Boyang Sun, Yu Yao, Huangyuan Hao, Yumou Qiu, Kun Zhang
Abstract: Testing conditional independence has many applications, such as in Bayesian network learning and causal discovery. Different test methods have been proposed. However, existing methods generally can not work when only discretized observations are available. Specifically, consider $X_1$, $\tilde{X}_2$ and $X_3$ are observed variables, where $\tilde{X}_2$ is a discretization of latent variables $X_2$. Applying existing test methods to the observations of $X_1$, $\tilde{X}_2$ and $X_3$ can lead to a false conclusion about the underlying conditional independence of variables $X_1$, $X_2$ and $X_3$. Motivated by this, we propose a conditional independence test specifically designed to accommodate the presence of such discretization. To achieve this, we design the bridge equations to recover the parameter reflecting the statistical information of the underlying latent continuous variables. An appropriate test statistic and its asymptotic distribution under the null hypothesis of conditional independence have also been derived. Both theoretical results and empirical validation have been provided, demonstrating the effectiveness of our test methods.
Authors: Abhinav Lalwani, Tasha Kim, Lovish Chopra, Christopher Hahn, Zhijing Jin, Mrinmaya Sachan
Abstract: Translating natural language into formal language such as First-Order Logic (FOL) is a foundational challenge in NLP with wide-ranging applications in automated reasoning, misinformation tracking, and knowledge validation. In this paper, we introduce Natural Language to First-Order Logic (NL2FOL), a framework to autoformalize natural language to FOL step by step using Large Language Models (LLMs). Our approach addresses key challenges in this translation process, including the integration of implicit background knowledge. By leveraging structured representations generated by NL2FOL, we use Satisfiability Modulo Theory (SMT) solvers to reason about the logical validity of natural language statements. We present logical fallacy detection as a case study to evaluate the efficacy of NL2FOL. Being neurosymbolic, our approach also provides interpretable insights into the reasoning process and demonstrates robustness without requiring model fine-tuning or labeled training data. Our framework achieves strong performance on multiple datasets. On the LOGIC dataset, NL2FOL achieves an F1-score of 78%, while generalizing effectively to the LOGICCLIMATE dataset with an F1-score of 80%.
Authors: Petar Bevanda, Bas Driessen, Lucian Cristian Iacob, Stefan Sosnowski Roland T\'oth, Sandra Hirche
Abstract: This paper presents a novel Koopman (composition) operator representation framework for control systems in reproducing kernel Hilbert spaces (RKHSs) that is free of explicit dictionary or input parametrizations. By establishing fundamental equivalences between different model representations, we are able to close the gap of control system operator learning and infinite-dimensional regression, enabling various empirical estimators and the connection to well-understood learning theory in RKHSs under one unified framework. As a consequence, our proposed framework allows for arbitrary accurate finite-rank approximations in infinite-dimensional spaces and leads to finite-dimensional predictors without apriori restrictions to a finite span of functions or inputs. To enable applications to high-dimensional control systems, we improve the scalability of our proposed control Koopman operator estimates by utilizing sketching techniques. Numerical experiments demonstrate superior prediction accuracy compared to bilinear EDMD, especially in high dimensions. Finally, we show that our learned models are readily interfaced with linear-parameter-varying techniques for model predictive control.
Authors: Sihan Zeng, Thinh T. Doan
Abstract: Two-time-scale optimization is a framework introduced in Zeng et al. (2024) that abstracts a range of policy evaluation and policy optimization problems in reinforcement learning (RL). Akin to bi-level optimization under a particular type of stochastic oracle, the two-time-scale optimization framework has an upper level objective whose gradient evaluation depends on the solution of a lower level problem, which is to find the root of a strongly monotone operator. In this work, we propose a new method for solving two-time-scale optimization that achieves significantly faster convergence than the prior arts. The key idea of our approach is to leverage an averaging step to improve the estimates of the operators in both lower and upper levels before using them to update the decision variables. These additional averaging steps eliminate the direct coupling between the main variables, enabling the accelerated performance of our algorithm. We characterize the finite-time convergence rates of the proposed algorithm under various conditions of the underlying objective function, including strong convexity, Polyak-Lojasiewicz condition, and general non-convexity. These rates significantly improve over the best-known complexity of the standard two-time-scale stochastic approximation algorithm. When applied to RL, we show how the proposed algorithm specializes to novel online sample-based methods that surpass or match the performance of the existing state of the art. Finally, we support our theoretical results with numerical simulations in RL.
Authors: Nadav Timor, Jonathan Mamou, Daniel Korat, Moshe Berchansky, Oren Pereg, Moshe Wasserblat, Tomer Galanti, Michal Gordon, David Harel
Abstract: This paper introduces distributed speculative inference (DSI), a novel inference algorithm that is provably faster than speculative inference (SI) [leviathan2023, chen2023, miao2024, sun2025, timor2025] and standard autoregressive inference (non-SI). Like other SI algorithms, DSI operates on frozen language models (LMs), requiring no training or architectural modifications, and it preserves the target distribution. Prior studies on SI have demonstrated empirical speedups over non-SI--but rely on sufficiently fast and accurate drafters, which are often unavailable in practice. We identify a gap where SI can be slower than non-SI if drafters are too slow or inaccurate. We close this gap by proving that DSI is faster than both SI and non-SI--given any drafters. DSI is therefore not only faster than SI, but also unlocks the acceleration of LMs for which SI fails. DSI leverages speculation parallelism (SP), a novel type of task parallelism, to orchestrate target and drafter instances that overlap in time, establishing a new foundational tradeoff between computational resources and latency. Our simulations show that DSI is 1.29-1.92x faster than SI in single-node setups for various off-the-shelf LMs and tasks. We open-source all our code.
Authors: Angelo Porrello, Lorenzo Bonicelli, Pietro Buzzega, Monica Millunzi, Simone Calderara, Rita Cucchiara
Abstract: The fine-tuning of deep pre-trained models has revealed compositional properties, with multiple specialized modules that can be arbitrarily composed into a single, multi-task model. However, identifying the conditions that promote compositionality remains an open issue, with recent efforts concentrating mainly on linearized networks. We conduct a theoretical study that attempts to demystify compositionality in standard non-linear networks through the second-order Taylor approximation of the loss function. The proposed formulation highlights the importance of staying within the pre-training basin to achieve composable modules. Moreover, it provides the basis for two dual incremental training algorithms: the one from the perspective of multiple models trained individually, while the other aims to optimize the composed model as a whole. We probe their application in incremental classification tasks and highlight some valuable skills. In fact, the pool of incrementally learned modules not only supports the creation of an effective multi-task model but also enables unlearning and specialization in certain tasks. Code available at https://github.com/aimagelab/mammoth.
Authors: Mingyang Jiang, Yueyuan Li, Songan Zhang, Siyuan Chen, Chunxiang Wang, Ming Yang
Abstract: Automated parking stands as a highly anticipated application of autonomous driving technology. However, existing path planning methodologies fall short of addressing this need due to their incapability to handle the diverse and complex parking scenarios in reality. While non-learning methods provide reliable planning results, they are vulnerable to intricate occasions, whereas learning-based ones are good at exploration but unstable in converging to feasible solutions. To leverage the strengths of both approaches, we introduce Hybrid pOlicy Path plannEr (HOPE). This novel solution integrates a reinforcement learning agent with Reeds-Shepp curves, enabling effective planning across diverse scenarios. HOPE guides the exploration of the reinforcement learning agent by applying an action mask mechanism and employs a transformer to integrate the perceived environmental information with the mask. To facilitate the training and evaluation of the proposed planner, we propose a criterion for categorizing the difficulty level of parking scenarios based on space and obstacle distribution. Experimental results demonstrate that our approach outperforms typical rule-based algorithms and traditional reinforcement learning methods, showing higher planning success rates and generalization across various scenarios. We also conduct real-world experiments to verify the practicability of HOPE. The code for our solution is openly available on https://github.com/jiamiya/HOPE.
Authors: Suraj Anand, Michael A. Lepori, Jack Merullo, Ellie Pavlick
Abstract: Language models have the ability to perform in-context learning (ICL), allowing them to flexibly adapt their behavior based on context. This contrasts with in-weights learning (IWL), where memorized information is encoded in model parameters after iterated observations of data. An ideal model should be able to flexibly deploy both of these abilities. Despite their apparent ability to learn in-context, language models are known to struggle when faced with unseen or rarely seen tokens (Land & Bartolo, 2024). Hence, we study $\textbf{structural in-context learning}$, which we define as the ability of a model to execute in-context learning on arbitrary novel tokens -- so called because the model must generalize on the basis of e.g. sentence structure or task structure, rather than content encoded in token embeddings. We study structural in-context algorithms on both synthetic and naturalistic tasks using toy models, masked language models, and autoregressive language models. We find that structural ICL appears before quickly disappearing early in LM pretraining. While it has been shown that ICL can diminish during training (Singh et al., 2023), we find that prior work does not account for structural ICL. Building on Chen et al. (2024) 's active forgetting method, we introduce pretraining and finetuning methods that can modulate the preference for structural ICL and IWL. Importantly, this allows us to induce a $\textit{dual process strategy}$ where in-context and in-weights solutions coexist within a single model.
Authors: Harin Lee, Taehyun Hwang, Min-hwan Oh
Abstract: We consider a stochastic sparse linear bandit problem where only a sparse subset of context features affects the expected reward function, i.e., the unknown reward parameter has a sparse structure. In the existing Lasso bandit literature, the compatibility conditions, together with additional diversity conditions on the context features are imposed to achieve regret bounds that only depend logarithmically on the ambient dimension $d$. In this paper, we demonstrate that even without the additional diversity assumptions, the \textit{compatibility condition on the optimal arm} is sufficient to derive a regret bound that depends logarithmically on $d$, and our assumption is strictly weaker than those used in the lasso bandit literature under the single-parameter setting. We propose an algorithm that adapts the forced-sampling technique and prove that the proposed algorithm achieves $O(\text{poly}\log dT)$ regret under the margin condition. To our knowledge, the proposed algorithm requires the weakest assumptions among Lasso bandit algorithms under the single-parameter setting that achieve $O(\text{poly}\log dT)$ regret. Through numerical experiments, we confirm the superior performance of our proposed algorithm.
Authors: Satoshi Yamamori, Jun Morimoto
Abstract: In this study, we propose the use of the phase-amplitude reduction method to construct an imitation learning framework. Imitating human movement trajectories is recognized as a promising strategy for generating a range of human-like robot movements. Unlike previous dynamical system-based imitation learning approaches, our proposed method allows the robot not only to imitate a limit cycle trajectory but also to replicate the transient movement from the initial or disturbed state to the limit cycle. Consequently, our method offers a safer imitation learning approach that avoids generating unpredictable motions immediately after disturbances or from a specified initial state. We first validated our proposed method by reconstructing a simple limit-cycle attractor. We then compared the proposed approach with a conventional method on a lemniscate trajectory tracking task with a simulated robot arm. Our findings confirm that our proposed method can more accurately generate transient movements to converge on a target periodic attractor compared to the previous standard approach. Subsequently, we applied our method to a real robot arm to imitate periodic human movements.
Authors: Manuel Cherep, Nikhil Singh
Abstract: Learning robust audio representations currently demands extensive datasets of real-world sound recordings. By applying artificial transformations to these recordings, models can learn to recognize similarities despite subtle variations through techniques like contrastive learning. However, these transformations are only approximations of the true diversity found in real-world sounds, which are generated by complex interactions of physical processes, from vocal cord vibrations to the resonance of musical instruments. We propose a solution to both the data scale and transformation limitations, leveraging synthetic audio. By randomly perturbing the parameters of a sound synthesizer, we generate audio doppelg\"angers-synthetic positive pairs with causally manipulated variations in timbre, pitch, and temporal envelopes. These variations, difficult to achieve through augmentations of existing audio, provide a rich source of contrastive information. Despite the shift to randomly generated synthetic data, our method produces strong representations, outperforming real data on several standard audio classification tasks. Notably, our approach is lightweight, requires no data storage, and has only a single hyperparameter, which we extensively analyze. We offer this method as a complement to existing strategies for contrastive learning in audio, using synthesized sounds to reduce the data burden on practitioners.
Authors: Seungeun Rho, Laura Smith, Tianyu Li, Sergey Levine, Xue Bin Peng, Sehoon Ha
Abstract: Skill discovery methods enable agents to learn diverse emergent behaviors without explicit rewards. To make learned skills useful for unknown downstream tasks, obtaining a semantically diverse repertoire of skills is essential. While some approaches introduce a discriminator to distinguish skills and others aim to increase state coverage, no existing work directly addresses the "semantic diversity" of skills. We hypothesize that leveraging the semantic knowledge of large language models (LLMs) can lead us to improve semantic diversity of resulting behaviors. In this sense, we introduce Language Guided Skill Discovery (LGSD), a skill discovery framework that aims to directly maximize the semantic diversity between skills. LGSD takes user prompts as input and outputs a set of semantically distinctive skills. The prompts serve as a means to constrain the search space into a semantically desired subspace, and the generated LLM outputs guide the agent to visit semantically diverse states within the subspace. We demonstrate that LGSD enables legged robots to visit different user-intended areas on a plane by simply changing the prompt. Furthermore, we show that language guidance aids in discovering more diverse skills compared to five existing skill discovery methods in robot-arm manipulation environments. Lastly, LGSD provides a simple way of utilizing learned skills via natural language.
Authors: Yulong Yang, Felix O'Mahony, Christine Allen-Blanchette
Abstract: In this paper, we introduce group convolutional neural networks (GCNNs) equivariant to color variation. GCNNs have been designed for a variety of geometric transformations from 2D and 3D rotation groups, to semi-groups such as scale. Despite the improved interpretability, accuracy and generalizability of these architectures, GCNNs have seen limited application in the context of perceptual quantities. Notably, the recent CEConv network uses a GCNN to achieve equivariance to hue transformations by convolving input images with a hue rotated RGB filter. However, this approach leads to invalid RGB values which break equivariance and degrade performance. We resolve these issues with a lifting layer that transforms the input image directly, thereby circumventing the issue of invalid RGB values and improving equivariance error by over three orders of magnitude. Moreover, we extend the notion of color equivariance to include equivariance to saturation and luminance shift. Our hue-, saturation-, luminance- and color-equivariant networks achieve strong generalization to out-of-distribution perceptual variations and improved sample efficiency over conventional architectures. We demonstrate the utility of our approach on synthetic and real world datasets where we consistently outperform competitive baselines.
Authors: Joseph Spracklen, Raveen Wijewickrama, A H M Nazmus Sakib, Anindya Maiti, Bimal Viswanath, Murtuza Jadliwala
Abstract: The reliance of popular programming languages such as Python and JavaScript on centralized package repositories and open-source software, combined with the emergence of code-generating Large Language Models (LLMs), has created a new type of threat to the software supply chain: package hallucinations. These hallucinations, which arise from fact-conflicting errors when generating code using LLMs, represent a novel form of package confusion attack that poses a critical threat to the integrity of the software supply chain. This paper conducts a rigorous and comprehensive evaluation of package hallucinations across different programming languages, settings, and parameters, exploring how a diverse set of models and configurations affect the likelihood of generating erroneous package recommendations and identifying the root causes of this phenomenon. Using 16 popular LLMs for code generation and two unique prompt datasets, we generate 576,000 code samples in two programming languages that we analyze for package hallucinations. Our findings reveal that that the average percentage of hallucinated packages is at least 5.2% for commercial models and 21.7% for open-source models, including a staggering 205,474 unique examples of hallucinated package names, further underscoring the severity and pervasiveness of this threat. To overcome this problem, we implement several hallucination mitigation strategies and show that they are able to significantly reduce the number of package hallucinations while maintaining code quality. Our experiments and findings highlight package hallucinations as a persistent and systemic phenomenon while using state-of-the-art LLMs for code generation, and a significant challenge which deserves the research community's urgent attention.
Authors: Zhepei Wei, Wei-Lin Chen, Yu Meng
Abstract: Retrieval-augmented generation (RAG) has shown promising potential to enhance the accuracy and factuality of language models (LMs). However, imperfect retrievers or noisy corpora can introduce misleading or even erroneous information to the retrieved contents, posing a significant challenge to the generation quality. Existing RAG methods typically address this challenge by directly predicting final answers despite potentially noisy inputs, resulting in an implicit denoising process that is difficult to interpret and verify. On the other hand, the acquisition of explicit denoising supervision is often costly, involving significant human efforts. In this work, we propose InstructRAG, where LMs explicitly learn the denoising process through self-synthesized rationales -- First, we instruct the LM to explain how the ground-truth answer is derived from retrieved documents. Then, these rationales can be used either as demonstrations for in-context learning of explicit denoising or as supervised fine-tuning data to train the model. Compared to standard RAG approaches, InstructRAG requires no additional supervision, allows for easier verification of the predicted answers, and effectively improves generation accuracy. Experiments show InstructRAG consistently outperforms existing RAG methods in both training-free and trainable scenarios, achieving a relative improvement of 8.3% over the best baseline method on average across five knowledge-intensive benchmarks. Extensive analysis indicates that InstructRAG scales well with increased numbers of retrieved documents and consistently exhibits robust denoising ability even in out-of-domain datasets, demonstrating strong generalizability.
Authors: Aliyah R. Hsu, Georgia Zhou, Yeshwanth Cherapanamjeri, Yaxuan Huang, Anobel Y. Odisho, Peter R. Carroll, Bin Yu
Abstract: Automated mechanistic interpretation research has attracted great interest due to its potential to scale explanations of neural network internals to large models. Existing automated circuit discovery work relies on activation patching or its approximations to identify subgraphs in models for specific tasks (circuits). They often suffer from slow runtime, approximation errors, and specific requirements of metrics, such as non-zero gradients. In this work, we introduce contextual decomposition for transformers (CD-T) to build interpretable circuits in large language models. CD-T can produce circuits of arbitrary level of abstraction, and is the first able to produce circuits as fine-grained as attention heads at specific sequence positions efficiently. CD-T consists of a set of mathematical equations to isolate contribution of model features. Through recursively computing contribution of all nodes in a computational graph of a model using CD-T followed by pruning, we are able to reduce circuit discovery runtime from hours to seconds compared to state-of-the-art baselines. On three standard circuit evaluation datasets (indirect object identification, greater-than comparisons, and docstring completion), we demonstrate that CD-T outperforms ACDC and EAP by better recovering the manual circuits with an average of 97% ROC AUC under low runtimes. In addition, we provide evidence that faithfulness of CD-T circuits is not due to random chance by showing our circuits are 80% more faithful than random circuits of up to 60% of the original model size. Finally, we show CD-T circuits are able to perfectly replicate original models' behavior (faithfulness $ = 1$) using fewer nodes than the baselines for all tasks. Our results underscore the great promise of CD-T for efficient automated mechanistic interpretability, paving the way for new insights into the workings of large language models.
Authors: Trevor Ablett, Bryan Chan, Jayce Haoran Wang, Jonathan Kelly
Abstract: Common approaches to providing feedback in reinforcement learning are the use of hand-crafted rewards or full-trajectory expert demonstrations. Alternatively, one can use examples of completed tasks, but such an approach can be extremely sample inefficient. We introduce value-penalized auxiliary control from examples (VPACE), an algorithm that significantly improves exploration in example-based control by adding examples of simple auxiliary tasks and an above-success-level value penalty. Across both simulated and real robotic environments, we show that our approach substantially improves learning efficiency for challenging tasks, while maintaining bounded value estimates. Preliminary results also suggest that VPACE may learn more efficiently than the more common approaches of using full trajectories or true sparse rewards. Project site: https://papers.starslab.ca/vpace/ .
Authors: Kotaro Ikeda, Tomoya Uda, Daisuke Okanohara, Sosuke Ito
Abstract: We discuss a connection between a generative model, called the diffusion model, and nonequilibrium thermodynamics for the Fokker-Planck equation, called stochastic thermodynamics. Based on the techniques of stochastic thermodynamics, we derive the speed-accuracy relations for the diffusion models, which are inequalities that relate the accuracy of data generation to the entropy production rate, which can be interpreted as the speed of the diffusion dynamics in the absence of the non-conservative force. From a stochastic thermodynamic perspective, our results provide a quantitative insight into how best to generate data in diffusion models. The optimal learning protocol is introduced by the geodesic of space of the 2-Wasserstein distance in optimal transport theory. We numerically illustrate the validity of the speed-accuracy relations for the diffusion models with different noise schedules and the different data. We numerically discuss our results for the optimal and suboptimal learning protocols. We also show the inaccurate data generation due to the non-conservative force, and the applicability of our results to data generation from the real-world image datasets.
Authors: Chenyang Cao, Yucheng Xin, Silang Wu, Longxiang He, Zichen Yan, Junbo Tan, Xueqian Wang
Abstract: Offline Safe Reinforcement Learning (RL) seeks to address safety constraints by learning from static datasets and restricting exploration. However, these approaches heavily rely on the dataset and struggle to generalize to unseen scenarios safely. In this paper, we aim to improve safety during the deployment of vision-based robotic tasks through online fine-tuning an offline pretrained policy. To facilitate effective fine-tuning, we introduce model-based RL, which is known for its data efficiency. Specifically, our method employs in-sample optimization to improve offline training efficiency while incorporating reachability guidance to ensure safety. After obtaining an offline safe policy, a safe policy expansion approach is leveraged for online fine-tuning. The performance of our method is validated on simulation benchmarks with five vision-only tasks and through real-world robot deployment using limited data. It demonstrates that our approach significantly improves the generalization of offline policies to unseen safety-constrained scenarios. To the best of our knowledge, this is the first work to explore offline-to-online RL for safe generalization tasks.
Authors: Afra Amini, Tim Vieira, Elliott Ash, Ryan Cotterell
Abstract: Best-of-N (BoN) is a popular and effective algorithm for aligning language models to human preferences. The algorithm works as follows: at inference time, N samples are drawn from the language model, and the sample with the highest reward, as judged by a reward model, is returned as the output. Despite its effectiveness, BoN is computationally expensive; it reduces sampling throughput by a factor of N. To make BoN more efficient at inference time, one strategy is to fine-tune the language model to mimic what BoN does during inference. To achieve this, we derive the distribution induced by the BoN algorithm. We then propose to fine-tune the language model to minimize backward KL divergence to the BoN distribution. Our approach is analogous to mean-field variational inference and, thus, we term it variational BoN (vBoN). To the extent this fine-tuning is successful and we end up with a good approximation, we have reduced the inference cost by a factor of N. Our experiments on controlled generation and summarization tasks show that BoN is the most effective alignment method, and our variational approximation to BoN achieves the closest performance to BoN and surpasses models fine-tuned using the standard KL-constrained RL objective. In the controlled generation task, vBoN appears more frequently on the Pareto frontier of reward and KL divergence compared to other alignment methods. In the summarization task, vBoN achieves high reward values across various sampling temperatures.
Authors: Alessandro Palma, Till Richter, Hanyi Zhang, Manuel Lubetzki, Alexander Tong, Andrea Dittadi, Fabian Theis
Abstract: Generative modeling of single-cell RNA-seq data is crucial for tasks like trajectory inference, batch effect removal, and simulation of realistic cellular data. However, recent deep generative models simulating synthetic single cells from noise operate on pre-processed continuous gene expression approximations, overlooking the discrete nature of single-cell data, which limits their effectiveness and hinders the incorporation of robust noise models. Additionally, aspects like controllable multi-modal and multi-label generation of cellular data remain underexplored. This work introduces CellFlow for Generation (CFGen), a flow-based conditional generative model that preserves the inherent discreteness of single-cell data. CFGen generates whole-genome multi-modal single-cell data reliably, improving the recovery of crucial biological data characteristics while tackling relevant generative tasks such as rare cell type augmentation and batch correction. We also introduce a novel framework for compositional data generation using Flow Matching. By showcasing CFGen on a diverse set of biological datasets and settings, we provide evidence of its value to the fields of computational biology and deep generative models.
Authors: James Giroux, Ariyarathne Gangani, Alexander C. Nwala, Cristiano Fanelli
Abstract: Social bots remain a major vector for spreading disinformation on social media and a menace to the public. Despite the progress made in developing multiple sophisticated social bot detection algorithms and tools, bot detection remains a challenging, unsolved problem that is fraught with uncertainty due to the heterogeneity of bot behaviors, training data, and detection algorithms. Detection models often disagree on whether to label the same account as bot or human-controlled. However, they do not provide any measure of uncertainty to indicate how much we should trust their results. We propose to address both bot detection and the quantification of uncertainty at the account level - a novel feature of this research. This dual focus is crucial as it allows us to leverage additional information related to the quantified uncertainty of each prediction, thereby enhancing decision-making and improving the reliability of bot classifications. Specifically, our approach facilitates targeted interventions for bots when predictions are made with high confidence and suggests caution (e.g., gathering more data) when predictions are uncertain.
Authors: Xinyi Wang, Antonis Antoniades, Yanai Elazar, Alfonso Amayuelas, Alon Albalak, Kexun Zhang, William Yang Wang
Abstract: The impressive capabilities of large language models (LLMs) have sparked debate over whether these models genuinely generalize to unseen tasks or predominantly rely on memorizing vast amounts of pretraining data. To explore this issue, we introduce an extended concept of memorization, distributional memorization, which measures the correlation between the LLM output probabilities and the pretraining data frequency. To effectively capture task-specific pretraining data frequency, we propose a novel task-gram language model, which is built by counting the co-occurrence of semantically related $n$-gram pairs from task inputs and outputs in the pretraining corpus. Using the Pythia models trained on the Pile dataset, we evaluate four distinct tasks: machine translation, factual question answering, world knowledge understanding, and math reasoning. Our findings reveal varying levels of memorization, with the strongest effect observed in factual question answering. Furthermore, while model performance improves across all tasks as LLM size increases, only factual question answering shows an increase in memorization, whereas machine translation and reasoning tasks exhibit greater generalization, producing more novel outputs. This study demonstrates that memorization plays a larger role in simpler, knowledge-intensive tasks, while generalization is the key for harder, reasoning-based tasks, providing a scalable method for analyzing large pretraining corpora in greater depth.
Authors: Amir Mohammad Karimi Mamaghan, Samuele Papa, Karl Henrik Johansson, Stefan Bauer, Andrea Dittadi
Abstract: Object-centric (OC) representations, which model visual scenes as compositions of discrete objects, have the potential to be used in various downstream tasks to achieve systematic compositional generalization and facilitate reasoning. However, these claims have yet to be thoroughly validated empirically. Recently, foundation models have demonstrated unparalleled capabilities across diverse domains, from language to computer vision, positioning them as a potential cornerstone of future research for a wide range of computational tasks. In this paper, we conduct an extensive empirical study on representation learning for downstream Visual Question Answering (VQA), which requires an accurate compositional understanding of the scene. We thoroughly investigate the benefits and trade-offs of OC models and alternative approaches including large pre-trained foundation models on both synthetic and real-world data, ultimately identifying a promising path to leverage the strengths of both paradigms. The extensiveness of our study, encompassing over 600 downstream VQA models and 15 different types of upstream representations, also provides several additional insights that we believe will be of interest to the community at large.
Authors: Danfeng Guo, Demetri Terzopoulos
Abstract: Large Vision-Language Models (LVLMs) have achieved significant success in recent years, and they have been extended to the medical domain. Although demonstrating satisfactory performance on medical Visual Question Answering (VQA) tasks, Medical LVLMs (MLVLMs) suffer from the hallucination problem, which makes them fail to diagnose complex pathologies. Moreover, they readily fail to learn minority pathologies due to imbalanced training data. We propose two prompting strategies for MLVLMs that reduce hallucination and improve VQA performance. In the first strategy, we provide a detailed explanation of the queried pathology. In the second strategy, we fine-tune a cheap, weak learner to achieve high performance on a specific metric, and textually provide its judgment to the MLVLM. Tested on the MIMIC-CXR-JPG and Chexpert datasets, our methods significantly improve the diagnostic F1 score, with the highest increase being 0.27. We also demonstrate that our prompting strategies can be extended to general LVLM domains. Based on POPE metrics, it effectively suppresses the false negative predictions of existing LVLMs and improves Recall by approximately 0.07.
Authors: Jung In Park, Mahyar Abbasian, Iman Azimi, Dawn T. Bounds, Angela Jun, Jaesu Han, Robert M. McCarron, Jessica Borelli, Parmida Safavi, Sanaz Mirbaha, Jia Li, Mona Mahmoudi, Carmen Wiedenhoeft, Amir M. Rahmani
Abstract: Objective: This study aims to develop and validate an evaluation framework to ensure the safety and reliability of mental health chatbots, which are increasingly popular due to their accessibility, human-like interactions, and context-aware support. Materials and Methods: We created an evaluation framework with 100 benchmark questions and ideal responses, and five guideline questions for chatbot responses. This framework, validated by mental health experts, was tested on a GPT-3.5-turbo-based chatbot. Automated evaluation methods explored included large language model (LLM)-based scoring, an agentic approach using real-time data, and embedding models to compare chatbot responses against ground truth standards. Results: The results highlight the importance of guidelines and ground truth for improving LLM evaluation accuracy. The agentic method, dynamically accessing reliable information, demonstrated the best alignment with human assessments. Adherence to a standardized, expert-validated framework significantly enhanced chatbot response safety and reliability. Discussion: Our findings emphasize the need for comprehensive, expert-tailored safety evaluation metrics for mental health chatbots. While LLMs have significant potential, careful implementation is necessary to mitigate risks. The superior performance of the agentic approach underscores the importance of real-time data access in enhancing chatbot reliability. Conclusion: The study validated an evaluation framework for mental health chatbots, proving its effectiveness in improving safety and reliability. Future work should extend evaluations to accuracy, bias, empathy, and privacy to ensure holistic assessment and responsible integration into healthcare. Standardized evaluations will build trust among users and professionals, facilitating broader adoption and improved mental health support through technology.
Authors: Hee-Geon Kang, Sunggon Kim
Abstract: The problem of finding the expected value of a statistic of a locally stable point process in a bounded region is addressed. We propose an adaptive importance sampling for solving the problem. In our proposal, we restrict the importance point process to the family of homogeneous Poisson point processes, which enables us to generate quickly independent samples of the importance point process. The optimal intensity of the importance point process is found by applying the cross-entropy minimization method. In the proposed scheme, the expected value of the function and the optimal intensity are iteratively estimated in an adaptive manner. We show that the proposed estimator converges to the target value almost surely, and prove the asymptotic normality of it. We explain how to apply the proposed scheme to the estimation of the intensity of a stationary pairwise interaction point process. The performance of the proposed scheme is compared numerically with the Markov chain Monte Carlo simulation and the perfect sampling.
Authors: Hossein Jafarinia, Alireza Alipanah, Danial Hamdi, Saeed Razavi, Nahal Mirzaie, Mohammad Hossein Rohban
Abstract: Whole Slide Image (WSI) classification with multiple instance learning (MIL) in digital pathology faces significant computational challenges. Current methods mostly rely on extensive self-supervised learning (SSL) for satisfactory performance, requiring long training periods and considerable computational resources. At the same time, no pre-training affects performance due to domain shifts from natural images to WSIs. We introduce Snuffy architecture, a novel MIL-pooling method based on sparse transformers that mitigates performance loss with limited pre-training and enables continual few-shot pre-training as a competitive option. Our sparsity pattern is tailored for pathology and is theoretically proven to be a universal approximator with the tightest probabilistic sharp bound on the number of layers for sparse transformers, to date. We demonstrate Snuffy's effectiveness on CAMELYON16 and TCGA Lung cancer datasets, achieving superior WSI and patch-level accuracies. The code is available on https://github.com/jafarinia/snuffy.
Authors: Muhammad Aqeel, Shakiba Sharifi, Marco Cristani, Francesco Setti
Abstract: This study introduces the Iterative Refinement Process (IRP), a robust anomaly detection methodology designed for high-stakes industrial quality control. The IRP enhances defect detection accuracy through a cyclic data refinement strategy, iteratively removing misleading data points to improve model performance and robustness. We validate the IRP's effectiveness using two benchmark datasets, Kolektor SDD2 (KSDD2) and MVTec AD, covering a wide range of industrial products and defect types. Our experimental results demonstrate that the IRP consistently outperforms traditional anomaly detection models, particularly in environments with high noise levels. This study highlights the IRP's potential to significantly enhance anomaly detection processes in industrial settings, effectively managing the challenges of sparse and noisy data.
Authors: Junwon Lee, Jaekwon Im, Dabin Kim, Juhan Nam
Abstract: Foley sound synthesis is crucial for multimedia production, enhancing user experience by synchronizing audio and video both temporally and semantically. Recent studies on automating this labor-intensive process through video-to-sound generation face significant challenges. Systems lacking explicit temporal features suffer from poor alignment and controllability, while timestamp-based models require costly and subjective human annotation. We propose Video-Foley, a video-to-sound system using Root Mean Square (RMS) as an intuitive condition with semantic timbre prompts (audio or text). RMS, a frame-level intensity envelope closely related to audio semantics, acts as a temporal event feature to guide audio generation from video. The annotation-free self-supervised learning framework consists of two stages, Video2RMS and RMS2Sound, incorporating novel ideas including RMS discretization and RMS-ControlNet with a pretrained text-to-audio model. Our extensive evaluation shows that Video-Foley achieves state-of-the-art performance in audio-visual alignment and controllability for sound timing, intensity, timbre, and nuance. Source code, model weights and demos are available on our companion website. (https://jnwnlee.github.io/video-foley-demo)
Authors: Min Shi, Fuxiao Liu, Shihao Wang, Shijia Liao, Subhashree Radhakrishnan, Yilin Zhao, De-An Huang, Hongxu Yin, Karan Sapra, Yaser Yacoob, Humphrey Shi, Bryan Catanzaro, Andrew Tao, Jan Kautz, Zhiding Yu, Guilin Liu
Abstract: The ability to accurately interpret complex visual information is a crucial topic of multimodal large language models (MLLMs). Recent work indicates that enhanced visual perception significantly reduces hallucinations and improves performance on resolution-sensitive tasks, such as optical character recognition and document analysis. A number of recent MLLMs achieve this goal using a mixture of vision encoders. Despite their success, there is a lack of systematic comparisons and detailed ablation studies addressing critical aspects, such as expert selection and the integration of multiple vision experts. This study provides an extensive exploration of the design space for MLLMs using a mixture of vision encoders and resolutions. Our findings reveal several underlying principles common to various existing strategies, leading to a streamlined yet effective design approach. We discover that simply concatenating visual tokens from a set of complementary vision encoders is as effective as more complex mixing architectures or strategies. We additionally introduce Pre-Alignment to bridge the gap between vision-focused encoders and language tokens, enhancing model coherence. The resulting family of MLLMs, Eagle, surpasses other leading open-source models on major MLLM benchmarks.
Authors: Benjamin L. Badger
Abstract: Attention mechanisms that confer selective focus on a strict subset of input elements are nearly ubiquitous in language models today. We posit there to be downside to the use of attention: most input information is lost. In support of this idea we observe poor input representation accuracy in transformers and more accurate representation in what we term masked mixers, which replace self-attention with masked convolutions. The masked mixer learns causal language modeling more efficiently than early transformer implementations and even outperforms optimized, current transformers when training on small (<512) but not larger context windows. Evidence is presented for the hypothesis that differences in transformer and masked mixer training efficiencies for various tasks are best predicted by input representation accuracy, or equivalently global invertibility. We hypothesize that the information loss exhibited by transformers would be more detrimental to retrieval than generation, as the former is more closely approximated by a bijective and thus invertible function. We find that masked mixers are more effective retrieval models both when the pretrained embedding model is unchanged as well as when the embedding model is modified via cosine similarity-based InfoNCE loss minimization. A small masked mixer is shown to outperform a large and near state-of-the-art transformer-based retrieval model, despite the latter being trained with many orders of magnitude more data and compute.
Authors: Niklas Muennighoff, Luca Soldaini, Dirk Groeneveld, Kyle Lo, Jacob Morrison, Sewon Min, Weijia Shi, Pete Walsh, Oyvind Tafjord, Nathan Lambert, Yuling Gu, Shane Arora, Akshita Bhagia, Dustin Schwenk, David Wadden, Alexander Wettig, Binyuan Hui, Tim Dettmers, Douwe Kiela, Ali Farhadi, Noah A. Smith, Pang Wei Koh, Amanpreet Singh, Hannaneh Hajishirzi
Abstract: We introduce OLMoE, a fully open, state-of-the-art language model leveraging sparse Mixture-of-Experts (MoE). OLMoE-1B-7B has 7 billion (B) parameters but uses only 1B per input token. We pretrain it on 5 trillion tokens and further adapt it to create OLMoE-1B-7B-Instruct. Our models outperform all available models with similar active parameters, even surpassing larger ones like Llama2-13B-Chat and DeepSeekMoE-16B. We present various experiments on MoE training, analyze routing in our model showing high specialization, and open-source all aspects of our work: model weights, training data, code, and logs.
Authors: Ziwei Yang, Rikuto Kotoge, Xihao Piao, Zheng Chen, Lingwei Zhu, Peng Gao, Yasuko Matsubara, Yasushi Sakurai, Jimeng Sun
Abstract: Framing the investigation of diverse cancers as a machine learning problem has recently shown significant potential in multi-omics analysis and cancer research. Empowering these successful machine learning models are the high-quality training datasets with sufficient data volume and adequate preprocessing. However, while there exist several public data portals including The Cancer Genome Atlas (TCGA) multi-omics initiative or open-bases such as the LinkedOmics, these databases are not off-the-shelf for existing machine learning models. In this paper we propose MLOmics, an open cancer multi-omics benchmark aiming at serving better the development and evaluation of bioinformatics and machine learning models. MLOmics contains 8,314 patient samples covering all 32 cancer types with four omics types, stratified features, and extensive baselines. Complementary support for downstream analysis and bio-knowledge linking are also included to support interdisciplinary analysis.
Authors: Michael Sun, Alston Lo, Minghao Guo, Jie Chen, Connor Coley, Wojciech Matusik
Abstract: Designing synthetically accessible molecules and recommending analogs to unsynthesizable molecules are important problems for accelerating molecular discovery. We reconceptualize both problems using ideas from program synthesis. Drawing inspiration from syntax-guided synthesis approaches, we decouple the syntactic skeleton from the semantics of a synthetic tree to create a bilevel framework for reasoning about the combinatorial space of synthesis pathways. Given a molecule we aim to generate analogs for, we iteratively refine its skeletal characteristics via Markov Chain Monte Carlo simulations over the space of syntactic skeletons. Given a black-box oracle to optimize, we formulate a joint design space over syntactic templates and molecular descriptors and introduce evolutionary algorithms that optimize both syntactic and semantic dimensions synergistically. Our key insight is that once the syntactic skeleton is set, we can amortize over the search complexity of deriving the program's semantics by training policies to fully utilize the fixed horizon Markov Decision Process imposed by the syntactic template. We demonstrate performance advantages of our bilevel framework for synthesizable analog generation and synthesizable molecule design. Notably, our approach offers the user explicit control over the resources required to perform synthesis and biases the design space towards simpler solutions, making it particularly promising for autonomous synthesis platforms. Code is at https://github.com/shiningsunnyday/SynthesisNet.
Authors: Daniel M. Steinberg, Rafael Oliveira, Cheng Soon Ong, Edwin V. Bonilla
Abstract: We develop VSD, a method for conditioning a generative model of discrete, combinatorial designs on a rare desired class by efficiently evaluating a black-box (e.g. experiment, simulation) in a batch sequential manner. We call this task active generation; we formalize active generation's requirements and desiderata, and formulate a solution via variational inference. VSD uses off-the-shelf gradient based optimization routines, can learn powerful generative models for desirable designs, and can take advantage of scalable predictive models. We derive asymptotic convergence rates for learning the true conditional generative distribution of designs with certain configurations of our method. After illustrating the generative model on images, we empirically demonstrate that VSD can outperform existing baseline methods on a set of real sequence-design problems in various protein and DNA/RNA engineering tasks.
Authors: Boming Miao, Chunxiao Li, Yao Zhu, Weixiang Sun, Zizhe Wang, Xiaoyi Wang, Chuanlong Xie
Abstract: With the rapid development of deep learning, object detectors have demonstrated impressive performance; however, vulnerabilities still exist in certain scenarios. Current research exploring the vulnerabilities using adversarial patches often struggles to balance the trade-off between attack effectiveness and visual quality. To address this problem, we propose a novel framework of patch attack from semantic perspective, which we refer to as AdvLogo. Based on the hypothesis that every semantic space contains an adversarial subspace where images can cause detectors to fail in recognizing objects, we leverage the semantic understanding of the diffusion denoising process and drive the process to adversarial subareas by perturbing the latent and unconditional embeddings at the last timestep. To mitigate the distribution shift that exposes a negative impact on image quality, we apply perturbation to the latent in frequency domain with the Fourier Transform. Experimental results demonstrate that AdvLogo achieves strong attack performance while maintaining high visual quality.
Authors: Tianqi Chen, Shujian Zhang, Mingyuan Zhou
Abstract: The machine learning community is increasingly recognizing the importance of fostering trust and safety in modern generative AI (GenAI) models. We posit machine unlearning (MU) as a crucial foundation for developing safe, secure, and trustworthy GenAI models. Traditional MU methods often rely on stringent assumptions and require access to real data. This paper introduces Score Forgetting Distillation (SFD), an innovative MU approach that promotes the forgetting of undesirable information in diffusion models by aligning the conditional scores of "unsafe" classes or concepts with those of "safe" ones. To eliminate the need for real data, our SFD framework incorporates a score-based MU loss into the score distillation objective of a pretrained diffusion model. This serves as a regularization term that preserves desired generation capabilities while enabling the production of synthetic data through a one-step generator. Our experiments on pretrained label-conditional and text-to-image diffusion models demonstrate that our method effectively accelerates the forgetting of target classes or concepts during generation, while preserving the quality of other classes or concepts. This unlearned and distilled diffusion not only pioneers a novel concept in MU but also accelerates the generation speed of diffusion models. Our experiments and studies on a range of diffusion models and datasets confirm that our approach is generalizable, effective, and advantageous for MU in diffusion models. Code is available at https://github.com/tqch/score-forgetting-distillation. ($\textbf{Warning:}$ This paper contains sexually explicit imagery, discussions of pornography, racially-charged terminology, and other content that some readers may find disturbing, distressing, and/or offensive.)
URLs: https://github.com/tqch/score-forgetting-distillation.
Authors: Sagar Parekh, Lauren Bramblett, Nicola Bezzo, Dylan P. Losey
Abstract: Humans interacting with robots often form predictions of what the robot will do next. For instance, based on the recent behavior of an autonomous car, a nearby human driver might predict that the car is going to remain in the same lane. It is important for the robot to understand the human's prediction for safe and seamless interaction: e.g., if the autonomous car knows the human thinks it is not merging -- but the autonomous car actually intends to merge -- then the car can adjust its behavior to prevent an accident. Prior works typically assume that humans make precise predictions of robot behavior. However, recent research on human-human prediction suggests the opposite: humans tend to approximate other agents by predicting their high-level behaviors. We apply this finding to develop a second-order theory of mind approach that enables robots to estimate how humans predict they will behave. To extract these high-level predictions directly from data, we embed the recent human and robot trajectories into a discrete latent space. Each element of this latent space captures a different type of behavior (e.g., merging in front of the human, remaining in the same lane) and decodes into a vector field across the state space that is consistent with the underlying behavior type. We hypothesize that our resulting high-level and course predictions of robot behavior will correspond to actual human predictions. We provide initial evidence in support of this hypothesis through proof-of-concept simulations, testing our method's predictions against those of real users, and experiments on a real-world interactive driving dataset.
Authors: Ninad Khargonkar, Luis Felipe Casas, Balakrishnan Prabhakaran, Yu Xiang
Abstract: We introduce a novel grasp representation named the Unified Gripper Coordinate Space (UGCS) for grasp synthesis and grasp transfer. Our representation leverages spherical coordinates to create a shared coordinate space across different robot grippers, enabling it to synthesize and transfer grasps for both novel objects and previously unseen grippers. The strength of this representation lies in the ability to map palm and fingers of a gripper and the unified coordinate space. Grasp synthesis is formulated as predicting the unified spherical coordinates on object surface points via a conditional variational autoencoder. The predicted unified gripper coordinates establish exact correspondences between the gripper and object points, which is used to optimize grasp pose and joint values. Grasp transfer is facilitated through the point-to-point correspondence between any two (potentially unseen) grippers and solved via a similar optimization. Extensive simulation and real-world experiments showcase the efficacy of the unified grasp representation for grasp synthesis in generating stable and diverse grasps. Similarly, we showcase real-world grasp transfer from human demonstrations across different objects.
Authors: Maximilian Weiherer, Bernhard Egger
Abstract: We propose to use the family of Mat\'ern kernels for implicit surface reconstruction, building upon the recent success of kernel methods for 3D reconstruction of oriented point clouds. As we show from a theoretical and practical perspective, Mat\'ern kernels have some appealing properties which make them particularly well suited for surface reconstruction -- outperforming state-of-the-art methods based on the arc-cosine kernel while being significantly easier to implement, faster to compute, and scalable. Being stationary, we demonstrate that Mat\'ern kernels allow for tunable surface reconstruction in the same way as Fourier feature mappings help coordinate-based MLPs overcome spectral bias. Moreover, we theoretically analyze Mat\'ern kernels' connection to SIREN networks as well as their relation to previously employed arc-cosine kernels. Finally, based on recently introduced Neural Kernel Fields, we present data-dependent Mat\'ern kernels and conclude that especially the Laplace kernel (being part of the Mat\'ern family) is extremely competitive, performing almost on par with state-of-the-art methods in the noise-free case while having a more than five times shorter training time.
Authors: Hong Li, Nanxi Li, Yuanjie Chen, Jianbin Zhu, Qinlu Guo, Cewu Lu, Yong-Lu Li
Abstract: Multi-modal Large Language Models (MLLMs) have exhibited impressive capability. However, recently many deficiencies of MLLMs have been found compared to human intelligence, $\textit{e.g.}$, hallucination. To drive the MLLMs study, the community dedicated efforts to building larger benchmarks with complex tasks. In this paper, we propose benchmarking an essential but usually overlooked intelligence: $\textbf{association}$, a human's basic capability to link observation and prior practice memory. To comprehensively investigate MLLM's performance on the association, we formulate the association task and devise a standard benchmark based on adjective and verb semantic concepts. Instead of costly data annotation and curation, we propose a convenient $\textbf{annotation-free}$ construction method transforming the general dataset for our association tasks. Simultaneously, we devise a rigorous data refinement process to eliminate confusion in the raw dataset. Building on this database, we establish three levels of association tasks: single-step, synchronous, and asynchronous associations. Moreover, we conduct a comprehensive investigation into the MLLMs' zero-shot association capabilities, addressing multiple dimensions, including three distinct memory strategies, both open-source and closed-source MLLMs, cutting-edge Mixture-of-Experts (MoE) models, and the involvement of human experts. Our systematic investigation shows that current open-source MLLMs consistently exhibit poor capability in our association tasks, even the currently state-of-the-art GPT-4V(vision) also has a significant gap compared to humans. We believe our benchmark would pave the way for future MLLM studies. $\textit{Our data and code are available at:}$ https://mvig-rhos.com/llm_inception.
Authors: Genta Indra Winata, David Anugraha, Lucky Susanto, Garry Kuwanto, Derry Tanti Wijaya
Abstract: Understanding the quality of a performance evaluation metric is crucial for ensuring that model outputs align with human preferences. However, it remains unclear how well each metric captures the diverse aspects of these preferences, as metrics often excel in one particular area but not across all dimensions. To address this, it is essential to systematically calibrate metrics to specific aspects of human preference, catering to the unique characteristics of each aspect. We introduce MetaMetrics, a calibrated meta-metric designed to evaluate generation tasks across different modalities in a supervised manner. MetaMetrics optimizes the combination of existing metrics to enhance their alignment with human preferences. Our metric demonstrates flexibility and effectiveness in both language and vision downstream tasks, showing significant benefits across various multilingual and multi-domain scenarios. MetaMetrics aligns closely with human preferences and is highly extendable and easily integrable into any application. This makes MetaMetrics a powerful tool for improving the evaluation of generation tasks, ensuring that metrics are more representative of human judgment across diverse contexts.
Authors: S\'egol\`ene Martin, Anne Gagneux, Paul Hagemann, Gabriele Steidl
Abstract: In this paper, we introduce Plug-and-Play (PnP) Flow Matching, an algorithm for solving imaging inverse problems. PnP methods leverage the strength of pre-trained denoisers, often deep neural networks, by integrating them in optimization schemes. While they achieve state-of-the-art performance on various inverse problems in imaging, PnP approaches face inherent limitations on more generative tasks like inpainting. On the other hand, generative models such as Flow Matching pushed the boundary in image sampling yet lack a clear method for efficient use in image restoration. We propose to combine the PnP framework with Flow Matching (FM) by defining a time-dependent denoiser using a pre-trained FM model. Our algorithm alternates between gradient descent steps on the data-fidelity term, reprojections onto the learned FM path, and denoising. Notably, our method is computationally efficient and memory-friendly, as it avoids backpropagation through ODEs and trace computations. We evaluate its performance on denoising, super-resolution, deblurring, and inpainting tasks, demonstrating superior results compared to existing PnP algorithms and Flow Matching based state-of-the-art methods.
Authors: Yu Ying Chiu, Liwei Jiang, Yejin Choi
Abstract: As users increasingly seek guidance from LLMs for decision-making in daily life, many of these decisions are not clear-cut and depend significantly on the personal values and ethical standards of people. We present DailyDilemmas, a dataset of 1,360 moral dilemmas encountered in everyday life. Each dilemma presents two possible actions, along with affected parties and relevant human values for each action. Based on these dilemmas, we gather a repository of human values covering diverse everyday topics, such as interpersonal relationships, workplace, and environmental issues. With DailyDilemmas, we evaluate LLMs on these dilemmas to determine what action they will choose and the values represented by these action choices. Then, we analyze values through the lens of five theoretical frameworks inspired by sociology, psychology, and philosophy, including the World Values Survey, Moral Foundations Theory, Maslow's Hierarchy of Needs, Aristotle's Virtues, and Plutchik's Wheel of Emotions. For instance, we find LLMs are most aligned with self-expression over survival in World Values Survey and care over loyalty in Moral Foundations Theory. Interestingly, we find substantial preference differences in models for some core values. For example, for truthfulness, Mixtral-8x7B neglects it by 9.7% while GPT-4-turbo selects it by 9.4%. We also study the recent guidance released by OpenAI (ModelSpec), and Anthropic (Constitutional AI) to understand how their designated principles reflect their models' actual value prioritization when facing nuanced moral reasoning in daily-life settings. Finally, we find that end users cannot effectively steer such prioritization using system prompts.
Authors: Michael E. Sander, Gabriel Peyr\'e
Abstract: Causal Transformers are trained to predict the next token for a given context. While it is widely accepted that self-attention is crucial for encoding the causal structure of sequences, the precise underlying mechanism behind this in-context autoregressive learning ability remains unclear. In this paper, we take a step towards understanding this phenomenon by studying the approximation ability of Transformers for next-token prediction. Specifically, we explore the capacity of causal Transformers to predict the next token $x_{t+1}$ given an autoregressive sequence $(x_1, \dots, x_t)$ as a prompt, where $ x_{t+1} = f(x_t) $, and $ f $ is a context-dependent function that varies with each sequence. On the theoretical side, we focus on specific instances, namely when $ f $ is linear or when $ (x_t)_{t \geq 1} $ is periodic. We explicitly construct a Transformer (with linear, exponential, or softmax attention) that learns the mapping $f$ in-context through a causal kernel descent method. The causal kernel descent method we propose provably estimates $x_{t+1} $ based solely on past and current observations $ (x_1, \dots, x_t) $, with connections to the Kaczmarz algorithm in Hilbert spaces. We present experimental results that validate our theoretical findings and suggest their applicability to more general mappings $f$.
Authors: Marina Sheshukova, Denis Belomestny, Alain Durmus, Eric Moulines, Alexey Naumov, Sergey Samsonov
Abstract: We address the problem of solving strongly convex and smooth minimization problems using stochastic gradient descent (SGD) algorithm with a constant step size. Previous works suggested to combine the Polyak-Ruppert averaging procedure with the Richardson-Romberg extrapolation to reduce the asymptotic bias of SGD at the expense of a mild increase of the variance. We significantly extend previous results by providing an expansion of the mean-squared error of the resulting estimator with respect to the number of iterations $n$. We show that the root mean-squared error can be decomposed into the sum of two terms: a leading one of order $\mathcal{O}(n^{-1/2})$ with explicit dependence on a minimax-optimal asymptotic covariance matrix, and a second-order term of order $\mathcal{O}(n^{-3/4})$, where the power $3/4$ is best known. We also extend this result to the higher-order moment bounds. Our analysis relies on the properties of the SGD iterates viewed as a time-homogeneous Markov chain. In particular, we establish that this chain is geometrically ergodic with respect to a suitably defined weighted Wasserstein semimetric.
Authors: Serin Yang, Taesung Kwon, Jong Chul Ye
Abstract: Recent progress in large-scale text-to-video (T2V) and image-to-video (I2V) diffusion models has greatly enhanced video generation, especially in terms of keyframe interpolation. However, current image-to-video diffusion models, while powerful in generating videos from a single conditioning frame, need adaptation for two-frame (start & end) conditioned generation, which is essential for effective bounded interpolation. Unfortunately, existing approaches that fuse temporally forward and backward paths in parallel often suffer from off-manifold issues, leading to artifacts or requiring multiple iterative re-noising steps. In this work, we introduce a novel, bidirectional sampling strategy to address these off-manifold issues without requiring extensive re-noising or fine-tuning. Our method employs sequential sampling along both forward and backward paths, conditioned on the start and end frames, respectively, ensuring more coherent and on-manifold generation of intermediate frames. Additionally, we incorporate advanced guidance techniques, CFG++ and DDS, to further enhance the interpolation process. By integrating these, our method achieves state-of-the-art performance, efficiently generating high-quality, smooth videos between keyframes. On a single 3090 GPU, our method can interpolate 25 frames at 1024 x 576 resolution in just 195 seconds, establishing it as a leading solution for keyframe interpolation.
Authors: Will Dorrell, Kyle Hsu, Luke Hollingsworth, Jin Hwa Lee, Jiajun Wu, Chelsea Finn, Peter E Latham, Tim EJ Behrens, James CR Whittington
Abstract: Why do biological and artificial neurons sometimes modularise, each encoding a single meaningful variable, and sometimes entangle their representation of many variables? In this work, we develop a theory of when biologically inspired networks -- those that are nonnegative and energy efficient -- modularise their representation of source variables (sources). We derive necessary and sufficient conditions on a sample of sources that determine whether the neurons in an optimal biologically-inspired linear autoencoder modularise. Our theory applies to any dataset, extending far beyond the case of statistical independence studied in previous work. Rather we show that sources modularise if their support is ``sufficiently spread''. From this theory, we extract and validate predictions in a variety of empirical studies on how data distribution affects modularisation in nonlinear feedforward and recurrent neural networks trained on supervised and unsupervised tasks. Furthermore, we apply these ideas to neuroscience data, showing that range independence can be used to understand the mixing or modularising of spatial and reward information in entorhinal recordings in seemingly conflicting experiments. Further, we use these results to suggest alternate origins of mixed-selectivity, beyond the predominant theory of flexible nonlinear classification. In sum, our theory prescribes precise conditions on when neural activities modularise, providing tools for inducing and elucidating modular representations in brains and machines.
Authors: Putra Manggala, Atalanti Mastakouri, Elke Kirschbaum, Shiva Prasad Kasiviswanathan, Aaditya Ramdas
Abstract: To use generative question-and-answering (QA) systems for decision-making and in any critical application, these systems need to provide well-calibrated confidence scores that reflect the correctness of their answers. Existing calibration methods aim to ensure that the confidence score is, *on average*, indicative of the likelihood that the answer is correct. We argue, however, that this standard (average-case) notion of calibration is difficult to interpret for decision-making in generative QA. To address this, we generalize the standard notion of average calibration and introduce QA-calibration, which ensures calibration holds across different question-and-answer groups. We then propose discretized posthoc calibration schemes for achieving QA-calibration. We establish distribution-free guarantees on the performance of this method and validate our method on confidence scores returned by elicitation prompts across multiple QA benchmarks and large language models (LLMs).
Authors: Avik Pal, Max van Spengler, Guido Maria D'Amely di Melendugno, Alessandro Flaborea, Fabio Galasso, Pascal Mettes
Abstract: Image-text representation learning forms a cornerstone in vision-language models, where pairs of images and textual descriptions are contrastively aligned in a shared embedding space. Since visual and textual concepts are naturally hierarchical, recent work has shown that hyperbolic space can serve as a high-potential manifold to learn vision-language representation with strong downstream performance. In this work, for the first time we show how to fully leverage the innate hierarchical nature of hyperbolic embeddings by looking beyond individual image-text pairs. We propose Compositional Entailment Learning for hyperbolic vision-language models. The idea is that an image is not only described by a sentence but is itself a composition of multiple object boxes, each with their own textual description. Such information can be obtained freely by extracting nouns from sentences and using openly available localized grounding models. We show how to hierarchically organize images, image boxes, and their textual descriptions through contrastive and entailment-based objectives. Empirical evaluation on a hyperbolic vision-language model trained with millions of image-text pairs shows that the proposed compositional learning approach outperforms conventional Euclidean CLIP learning, as well as recent hyperbolic alternatives, with better zero-shot and retrieval generalization and clearly stronger hierarchical performance.
Authors: Zonglin Yang, Wanhao Liu, Ben Gao, Tong Xie, Yuqiang Li, Wanli Ouyang, Soujanya Poria, Erik Cambria, Dongzhan Zhou
Abstract: Scientific discovery contributes largely to human society's prosperity, and recent progress shows that LLMs could potentially catalyze this process. However, it is still unclear whether LLMs can discover novel and valid hypotheses in chemistry. In this work, we investigate this central research question: Can LLMs automatically discover novel and valid chemistry research hypotheses given only a chemistry research background (consisting of a research question and/or a background survey), without limitation on the domain of the research question? After extensive discussions with chemistry experts, we propose an assumption that a majority of chemistry hypotheses can be resulted from a research background and several inspirations. With this key insight, we break the central question into three smaller fundamental questions. In brief, they are: (1) given a background question, whether LLMs can retrieve good inspirations; (2) with background and inspirations, whether LLMs can lead to hypothesis; and (3) whether LLMs can identify good hypotheses to rank them higher. To investigate these questions, we construct a benchmark consisting of 51 chemistry papers published in Nature, Science, or a similar level in 2024 (all papers are only available online since 2024). Every paper is divided by chemistry PhD students into three components: background, inspirations, and hypothesis. The goal is to rediscover the hypothesis, given only the background and a large randomly selected chemistry literature corpus consisting the ground truth inspiration papers, with LLMs trained with data up to 2023. We also develop an LLM-based multi-agent framework that leverages the assumption, consisting of three stages reflecting the three smaller questions. The proposed method can rediscover many hypotheses with very high similarity with the ground truth ones, covering the main innovations.
Authors: Xiaosen Zheng, Tianyu Pang, Chao Du, Qian Liu, Jing Jiang, Min Lin
Abstract: Automatic LLM benchmarks, such as AlpacaEval 2.0, Arena-Hard-Auto, and MT-Bench, have become popular for evaluating language models due to their cost-effectiveness and scalability compared to human evaluation. Achieving high win rates on these benchmarks can significantly boost the promotional impact of newly released language models. This promotional benefit may motivate tricks, such as manipulating model output length or style to game win rates, even though several mechanisms have been developed to control length and disentangle style to reduce gameability. Nonetheless, we show that even a "null model" that always outputs a constant response (irrelevant to input instructions) can cheat automatic benchmarks and achieve top-ranked win rates: an 86.5% LC win rate on AlpacaEval 2.0; an 83.0 score on Arena-Hard-Auto; and a 9.55 score on MT-Bench. Moreover, the crafted cheating outputs are transferable because we assume that the instructions of these benchmarks (e.g., 805 samples of AlpacaEval 2.0) are private and cannot be accessed. While our experiments are primarily proof-of-concept, an adversary could use LLMs to generate more imperceptible cheating responses, unethically benefiting from high win rates and promotional impact. Our findings call for the development of anti-cheating mechanisms for reliable automatic benchmarks. The code is available at https://github.com/sail-sg/Cheating-LLM-Benchmarks.
Authors: Pengcheng Ai, Xiangming Sun, Zhi Deng, Xinchi Ran
Abstract: Particle detectors based on scintillators are widely used in high-energy physics and astroparticle physics experiments, nuclear medicine imaging, industrial and environmental detection, etc. Precisely extracting scintillation signal characteristics at the event level is important for these applications, not only in respect of understanding the scintillator itself, but also kinds and physical property of incident particles. Recent researches demonstrate data-driven neural networks surpass traditional statistical methods, especially when the analytical form of signals is hard to obtain, or noise is significant. However, most densely connected or convolution-based networks fail to fully exploit the spectral and temporal structure of scintillation signals, leaving large space for performance improvement. In this paper, we propose a network architecture specially tailored for scintillation pulse characterization based on previous works on time series analysis. The core insight is that, by directly applying Fast Fourier Transform on original signals and utilizing different frequency components, the proposed network architecture can serve as a lightweight and enhanced representation learning backbone. We prove our idea in two case studies: (a) simulation data generated with the setting of the LUX dark matter detector, and (b) experimental electrical signals with fast electronics to emulate scintillation variations for the NICA/MPD calorimeter. The proposed model achieves significantly better results than the reference model in literature and densely connected models, and demonstrates higher cost-efficiency than conventional machine learning methods.
Authors: Shubham Ugare, Rohan Gumaste, Tarun Suresh, Gagandeep Singh, Sasa Misailovic
Abstract: Large Language Models (LLMs) are widely used for tasks such as natural language and code generation, but their outputs often suffer from issues like hallucination, toxicity, and incorrect results. Current libraries for structured LLM generation rely on left-to-right decoding without support for backtracking, limiting the ability to correct or refine outputs mid-generation. To address this, we introduce IterGen, a user-friendly library for iterative, grammar-guided LLM generation that enables users to move both forward and backward within the generated output based on grammar symbols. By leveraging a symbol-to-position mapping and maintaining the key-value (KV) cache state, IterGen ensures efficient and structured generation while allowing for corrections during the process. We demonstrate IterGen's effectiveness in two important applications: reducing privacy leakage in LLM outputs and improving the accuracy of LLM-generated SQL and Vega-Lite queries. Our code and additional resources are available at https://structuredllm.com.
Authors: Songming Liu, Lingxuan Wu, Bangguo Li, Hengkai Tan, Huayu Chen, Zhengyi Wang, Ke Xu, Hang Su, Jun Zhu
Abstract: Bimanual manipulation is essential in robotics, yet developing foundation models is extremely challenging due to the inherent complexity of coordinating two robot arms (leading to multi-modal action distributions) and the scarcity of training data. In this paper, we present the Robotics Diffusion Transformer (RDT), a pioneering diffusion foundation model for bimanual manipulation. RDT builds on diffusion models to effectively represent multi-modality, with innovative designs of a scalable Transformer to deal with the heterogeneity of multi-modal inputs and to capture the nonlinearity and high frequency of robotic data. To address data scarcity, we further introduce a Physically Interpretable Unified Action Space, which can unify the action representations of various robots while preserving the physical meanings of original actions, facilitating learning transferrable physical knowledge. With these designs, we managed to pre-train RDT on the largest collection of multi-robot datasets to date and scaled it up to 1.2B parameters, which is the largest diffusion-based foundation model for robotic manipulation. We finally fine-tuned RDT on a self-created multi-task bimanual dataset with over 6K+ episodes to refine its manipulation capabilities. Experiments on real robots demonstrate that RDT significantly outperforms existing methods. It exhibits zero-shot generalization to unseen objects and scenes, understands and follows language instructions, learns new skills with just 1~5 demonstrations, and effectively handles complex, dexterous tasks. We refer to https://rdt-robotics.github.io/rdt-robotics/ for the code and videos.
Authors: Xiaojian Yuan, Tianyu Pang, Chao Du, Kejiang Chen, Weiming Zhang, Min Lin
Abstract: Large language models (LLMs) may memorize sensitive or copyrighted content, raising privacy and legal concerns. Due to the high cost of retraining from scratch, researchers attempt to employ machine unlearning to remove specific content from LLMs while preserving the overall performance. In this paper, we discuss several issues in machine unlearning for LLMs and provide our insights on possible approaches. To address the issue of inadequate evaluation of model outputs after unlearning, we introduce three additional metrics to evaluate token diversity, sentence semantics, and factual correctness. We then categorize unlearning methods into untargeted and targeted, and discuss their issues respectively. Specifically, the behavior that untargeted unlearning attempts to approximate is unpredictable and may involve hallucinations, and existing regularization is insufficient for targeted unlearning. To alleviate these issues, we propose using the objective of maximizing entropy (ME) for untargeted unlearning and incorporate answer preservation (AP) loss as regularization for targeted unlearning. Experimental results across three scenarios, i.e., fictitious unlearning, continual unlearning, and real-world unlearning, demonstrate the effectiveness of our approaches. The code is available at https://github.com/sail-sg/closer-look-LLM-unlearning.
URLs: https://github.com/sail-sg/closer-look-LLM-unlearning.
Authors: Jiahao Lu, Yifan Zhang, Qiuhong Shen, Xinchao Wang, Shuicheng Yan
Abstract: 3D Gaussian splatting (3DGS), known for its groundbreaking performance and efficiency, has become a dominant 3D representation and brought progress to many 3D vision tasks. However, in this work, we reveal a significant security vulnerability that has been largely overlooked in 3DGS: the computation cost of training 3DGS could be maliciously tampered by poisoning the input data. By developing an attack named Poison-splat, we reveal a novel attack surface where the adversary can poison the input images to drastically increase the computation memory and time needed for 3DGS training, pushing the algorithm towards its worst computation complexity. In extreme cases, the attack can even consume all allocable memory, leading to a Denial-of-Service (DoS) that disrupts servers, resulting in practical damages to real-world 3DGS service vendors. Such a computation cost attack is achieved by addressing a bi-level optimization problem through three tailored strategies: attack objective approximation, proxy model rendering, and optional constrained optimization. These strategies not only ensure the effectiveness of our attack but also make it difficult to defend with simple defensive measures. We hope the revelation of this novel attack surface can spark attention to this crucial yet overlooked vulnerability of 3DGS systems. Our code is available at https://github.com/jiahaolu97/poison-splat .
Authors: Haoyi Zhu, Honghui Yang, Yating Wang, Jiange Yang, Limin Wang, Tong He
Abstract: In this paper, we introduce SPA, a novel representation learning framework that emphasizes the importance of 3D spatial awareness in embodied AI. Our approach leverages differentiable neural rendering on multi-view images to endow a vanilla Vision Transformer (ViT) with intrinsic spatial understanding. We present the most comprehensive evaluation of embodied representation learning to date, covering 268 tasks across 8 simulators with diverse policies in both single-task and language-conditioned multi-task scenarios. The results are compelling: SPA consistently outperforms more than 10 state-of-the-art representation methods, including those specifically designed for embodied AI, vision-centric tasks, and multi-modal applications, while using less training data. Furthermore, we conduct a series of real-world experiments to confirm its effectiveness in practical scenarios. These results highlight the critical role of 3D spatial awareness for embodied representation learning. Our strongest model takes more than 6000 GPU hours to train and we are committed to open-sourcing all code and model weights to foster future research in embodied representation learning. Project Page: https://haoyizhu.github.io/spa/.
Authors: Tong Wu, Shujian Zhang, Kaiqiang Song, Silei Xu, Sanqiang Zhao, Ravi Agrawal, Sathish Reddy Indurthi, Chong Xiang, Prateek Mittal, Wenxuan Zhou
Abstract: Large Language Models (LLMs) are susceptible to security and safety threats, such as prompt injection, prompt extraction, and harmful requests. One major cause of these vulnerabilities is the lack of an instruction hierarchy. Modern LLM architectures treat all inputs equally, failing to distinguish between and prioritize various types of instructions, such as system messages, user prompts, and data. As a result, lower-priority user prompts may override more critical system instructions, including safety protocols. Existing approaches to achieving instruction hierarchy, such as delimiters and instruction-based training, do not address this issue at the architectural level. We introduce the Instructional Segment Embedding (ISE) technique, inspired by BERT, to modern large language models, which embeds instruction priority information directly into the model. This approach enables models to explicitly differentiate and prioritize various instruction types, significantly improving safety against malicious prompts that attempt to override priority rules. Our experiments on the Structured Query and Instruction Hierarchy benchmarks demonstrate an average robust accuracy increase of up to 15.75% and 18.68%, respectively. Furthermore, we observe an improvement in instruction-following capability of up to 4.1% evaluated on AlpacaEval. Overall, our approach offers a promising direction for enhancing the safety and effectiveness of LLM architectures.
Authors: Xiangming Gu, Tianyu Pang, Chao Du, Qian Liu, Fengzhuo Zhang, Cunxiao Du, Ye Wang, Min Lin
Abstract: Language Models (LMs) assign significant attention to the first token, even if it is not semantically important, which is known as attention sink. This phenomenon has been widely adopted in applications such as streaming/long context generation, KV cache optimization, inference acceleration, model quantization, and others. Despite its widespread use, a deep understanding of attention sink in LMs is still lacking. In this work, we first demonstrate that attention sinks exist universally in LMs with various inputs, even in small models. Furthermore, attention sink is observed to emerge during the LM pre-training, motivating us to investigate how optimization, data distribution, loss function, and model architecture in LM pre-training influence its emergence. We highlight that attention sink emerges after effective optimization on sufficient training data. The sink position is highly correlated with the loss function and data distribution. Most importantly, we find that attention sink acts more like key biases, storing extra attention scores, which could be non-informative and not contribute to the value computation. We also observe that this phenomenon (at least partially) stems from tokens' inner dependence on attention scores as a result of softmax normalization. After relaxing such dependence by replacing softmax attention with other attention operations, such as sigmoid attention without normalization, attention sinks do not emerge in LMs up to 1B parameters. The code is available at https://github.com/sail-sg/Attention-Sink.
Authors: Luping Liu, Chao Du, Tianyu Pang, Zehan Wang, Chongxuan Li, Dong Xu
Abstract: The rapid advancement of text-to-image (T2I) diffusion models has enabled them to generate unprecedented results from given texts. However, as text inputs become longer, existing encoding methods like CLIP face limitations, and aligning the generated images with long texts becomes challenging. To tackle these issues, we propose LongAlign, which includes a segment-level encoding method for processing long texts and a decomposed preference optimization method for effective alignment training. For segment-level encoding, long texts are divided into multiple segments and processed separately. This method overcomes the maximum input length limits of pretrained encoding models. For preference optimization, we provide decomposed CLIP-based preference models to fine-tune diffusion models. Specifically, to utilize CLIP-based preference models for T2I alignment, we delve into their scoring mechanisms and find that the preference scores can be decomposed into two components: a text-relevant part that measures T2I alignment and a text-irrelevant part that assesses other visual aspects of human preference. Additionally, we find that the text-irrelevant part contributes to a common overfitting problem during fine-tuning. To address this, we propose a reweighting strategy that assigns different weights to these two components, thereby reducing overfitting and enhancing alignment. After fine-tuning $512 \times 512$ Stable Diffusion (SD) v1.5 for about 20 hours using our method, the fine-tuned SD outperforms stronger foundation models in T2I alignment, such as PixArt-$\alpha$ and Kandinsky v2.2. The code is available at https://github.com/luping-liu/LongAlign.
Authors: Paula Cordero-Encinar, Tobias Schr\"oder, Peter Yatsyshin, Andrew Duncan
Abstract: Selecting cost-effective optimal sensor configurations for subsequent inference of parameters in black-box stochastic systems faces significant computational barriers. We propose a novel and robust approach, modelling the joint distribution over input parameters and solution with a joint energy-based model, trained on simulation data. Unlike existing simulation-based inference approaches, which must be tied to a specific set of point evaluations, we learn a functional representation of parameters and solution. This is used as a resolution-independent plug-and-play surrogate for the joint distribution, which can be conditioned over any set of points, permitting an efficient approach to sensor placement. We demonstrate the validity of our framework on a variety of stochastic problems, showing that our method provides highly informative sensor locations at a lower computational cost compared to conventional approaches.
Authors: Fengyu Gao, Ruida Zhou, Tianhao Wang, Cong Shen, Jing Yang
Abstract: Large Language Models (LLMs) rely on the contextual information embedded in examples/demonstrations to perform in-context learning (ICL). To mitigate the risk of LLMs potentially leaking private information contained in examples in the prompt, we introduce a novel data-adaptive differentially private algorithm called AdaDPSyn to generate synthetic examples from the private dataset and then use these synthetic examples to perform ICL. The objective of AdaDPSyn is to adaptively adjust the noise level in the data synthesis mechanism according to the inherent statistical properties of the data, thereby preserving high ICL accuracy while maintaining formal differential privacy guarantees. A key innovation in AdaDPSyn is the Precision-Focused Iterative Radius Reduction technique, which dynamically refines the aggregation radius - the scope of data grouping for noise addition - based on patterns observed in data clustering, thereby minimizing the amount of additive noise. We conduct extensive experiments on standard benchmarks and compare AdaDPSyn with DP few-shot generation algorithm (Tang et al., 2023). The experiments demonstrate that AdaDPSyn not only outperforms DP few-shot generation, but also maintains high accuracy levels close to those of non-private baselines, providing an effective solution for ICL with privacy protection.
Authors: Caigao Jiang, Xiang Shu, Hong Qian, Xingyu Lu, Jun Zhou, Aimin Zhou, Yang Yu
Abstract: Optimization problems are prevalent across various scenarios. Formulating and then solving optimization problems described by natural language often requires highly specialized human expertise, which could block the widespread application of optimization-based decision making. To automate problem formulation and solving, leveraging large language models (LLMs) has emerged as a potential way. However, this kind of approach suffers from the issue of optimization generalization. Namely, the accuracy of most current LLM-based methods and the generality of optimization problem types that they can model are still limited. In this paper, we propose a unified learning-based framework called LLMOPT to boost optimization generalization. Starting from the natural language descriptions of optimization problems and a pre-trained LLM, LLMOPT constructs the introduced five-element formulation as a universal model for learning to define diverse optimization problem types. Then, LLMOPT employs the multi-instruction tuning to enhance both problem formalization and solver code generation accuracy and generality. After that, to prevent hallucinations in LLMs, such as sacrificing solving accuracy to avoid execution errors, the model alignment and self-correction mechanism are adopted in LLMOPT. We evaluate the optimization generalization ability of LLMOPT and compared methods across six real-world datasets covering roughly 20 fields such as health, environment, energy and manufacturing, etc. Extensive experiment results show that LLMOPT is able to model various optimization problem types such as linear/nonlinear programming, mixed integer programming, and combinatorial optimization, and achieves a notable 11.08% average solving accuracy improvement compared with the state-of-the-art methods. The code is available at https://github.com/caigaojiang/LLMOPT.
Authors: Antonio Sclocchi, Alessandro Favero, Noam Itzhak Levi, Matthieu Wyart
Abstract: High-dimensional data must be highly structured to be learnable. Although the compositional and hierarchical nature of data is often put forward to explain learnability, quantitative measurements establishing these properties are scarce. Likewise, accessing the latent variables underlying such a data structure remains a challenge. In this work, we show that forward-backward experiments in diffusion-based models, where data is noised and then denoised to generate new samples, are a promising tool to probe the latent structure of data. We predict in simple hierarchical models that, in this process, changes in data occur by correlated chunks, with a length scale that diverges at a noise level where a phase transition is known to take place. Remarkably, we confirm this prediction in both text and image datasets using state-of-the-art diffusion models. Our results show how latent variable changes manifest in the data and establish how to measure these effects in real data using diffusion models.
Authors: Maitreya Prafulla Chitale, Uday Bindal, Rajakrishnan Rajkumar, Rahul Mishra
Abstract: Summarizing movie screenplays presents a unique set of challenges compared to standard document summarization. Screenplays are not only lengthy, but also feature a complex interplay of characters, dialogues, and scenes, with numerous direct and subtle relationships and contextual nuances that are difficult for machine learning models to accurately capture and comprehend. Recent attempts at screenplay summarization focus on fine-tuning transformer-based pre-trained models, but these models often fall short in capturing long-term dependencies and latent relationships, and frequently encounter the "lost in the middle" issue. To address these challenges, we introduce DiscoGraMS, a novel resource that represents movie scripts as a movie character-aware discourse graph (CaD Graph). This approach is well-suited for various downstream tasks, such as summarization, question-answering, and salience detection. The model aims to preserve all salient information, offering a more comprehensive and faithful representation of the screenplay's content. We further explore a baseline method that combines the CaD Graph with the corresponding movie script through a late fusion of graph and text modalities, and we present very initial promising results.
Authors: Rafael Rivera Soto, Barry Chen, Nicholas Andrews
Abstract: High-quality paraphrases are easy to produce using instruction-tuned language models or specialized paraphrasing models. Although this capability has a variety of benign applications, paraphrasing attacks$\unicode{x2013}$paraphrases applied to machine-generated texts$\unicode{x2013}$are known to significantly degrade the performance of machine-text detectors. This motivates us to consider the novel problem of paraphrase inversion, where, given paraphrased text, the objective is to recover an approximation of the original text. The closer the approximation is to the original text, the better machine-text detectors will perform. We propose an approach which frames the problem as translation from paraphrased text back to the original text, which requires examples of texts and corresponding paraphrases to train the inversion model. Fortunately, such training data can easily be generated, given a corpus of original texts and one or more paraphrasing models. We find that language models such as GPT-4 and Llama-3 exhibit biases when paraphrasing which an inversion model can learn with a modest amount of data. Perhaps surprisingly, we also find that such models generalize well, including to paraphrase models unseen at training time. Finally, we show that when combined with a paraphrased-text detector, our inversion models provide an effective defense against paraphrasing attacks, and overall our approach yields an average improvement of +22% AUROC across seven machine-text detectors and three different domains.
Authors: Vincent Guan, Joseph Janssen, Hossein Rahmani, Andrew Warren, Stephen Zhang, Elina Robeva, Geoffrey Schiebinger
Abstract: Stochastic differential equations (SDEs) are a fundamental tool for modelling dynamic processes, including gene regulatory networks (GRNs), contaminant transport, financial markets, and image generation. However, learning the underlying SDE from data is a challenging task, especially if individual trajectories are not observable. Motivated by burgeoning research in single-cell datasets, we present the first comprehensive approach for jointly identifying the drift and diffusion of an SDE from its temporal marginals. Assuming linear drift and additive diffusion, we prove that these parameters are identifiable from marginals if and only if the initial distribution lacks any generalized rotational symmetries. We further prove that the causal graph of any SDE with additive diffusion can be recovered from the SDE parameters. To complement this theory, we adapt entropy-regularized optimal transport to handle anisotropic diffusion, and introduce APPEX (Alternating Projection Parameter Estimation from $X_0$), an iterative algorithm designed to estimate the drift, diffusion, and causal graph of an additive noise SDE, solely from temporal marginals. We show that APPEX iteratively decreases Kullback-Leibler divergence to the true solution, and demonstrate its effectiveness on simulated data from linear additive noise SDEs.
Authors: Yichao Liang, Nishanth Kumar, Hao Tang, Adrian Weller, Joshua B. Tenenbaum, Tom Silver, Jo\~ao F. Henriques, Kevin Ellis
Abstract: Broadly intelligent agents should form task-specific abstractions that selectively expose the essential elements of a task, while abstracting away the complexity of the raw sensorimotor space. In this work, we present Neuro-Symbolic Predicates, a first-order abstraction language that combines the strengths of symbolic and neural knowledge representations. We outline an online algorithm for inventing such predicates and learning abstract world models. We compare our approach to hierarchical reinforcement learning, vision-language model planning, and symbolic predicate invention approaches, on both in- and out-of-distribution tasks across five simulated robotic domains. Results show that our approach offers better sample complexity, stronger out-of-distribution generalization, and improved interpretability.
Authors: S Balasubramanian, M Sai Subramaniam, Sai Sriram Talasu, Yedu Krishna P, Manepalli Pranav Phanindra Sai, Ravi Mukkamala, Darshan Gera
Abstract: Deep neural networks (DNNS) excel at learning from static datasets but struggle with continual learning, where data arrives sequentially. Catastrophic forgetting, the phenomenon of forgetting previously learned knowledge, is a primary challenge. This paper introduces EXponentially Averaged Class-wise Feature Significance (EXACFS) to mitigate this issue in the class incremental learning (CIL) setting. By estimating the significance of model features for each learned class using loss gradients, gradually aging the significance through the incremental tasks and preserving the significant features through a distillation loss, EXACFS effectively balances remembering old knowledge (stability) and learning new knowledge (plasticity). Extensive experiments on CIFAR-100 and ImageNet-100 demonstrate EXACFS's superior performance in preserving stability while acquiring plasticity.
Authors: Lizhe Fang, Yifei Wang, Zhaoyang Liu, Chenheng Zhang, Stefanie Jegelka, Jinyang Gao, Bolin Ding, Yisen Wang
Abstract: Handling long-context inputs is crucial for large language models (LLMs) in tasks such as extended conversations, document summarization, and many-shot in-context learning. While recent approaches have extended the context windows of LLMs and employed perplexity (PPL) as a standard evaluation metric, PPL has proven unreliable for assessing long-context capabilities. The underlying cause of this limitation has remained unclear. In this work, we provide a comprehensive explanation for this issue. We find that PPL overlooks key tokens, which are essential for long-context understanding, by averaging across all tokens and thereby obscuring the true performance of models in long-context scenarios. To address this, we propose \textbf{LongPPL}, a novel metric that focuses on key tokens by employing a long-short context contrastive method to identify them. Our experiments demonstrate that LongPPL strongly correlates with performance on various long-context benchmarks (e.g., Pearson correlation of -0.96), significantly outperforming traditional PPL in predictive accuracy. Additionally, we introduce \textbf{LongCE} (Long-context Cross-Entropy) loss, a re-weighting strategy for fine-tuning that prioritizes key tokens, leading to consistent improvements across diverse benchmarks. In summary, these contributions offer deeper insights into the limitations of PPL and present effective solutions for accurately evaluating and enhancing the long-context capabilities of LLMs. Code is available at https://github.com/PKU-ML/LongPPL.
Authors: Neel Dey, Benjamin Billot, Hallee E. Wong, Clinton J. Wang, Mengwei Ren, P. Ellen Grant, Adrian V. Dalca, Polina Golland
Abstract: Current volumetric biomedical foundation models struggle to generalize as public 3D datasets are small and do not cover the broad diversity of medical procedures, conditions, anatomical regions, and imaging protocols. We address this by creating a representation learning method that instead anticipates strong domain shifts at training time itself. We first propose a data engine that synthesizes highly variable training samples that would enable generalization to new biomedical contexts. To then train a single 3D network for any voxel-level task, we develop a contrastive learning method that pretrains the network to be stable against nuisance imaging variation simulated by the data engine, a key inductive bias for generalization. This network's features can be used as robust representations of input images for downstream tasks and its weights provide a strong, dataset-agnostic initialization for finetuning on new datasets. As a result, we set new standards across both multimodality registration and few-shot segmentation, a first for any 3D biomedical vision model, all without (pre-)training on any existing dataset of real images.
Authors: Wei Wu, Zhuoshi Pan, Chao Wang, Liyi Chen, Yunchu Bai, Tianfu Wang, Kun Fu, Zheng Wang, Hui Xiong
Abstract: The rapid advancement of Large Language Models (LLMs) has driven growing demand for processing extended context sequences in contemporary applications. However, this progress faces two major challenges: performance degradation due to sequence lengths out-of-distribution, and excessively long inference times caused by the quadratic computational complexity of attention. These issues hinder the application of LLMs in long-context scenarios. In this paper, we propose Dynamic Token-Level KV Cache Selection (TokenSelect), a training-free method for efficient and accurate long-context inference. TokenSelect builds upon the observation of non-contiguous attention sparsity, using Query-Key dot products to measure per-head KV Cache criticality at token-level. By per-head soft voting mechanism, TokenSelect selectively involves a few critical KV cache tokens in attention calculation without sacrificing accuracy. To further accelerate TokenSelect, we design the Selection Cache based on observations of consecutive Query similarity and implemented efficient dot product kernel, significantly reducing the overhead. A comprehensive evaluation of TokenSelect demonstrates up to 23.84x speedup in attention computation and up to 2.28x acceleration in end-to-end latency, while providing superior performance compared to state-of-the-art long-context inference methods.
Authors: Harin Lee, Min-hwan Oh
Abstract: In this work, we close the fundamental gap of theory and practice by providing an improved regret bound for linear ensemble sampling. We prove that with an ensemble size logarithmic in $T$, linear ensemble sampling can achieve a frequentist regret bound of $\tilde{\mathcal{O}}(d^{3/2}\sqrt{T})$, matching state-of-the-art results for randomized linear bandit algorithms, where $d$ and $T$ are the dimension of the parameter and the time horizon respectively. Our approach introduces a general regret analysis framework for linear bandit algorithms. Additionally, we reveal a significant relationship between linear ensemble sampling and Linear Perturbed-History Exploration (LinPHE), showing that LinPHE is a special case of linear ensemble sampling when the ensemble size equals $T$. This insight allows us to derive a new regret bound of $\tilde{\mathcal{O}}(d^{3/2}\sqrt{T})$ for LinPHE, independent of the number of arms. Our contributions advance the theoretical foundation of ensemble sampling, bringing its regret bounds in line with the best known bounds for other randomized exploration algorithms.
Authors: Chenrui Tie, Yue Chen, Ruihai Wu, Boxuan Dong, Zeyi Li, Chongkai Gao, Hao Dong
Abstract: Imitation learning, e.g., diffusion policy, has been proven effective in various robotic manipulation tasks. However, extensive demonstrations are required for policy robustness and generalization. To reduce the demonstration reliance, we leverage spatial symmetry and propose ET-SEED, an efficient trajectory-level SE(3) equivariant diffusion model for generating action sequences in complex robot manipulation tasks. Further, previous equivariant diffusion models require the per-step equivariance in the Markov process, making it difficult to learn policy under such strong constraints. We theoretically extend equivariant Markov kernels and simplify the condition of equivariant diffusion process, thereby significantly improving training efficiency for trajectory-level SE(3) equivariant diffusion policy in an end-to-end manner. We evaluate ET-SEED on representative robotic manipulation tasks, involving rigid body, articulated and deformable object. Experiments demonstrate superior data efficiency and manipulation proficiency of our proposed method, as well as its ability to generalize to unseen configurations with only a few demonstrations. Website: https://et-seed.github.io/
Authors: Muyang Li, Yujun Lin, Zhekai Zhang, Tianle Cai, Xiuyu Li, Junxian Guo, Enze Xie, Chenlin Meng, Jun-Yan Zhu, Song Han
Abstract: Diffusion models can effectively generate high-quality images. However, as they scale, rising memory demands and higher latency pose substantial deployment challenges. In this work, we aim to accelerate diffusion models by quantizing their weights and activations to 4 bits. At such an aggressive level, both weights and activations are highly sensitive, where existing post-training quantization methods like smoothing become insufficient. To overcome this limitation, we propose SVDQuant, a new 4-bit quantization paradigm. Different from smoothing, which redistributes outliers between weights and activations, our approach absorbs these outliers using a low-rank branch. We first consolidate the outliers by shifting them from activations to weights. Then, we use a high-precision, low-rank branch to take in the weight outliers with Singular Value Decomposition (SVD), while a low-bit quantized branch handles the residuals. This process eases the quantization on both sides. However, naively running the low-rank branch independently incurs significant overhead due to extra data movement of activations, negating the quantization speedup. To address this, we co-design an inference engine Nunchaku that fuses the kernels of the low-rank branch into those of the low-bit branch to cut off redundant memory access. It can also seamlessly support off-the-shelf low-rank adapters (LoRAs) without re-quantization. Extensive experiments on SDXL, PixArt-$\Sigma$, and FLUX.1 validate the effectiveness of SVDQuant in preserving image quality. We reduce the memory usage for the 12B FLUX.1 models by 3.5$\times$, achieving 3.0$\times$ speedup over the 4-bit weight-only quantization (W4A16) baseline on the 16GB laptop 4090 GPU with INT4 precision. On the latest RTX 5090 desktop with Blackwell architecture, we achieve a 3.1$\times$ speedup compared to the W4A16 model using NVFP4 precision.
Authors: Shauli Ravfogel, Anej Svete, V\'esteinn Sn{\ae}bjarnarson, Ryan Cotterell
Abstract: Understanding and manipulating the causal generation mechanisms in language models is essential for controlling their behavior. Previous work has primarily relied on techniques such as representation surgery -- e.g., model ablations or manipulation of linear subspaces tied to specific concepts -- to \emph{intervene} on these models. To understand the impact of interventions precisely, it is useful to examine \emph{counterfactuals} -- e.g., how a given sentence would have appeared had it been generated by the model following a specific intervention. We highlight that counterfactual reasoning is conceptually distinct from interventions, as articulated in Pearl's causal hierarchy. Based on this observation, we propose a framework for generating true string counterfactuals by reformulating language models as a structural equation model using the Gumbel-max trick, which we called Gumbel counterfactual generation. This reformulation allows us to model the joint distribution over original strings and their counterfactuals resulting from the same instantiation of the sampling noise. We develop an algorithm based on hindsight Gumbel sampling that allows us to infer the latent noise variables and generate counterfactuals of observed strings. Our experiments demonstrate that the approach produces meaningful counterfactuals while at the same time showing that commonly used intervention techniques have considerable undesired side effects.
Authors: Ho Fung Tsoi, Dylan Rankin, Cecile Caillol, Miles Cranmer, Sridhara Dasu, Javier Duarte, Philip Harris, Elliot Lipeles, Vladimir Loncar
Abstract: We introduce SymbolFit, a framework that automates parametric modeling by using symbolic regression to perform a machine-search for functions that fit the data while simultaneously providing uncertainty estimates in a single run. Traditionally, constructing a parametric model to accurately describe binned data has been a manual and iterative process, requiring an adequate functional form to be determined before the fit can be performed. The main challenge arises when the appropriate functional forms cannot be derived from first principles, especially when there is no underlying true closed-form function for the distribution. In this work, we develop a framework that automates and streamlines the process by utilizing symbolic regression, a machine learning technique that explores a vast space of candidate functions without requiring a predefined functional form because the functional form itself is treated as a trainable parameter, making the process far more efficient and effortless than traditional regression methods. We demonstrate the framework in high-energy physics experiments at the CERN Large Hadron Collider (LHC) using five real proton-proton collision datasets from new physics searches, including background modeling in resonance searches for high-mass dijet, trijet, paired-dijet, diphoton, and dimuon events. We show that our framework can flexibly and efficiently generate a wide range of candidate functions that fit a nontrivial distribution well using a simple fit configuration that varies only by random seed, and that the same fit configuration, which defines a vast function space, can also be applied to distributions of different shapes, whereas achieving a comparable result with traditional methods would have required extensive manual effort.
Authors: Quang P. M. Pham, Khoi T. N. Nguyen, Lan C. Ngo, Truong Do, Dezhen Song, Truong-Son Hy
Abstract: Scene graphs have proven to be highly effective for various scene understanding tasks due to their compact and explicit representation of relational information. However, current methods often overlook the critical importance of preserving symmetry when generating scene graphs from 3D point clouds, which can lead to reduced accuracy and robustness, particularly when dealing with noisy, multi-view data. Furthermore, a major limitation of prior approaches is the lack of temporal modeling to capture time-dependent relationships among dynamically evolving entities in a scene. To address these challenges, we propose Temporal Equivariant Scene Graph Neural Network (TESGNN), consisting of two key components: (1) an Equivariant Scene Graph Neural Network (ESGNN), which extracts information from 3D point clouds to generate scene graph while preserving crucial symmetry properties, and (2) a Temporal Graph Matching Network, which fuses scene graphs generated by ESGNN across multiple time sequences into a unified global representation using an approximate graph-matching algorithm. Our combined architecture TESGNN outperforms current state-of-the-art methods in scene graph generation, achieving higher accuracy and faster training convergence. Moreover, we show that leveraging the symmetry-preserving property produces a more stable and accurate global scene representation compared to existing approaches. Last but not least, it is computationally efficient and easily implementable using existing frameworks, making it well-suited for real-time applications in robotics and computer vision. This approach paves the way for more robust and scalable solutions to complex multi-view scene understanding challenges. Our source code is publicly available at: https://github.com/HySonLab/TESGraph
Authors: Dengsheng Chen, Jie Hu, Tiezhu Yue, Xiaoming Wei, Enhua Wu
Abstract: Recently, autoregressive models have demonstrated remarkable performance in class-conditional image generation. However, the application of next-token prediction to high-resolution text-to-image generation remains largely unexplored. In this paper, we introduce \textbf{D-JEPA$\cdot$T2I}, an autoregressive model based on continuous tokens that incorporates innovations in both architecture and training strategy to generate high-quality, photorealistic images at arbitrary resolutions, up to 4K. Architecturally, we adopt the denoising joint embedding predictive architecture (D-JEPA) while leveraging a multimodal visual transformer to effectively integrate textual and visual features. Additionally, we introduce flow matching loss alongside the proposed Visual Rotary Positional Embedding (VoPE) to enable continuous resolution learning. In terms of training strategy, we propose a data feedback mechanism that dynamically adjusts the sampling procedure based on statistical analysis and an online learning critic model. This encourages the model to move beyond its comfort zone, reducing redundant training on well-mastered scenarios and compelling it to address more challenging cases with suboptimal generation quality. For the first time, we achieve state-of-the-art high-resolution image synthesis via next-token prediction.
Authors: Suramya Jadhav, Abhay Shanbhag, Amogh Thakurdesai, Ridhima Sinare, Raviraj Joshi
Abstract: Low-resource languages face significant challenges due to the lack of sufficient linguistic data, resources, and tools for tasks such as supervised learning, annotation, and classification. This shortage hinders the development of accurate models and datasets, making it difficult to perform critical NLP tasks like sentiment analysis or hate speech detection. To bridge this gap, Large Language Models (LLMs) present an opportunity for potential annotators, capable of generating datasets and resources for these underrepresented languages. In this paper, we focus on Marathi, a low-resource language, and evaluate the performance of both closed-source and open-source LLMs as annotators, while also comparing these results with fine-tuned BERT models. We assess models such as GPT-4o and Gemini 1.0 Pro, Gemma 2 (2B and 9B), and Llama 3.1 (8B and 405B) on classification tasks including sentiment analysis, news classification, and hate speech detection. Our findings reveal that while LLMs excel in annotation tasks for high-resource languages like English, they still fall short when applied to Marathi. Even advanced models like GPT-4o and Llama 3.1 405B underperform compared to fine-tuned BERT-based baselines, with GPT-4o and Llama 3.1 405B trailing fine-tuned BERT by accuracy margins of 10.2% and 14.1%, respectively. This highlights the limitations of LLMs as annotators for low-resource languages.
Authors: Yue Wang, Xu Cao, Yaojun Hu, Haochao Ying, Hongxia Xu, Ruijia Wu, James Matthew Rehg, Jimeng Sun, Jian Wu, Jintai Chen
Abstract: Electrocardiogram (ECG), a non-invasive and affordable tool for cardiac monitoring, is highly sensitive in detecting acute heart attacks. However, due to the lengthy nature of ECG recordings, numerous machine learning methods have been developed for automated heart disease detection to reduce human workload. Despite these efforts, performance remains suboptimal. A key obstacle is the inherent complexity of ECG data, which includes heterogeneity (e.g., varying sampling rates), high levels of noise, demographic-related pattern shifts, and intricate rhythm-event associations. To overcome these challenges, this paper introduces AnyECG, a foundational model designed to extract robust representations from any real-world ECG data. Specifically, a tailored ECG Tokenizer encodes each fixed-duration ECG fragment into a token and, guided by proxy tasks, converts noisy, continuous ECG features into discrete, compact, and clinically meaningful local rhythm codes. These codes encapsulate basic morphological, frequency, and demographic information (e.g., sex), effectively mitigating signal noise. We further pre-train the AnyECG to learn rhythmic pattern associations across ECG tokens, enabling the capture of cardiac event semantics. By being jointly pre-trained on diverse ECG data sources, AnyECG is capable of generalizing across a wide range of downstream tasks where ECG signals are recorded from various devices and scenarios. The experimental results show that AnyECG achieves an average performance improvement of 6% across four critical tasks-anomaly detection, arrhythmia classification, corrupted lead generation, and ultra-long ECG recognition. AnyECG learns common ECG rhythm from data and significantly outperforms state-of-the-art methods in each of these tasks.
Authors: Shuangqi Li, Hieu Le, Jingyi Xu, Mathieu Salzmann
Abstract: Text-to-image diffusion models have demonstrated remarkable capability in generating realistic images from arbitrary text prompts. However, they often produce inconsistent results for compositional prompts such as "two dogs" or "a penguin on the right of a bowl". Understanding these inconsistencies is crucial for reliable image generation. In this paper, we highlight the significant role of initial noise in these inconsistencies, where certain noise patterns are more reliable for compositional prompts than others. Our analyses reveal that different initial random seeds tend to guide the model to place objects in distinct image areas, potentially adhering to specific patterns of camera angles and image composition associated with the seed. To improve the model's compositional ability, we propose a method for mining these reliable cases, resulting in a curated training set of generated images without requiring any manual annotation. By fine-tuning text-to-image models on these generated images, we significantly enhance their compositional capabilities. For numerical composition, we observe relative increases of 29.3% and 19.5% for Stable Diffusion and PixArt-{\alpha}, respectively. Spatial composition sees even larger gains, with 60.7% for Stable Diffusion and 21.1% for PixArt-{\alpha}.
Authors: Andi Han, Wei Huang, Yuan Cao, Difan Zou
Abstract: The predominant success of diffusion models in generative modeling has spurred significant interest in understanding their theoretical foundations. In this work, we propose a feature learning framework aimed at analyzing and comparing the training dynamics of diffusion models with those of traditional classification models. Our theoretical analysis demonstrates that diffusion models, due to the denoising objective, are encouraged to learn more balanced and comprehensive representations of the data. In contrast, neural networks with a similar architecture trained for classification tend to prioritize learning specific patterns in the data, often focusing on easy-to-learn components. To support these theoretical insights, we conduct several experiments on both synthetic and real-world datasets, which empirically validate our findings and highlight the distinct feature learning dynamics in diffusion models compared to classification.
Authors: N. Sukumar, Amit Acharya
Abstract: Many partial differential equations (PDEs) such as Navier--Stokes equations in fluid mechanics, inelastic deformation in solids, and transient parabolic and hyperbolic equations do not have an exact, primal variational structure. Recently, a variational principle based on the dual (Lagrange multiplier) field was proposed. The essential idea in this approach is to treat the given PDEs as constraints, and to invoke an arbitrarily chosen auxiliary potential with strong convexity properties to be optimized. On requiring the vanishing of the gradient of the Lagrangian with respect to the primal variables, a mapping from the dual to the primal fields is obtained. This leads to requiring a convex dual functional to be minimized subject to Dirichlet boundary conditions on dual variables, with the guarantee that even PDEs that do not possess a variational structure in primal form can be solved via a variational principle. The vanishing of the first variation of the dual functional is, up to Dirichlet boundary conditions on dual fields, the weak form of the primal PDE problem with the dual-to-primal change of variables incorporated. We derive the dual weak form for the linear, one-dimensional, transient convection-diffusion equation. A Galerkin discretization is used, with the trial and test functions chosen as linear combination of either shallow neural networks with RePU activation functions or B-splines; the corresponding stiffness matrix is symmetric. For transient problems, a space-time Galerkin implementation is used with tensor-product B-splines as approximating functions. Numerical results are presented for the steady-state and transient convection-diffusion equation, and transient heat conduction. The proposed method delivers sound accuracy for ODEs and PDEs and rates of convergence are established in the $L^2$ norm and $H^1$ seminorm for the steady-state convection-diffusion problem.
Authors: Jiquan Wang, Sha Zhao, Zhiling Luo, Yangxuan Zhou, Haiteng Jiang, Shijian Li, Tao Li, Gang Pan
Abstract: Electroencephalography (EEG) is a non-invasive technique to measure and record brain electrical activity, widely used in various BCI and healthcare applications. Early EEG decoding methods rely on supervised learning, limited by specific tasks and datasets, hindering model performance and generalizability. With the success of large language models, there is a growing body of studies focusing on EEG foundation models. However, these studies still leave challenges: Firstly, most of existing EEG foundation models employ full EEG modeling strategy. It models the spatial and temporal dependencies between all EEG patches together, but ignores that the spatial and temporal dependencies are heterogeneous due to the unique structural characteristics of EEG signals. Secondly, existing EEG foundation models have limited generalizability on a wide range of downstream BCI tasks due to varying formats of EEG data, making it challenging to adapt to. To address these challenges, we propose a novel foundation model called CBraMod. Specifically, we devise a criss-cross transformer as the backbone to thoroughly leverage the structural characteristics of EEG signals, which can model spatial and temporal dependencies separately through two parallel attention mechanisms. And we utilize an asymmetric conditional positional encoding scheme which can encode positional information of EEG patches and be easily adapted to the EEG with diverse formats. CBraMod is pre-trained on a very large corpus of EEG through patch-based masked EEG reconstruction. We evaluate CBraMod on up to 10 downstream BCI tasks (12 public datasets). CBraMod achieves the state-of-the-art performance across the wide range of tasks, proving its strong capability and generalizability. The source code is publicly available at https://github.com/wjq-learning/CBraMod.
Authors: Ziyang Chen, Yiwen Ye, Feilong Tang, Yongsheng Pan, Yong Xia
Abstract: Domain generalization (DG) aims to enhance the ability of models trained on source domains to generalize effectively to unseen domains. Recently, Sharpness-Aware Minimization (SAM) has shown promise in this area by reducing the sharpness of the loss landscape to obtain more generalized models. However, SAM and its variants sometimes fail to guide the model toward a flat minimum, and their training processes exhibit limitations, hindering further improvements in model generalization. In this paper, we first propose an improved model training process aimed at encouraging the model to converge to a flat minima. To achieve this, we design a curvature metric that has a minimal effect when the model is far from convergence but becomes increasingly influential in indicating the curvature of the minima as the model approaches a local minimum. Then we derive a novel algorithm from this metric, called Meta Curvature-Aware Minimization (MeCAM), to minimize the curvature around the local minima. Specifically, the optimization objective of MeCAM simultaneously minimizes the regular training loss, the surrogate gap of SAM, and the surrogate gap of meta-learning. We provide theoretical analysis on MeCAM's generalization error and convergence rate, and demonstrate its superiority over existing DG methods through extensive experiments on five benchmark DG datasets, including PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet. Code will be available on GitHub.
Authors: Marco Minici, Luca Luceri, Francesco Fabbri, Emilio Ferrara
Abstract: Social media platforms have become vital spaces for public discourse, serving as modern agor\`as where a wide range of voices influence societal narratives. However, their open nature also makes them vulnerable to exploitation by malicious actors, including state-sponsored entities, who can conduct information operations (IOs) to manipulate public opinion. The spread of misinformation, false news, and misleading claims threatens democratic processes and societal cohesion, making it crucial to develop methods for the timely detection of inauthentic activity to protect the integrity of online discourse. In this work, we introduce a methodology designed to identify users orchestrating information operations, a.k.a. IO drivers, across various influence campaigns. Our framework, named IOHunter, leverages the combined strengths of Language Models and Graph Neural Networks to improve generalization in supervised, scarcely-supervised, and cross-IO contexts. Our approach achieves state-of-the-art performance across multiple sets of IOs originating from six countries, significantly surpassing existing approaches. This research marks a step toward developing Graph Foundation Models specifically tailored for the task of IO detection on social media platforms.
Authors: Arun K. Sharma, Shubhobrata Bhattacharya, Motahar Reza, Bishakh Bhattacharya
Abstract: Traditional biometric systems have encountered significant setbacks due to various unavoidable factors, for example, face recognition-based biometrics fails due to the wearing of face masks and fingerprints create hygiene concerns. This paper proposes a novel lightweight cross-spectral vision transformer (CS-ViT) for biometric authentication using forehead subcutaneous vein patterns and periocular patterns, offering a promising alternative to traditional methods, capable of performing well even with the face masks and without any physical touch. The proposed framework comprises a cross-spectral dual-channel architecture designed to handle two distinct biometric traits and to capture inter-dependencies in terms of relative spectral patterns. Each channel consists of a Phase-Only Correlation Cross-Spectral Attention (POC-CSA) that captures their individual as well as correlated patterns. The computation of cross-spectral attention using POC extracts the phase correlation in the spatial features. Therefore, it is robust against the resolution/intensity variations and illumination of the input images, assuming both biometric traits are from the same person. The lightweight model is suitable for edge device deployment. The performance of the proposed algorithm was rigorously evaluated using the Forehead Subcutaneous Vein Pattern and Periocular Biometric Pattern (FSVP-PBP) database. The results demonstrated the superiority of the algorithm over state-of-the-art methods, achieving a remarkable classification accuracy of 98.8% with the combined vein and periocular patterns.
Authors: Yijia Xiao, Edward Sun, Di Luo, Wei Wang
Abstract: Significant progress has been made in automated problem-solving using societies of agents powered by large language models (LLMs). In finance, efforts have largely focused on single-agent systems handling specific tasks or multi-agent frameworks independently gathering data. However, multi-agent systems' potential to replicate real-world trading firms' collaborative dynamics remains underexplored. TradingAgents proposes a novel stock trading framework inspired by trading firms, featuring LLM-powered agents in specialized roles such as fundamental analysts, sentiment analysts, technical analysts, and traders with varied risk profiles. The framework includes Bull and Bear researcher agents assessing market conditions, a risk management team monitoring exposure, and traders synthesizing insights from debates and historical data to make informed decisions. By simulating a dynamic, collaborative trading environment, this framework aims to improve trading performance. Detailed architecture and extensive experiments reveal its superiority over baseline models, with notable improvements in cumulative returns, Sharpe ratio, and maximum drawdown, highlighting the potential of multi-agent LLM frameworks in financial trading. TradingAgents is available at https://github.com/PioneerFintech.
Authors: Yixin Ji, Juntao Li, Hai Ye, Kaixin Wu, Kai Yao, Jia Xu, Linjian Mo, Min Zhang
Abstract: The remarkable performance of the o1 model in complex reasoning demonstrates that test-time compute scaling can further unlock the model's potential, enabling powerful System-2 thinking. However, there is still a lack of comprehensive surveys for test-time compute scaling. We trace the concept of test-time compute back to System-1 models. In System-1 models, test-time compute addresses distribution shifts and improves robustness and generalization through parameter updating, input modification, representation editing, and output calibration. In System-2 models, it enhances the model's reasoning ability to solve complex problems through repeated sampling, self-correction, and tree search. We organize this survey according to the trend of System-1 to System-2 thinking, highlighting the key role of test-time compute in the transition from System-1 models to weak System-2 models, and then to strong System-2 models. We also point out a few possible future directions.
Authors: Vighnesh Subramaniam, Yilun Du, Joshua B. Tenenbaum, Antonio Torralba, Shuang Li, Igor Mordatch
Abstract: Large language models (LLMs) have achieved remarkable performance in recent years but are fundamentally limited by the underlying training data. To improve models beyond the training data, recent works have explored how LLMs can be used to generate synthetic data for autonomous self-improvement. However, successive steps of self-improvement can reach a point of diminishing returns. In this work, we propose a complementary approach towards self-improvement where finetuning is applied to a multiagent society of language models. A group of language models, all starting from the same base model, are independently specialized by updating each one using data generated through multiagent interactions among the models. By training each model on independent sets of data, we illustrate how this approach enables specialization across models and diversification over the set of models. As a result, our overall system is able to preserve diverse reasoning chains and autonomously improve over many more rounds of fine-tuning than single-agent self-improvement methods. We quantitatively illustrate the efficacy of the approach across a wide suite of reasoning tasks.
Authors: Fei Wu, Jia Hu, Geyong Min, Shiqiang Wang
Abstract: Pre-trained Language Models (PLMs) have demonstrated their superiority and versatility in modern Natural Language Processing (NLP), effectively adapting to various downstream tasks through further fine-tuning. Federated Parameter-Efficient Fine-Tuning (FedPEFT) has emerged as a promising solution to address privacy and efficiency challenges in distributed training for PLMs on resource-constrained local devices. However, our measurements reveal two key limitations of FedPEFT: heterogeneous data across devices leads to significant performance degradation, and a fixed parameter configuration results in communication inefficiency. To overcome these limitations, we propose FedARA, a novel Adaptive Rank Allocation framework for federated parameter-efficient fine-tuning of language models. Specifically, FedARA employs truncated Singular Value Decomposition (SVD) adaptation to enhance similar feature representation across clients, significantly mitigating the adverse effects of data heterogeneity. Subsequently, it utilizes dynamic rank allocation to progressively identify critical ranks, effectively improving communication efficiency. Lastly, it leverages rank-based module pruning to automatically remove inactive modules, steadily reducing local computational cost and memory usage in each federated learning round. Extensive experiments show that FedARA consistently outperforms baselines by an average of 6.95% to 8.49% across various datasets and models under heterogeneous data while significantly improving communication efficiency by 2.40$ \times$. Moreover, experiments on various edge devices demonstrate substantial decreases in total training time and energy consumption by up to 48.90% and 46.95%, respectively.
Authors: Adil Kaan Akan, Yucel Yemez
Abstract: We present SlotAdapt, an object-centric learning method that combines slot attention with pretrained diffusion models by introducing adapters for slot-based conditioning. Our method preserves the generative power of pretrained diffusion models, while avoiding their text-centric conditioning bias. We also incorporate an additional guidance loss into our architecture to align cross-attention from adapter layers with slot attention. This enhances the alignment of our model with the objects in the input image without using external supervision. Experimental results show that our method outperforms state-of-the-art techniques in object discovery and image generation tasks across multiple datasets, including those with real images. Furthermore, we demonstrate through experiments that our method performs remarkably well on complex real-world images for compositional generation, in contrast to other slot-based generative methods in the literature. The project page can be found at https://kaanakan.github.io/SlotAdapt/.
Authors: Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Cand\`es, Tatsunori Hashimoto
Abstract: Test-time scaling is a promising new approach to language modeling that uses extra test-time compute to improve performance. Recently, OpenAI's o1 model showed this capability but did not publicly share its methodology, leading to many replication efforts. We seek the simplest approach to achieve test-time scaling and strong reasoning performance. First, we curate a small dataset s1K of 1,000 questions paired with reasoning traces relying on three criteria we validate through ablations: difficulty, diversity, and quality. Second, we develop budget forcing to control test-time compute by forcefully terminating the model's thinking process or lengthening it by appending "Wait" multiple times to the model's generation when it tries to end. This can lead the model to double-check its answer, often fixing incorrect reasoning steps. After supervised finetuning the Qwen2.5-32B-Instruct language model on s1K and equipping it with budget forcing, our model s1-32B exceeds o1-preview on competition math questions by up to 27% (MATH and AIME24). Further, scaling s1-32B with budget forcing allows extrapolating beyond its performance without test-time intervention: from 50% to 57% on AIME24. Our model, data, and code are open-source at https://github.com/simplescaling/s1
Authors: Shubham Malhotra, Fnu Yashu, Muhammad Saqib, Dipkumar Mehta, Jagdish Jangid, Sachin Dixit
Abstract: This report investigates the application of deep reinforcement learning (DRL) algorithms for dynamic resource allocation in wireless communication systems. An environment that includes a base station, multiple antennas, and user equipment is created. Using the RLlib library, various DRL algorithms such as Deep Q-Network (DQN) and Proximal Policy Optimization (PPO) are then applied. These algorithms are compared based on their ability to optimize resource allocation, focusing on the impact of different learning rates and scheduling policies. The findings demonstrate that the choice of algorithm and learning rate significantly influences system performance, with DRL providing more efficient resource allocation compared to traditional methods.
Authors: Juno Kim, Denny Wu, Jason Lee, Taiji Suzuki
Abstract: A key paradigm to improve the reasoning capabilities of large language models (LLMs) is to allocate more inference-time compute to search against a verifier or reward model. This process can then be utilized to refine the pretrained model or distill its reasoning patterns into more efficient models. In this paper, we study inference-time compute by viewing chain-of-thought (CoT) generation as a metastable Markov process: easy reasoning steps (e.g., algebraic manipulations) form densely connected clusters, while hard reasoning steps (e.g., applying a relevant theorem) create sparse, low-probability edges between clusters, leading to phase transitions at longer timescales. Under this framework, we prove that implementing a search protocol that rewards sparse edges improves CoT by decreasing the expected number of steps to reach different clusters. In contrast, we establish a limit on reasoning capability when the model is restricted to local information of the pretrained graph. We also show that the information gained by search can be utilized to obtain a better reasoning model: (1) the pretrained model can be directly finetuned to favor sparse edges via policy gradient methods, and moreover (2) a compressed metastable representation of the reasoning dynamics can be distilled into a smaller, more efficient model.
Authors: Andrew Van Horn, Lauryn Smith, Mahamad Mahmoud, Michael McMaster, Clara Pinchbeck, Ina Martin, Andrew Lininger, Anthony Ingrisano, Adam Lowe, Carlos Bayod, Elizabeth Bolman, Kenneth Singer, Michael Hinczewski
Abstract: The history of art has seen significant shifts in the manner in which artworks are created, making understanding of creative processes a central question in technical art history. In the Renaissance and Early Modern period, paintings were largely produced by master painters directing workshops of apprentices who often contributed to projects. The masters varied significantly in artistic and managerial styles, meaning different combinations of artists and implements might be seen both between masters and within workshops or even individual canvases. Information on how different workshops were managed and the processes by which artworks were created remains elusive. Machine learning methods have potential to unearth new information about artists' creative processes by extending the analysis of brushwork to a microscopic scale. Analysis of workshop paintings, however, presents a challenge in that documentation of the artists and materials involved is sparse, meaning external examples are not available to train networks to recognize their contributions. Here we present a novel machine learning approach we call pairwise assignment training for classifying heterogeneity (PATCH) that is capable of identifying individual artistic practice regimes with no external training data, or "ground truth." The method achieves unsupervised results by supervised means, and outperforms both simple statistical procedures and unsupervised machine learning methods. We apply this method to two historical paintings by the Spanish Renaissance master, El Greco: The Baptism of Christ and Christ on the Cross with Landscape, and our findings regarding the former potentially challenge previous work that has assigned the painting to workshop members. Further, the results of our analyses create a measure of heterogeneity of artistic practice that can be used to characterize artworks across time and space.
Authors: Yuri Chervonyi, Trieu H. Trinh, Miroslav Ol\v{s}\'ak, Xiaomeng Yang, Hoang Nguyen, Marcelo Menegali, Junehyuk Jung, Vikas Verma, Quoc V. Le, Thang Luong
Abstract: We present AlphaGeometry2, a significantly improved version of AlphaGeometry introduced in Trinh et al. (2024), which has now surpassed an average gold medalist in solving Olympiad geometry problems. To achieve this, we first extend the original AlphaGeometry language to tackle harder problems involving movements of objects, and problems containing linear equations of angles, ratios, and distances. This, together with support for non-constructive problems, has markedly improved the coverage rate of the AlphaGeometry language on International Math Olympiads (IMO) 2000-2024 geometry problems from 66% to 88%. The search process of AlphaGeometry2 has also been greatly improved through the use of Gemini architecture for better language modeling, and a novel knowledge-sharing mechanism that enables effective communication between search trees. Together with further enhancements to the symbolic engine and synthetic data generation, we have significantly boosted the overall solving rate of AlphaGeometry2 to 84% for $\textit{all}$ geometry problems over the last 25 years, compared to 54% previously. AlphaGeometry2 was also part of the system that achieved silver-medal standard at IMO 2024 https://dpmd.ai/imo-silver. Last but not least, we report progress towards using AlphaGeometry2 as a part of a fully automated system that reliably solves geometry problems directly from natural language input.
Authors: Shunxian Gu, Chaoqun You, Deke Guo, Zhihao Qu, Bangbang Ren, Zaipeng Xie, Lailong Luo
Abstract: Analytic Federated Learning (AFL) is an enhanced gradient-free federated learning (FL) paradigm designed to accelerate training by updating the global model in a single step with closed-form least-square (LS) solutions. However, the obtained global model suffers performance degradation across clients with heterogeneous data distribution. Meta-learning is a common approach to tackle this problem by delivering personalized local models for individual clients. Yet, integrating meta-learning with AFL presents significant challenges: First, conventional AFL frameworks cannot support deep neural network (DNN) training which can influence the fast adaption capability of meta-learning for complex FL tasks. Second, the existing meta-learning method requires gradient information, which is not involved in AFL. To overcome the first challenge, we propose an AFL framework, namely FedACnnL, in which a layer-wise DNN collaborative training method is designed by modeling the training of each layer as a distributed LS problem. For the second challenge, we further propose an analytic personalized federated meta-learning framework, namely pFedACnnL. It generates a personalized model for each client by analytically solving a local objective which bridges the gap between the global model and the individual data distribution. FedACnnL is theoretically proven to require significantly shorter training time than the conventional FL frameworks on DNN training while the reduction ratio is $83\%\sim99\%$ in the experiment. Meanwhile, pFedACnnL excels at test accuracy with the vanilla FedACnnL by $4\%\sim8\%$ and it achieves state-of-the-art (SOTA) model performance in most cases of convex and non-convex settings compared with previous SOTA frameworks.
Authors: Matthew Finlayson, Ilia Kulikov, Daniel M. Bikel, Barlas Oguz, Xilun Chen, Aasish Pappu
Abstract: Large language models (LLMs) often struggle with knowledge intensive NLP tasks, such as answering "Who won the latest World Cup?" because the knowledge they learn during training may be insufficient or outdated. Conditioning generation on retrieved documents -- a technique known as retrieval augmented generation (RAG) -- mitigates these shortcomings by allowing the model to leverage in-context information. Practitioners can improve LLM RAG performance by fine-tuning on retrieval-augmented instructions, but must beware that this can cause undesirable model behaviors like hallucinations. We attribute this degradation to the fact that the training data is likely to be out-of-distribution for the model and may suffer from quality issues, such as misalignment between retrievals and target responses (since retrievals are frequently added post-hoc). We propose a recipe for training RAG-enabled LLMs using self-generated demonstrations, thereby avoiding training on out-of-distribution text and integrating retrievals into the LLM responses. We evaluate our method on knowledge intensive question answering (QA) tasks and show that our method teaches LLMs to properly handle in-context retrievals and abstain from questions it will likely get wrong. Compared to conventional RA-IT methods, our method prevents model degradation in non-RAG settings while exhibiting superior QA performance.
Authors: Farbod Faraji, Maryam Reza
Abstract: Plasma systems exhibit complex multiscale dynamics, resolving which poses significant challenges for conventional numerical simulations. Machine learning (ML) offers an alternative by learning data-driven representations of these dynamics. Yet existing ML time-stepping models suffer from error accumulation, instability, and limited long-term forecasting horizons. This paper demonstrates the application of a hierarchical multiscale neural network architecture for autonomous plasma forecasting. The framework integrates multiple neural networks trained across different temporal scales to capture both fine-scale and large-scale behaviors while mitigating compounding error in recursive evaluation. Fine-scale networks accurately resolve fast-evolving features, while coarse-scale networks provide broader temporal context, reducing the frequency of recursive updates and limiting the accumulation of small prediction errors over time. We first evaluate the method using canonical nonlinear dynamical systems and compare its performance against classical single-scale neural networks. The results demonstrate that single-scale neural networks experience rapid divergence due to recursive error accumulation, whereas the multiscale approach improves stability and extends prediction horizons. Next, our ML model is applied to two plasma configurations of high scientific and applied significance, demonstrating its ability to preserve spatial structures and capture multiscale plasma dynamics. By leveraging multiple time-stepping resolutions, the applied framework is shown to outperform conventional single-scale networks for the studied plasma test cases. The results of this work position the hierarchical multiscale neural network as a promising tool for efficient plasma forecasting and digital twin applications.
Authors: Shao Zhang, Xihuai Wang, Wenhao Zhang, Chaoran Li, Junru Song, Tingyu Li, Lin Qiu, Xuezhi Cao, Xunliang Cai, Wen Yao, Weinan Zhang, Xinbing Wang, Ying Wen
Abstract: Agents built on large language models (LLMs) have excelled in turn-by-turn human-AI collaboration but struggle with simultaneous tasks requiring real-time interaction. Latency issues and the challenge of inferring variable human strategies hinder their ability to make autonomous decisions without explicit instructions. Through experiments with current independent System 1 and System 2 methods, we validate the necessity of using Dual Process Theory (DPT) in real-time tasks. We propose DPT-Agent, a novel language agent framework that integrates System 1 and System 2 for efficient real-time simultaneous human-AI collaboration. DPT-Agent's System 1 uses a Finite-state Machine (FSM) and code-as-policy for fast, intuitive, and controllable decision-making. DPT-Agent's System 2 integrates Theory of Mind (ToM) and asynchronous reflection to infer human intentions and perform reasoning-based autonomous decisions. We demonstrate the effectiveness of DPT-Agent through further experiments with rule-based agents and human collaborators, showing significant improvements over mainstream LLM-based frameworks. DPT-Agent can effectively help LLMs convert correct slow thinking and reasoning into executable actions, thereby improving performance. To the best of our knowledge, DPT-Agent is the first language agent framework that achieves successful real-time simultaneous human-AI collaboration autonomously. Code of DPT-Agent can be found in https://github.com/sjtu-marl/DPT-Agent.
Authors: Wei Dai, Steven Wang, Jun Liu
Abstract: Efficient evaluation of three-dimensional (3D) medical images is crucial for diagnostic and therapeutic practices in healthcare. Recent years have seen a substantial uptake in applying deep learning and computer vision to analyse and interpret medical images. Traditional approaches, such as convolutional neural networks (CNNs) and vision transformers (ViTs), face significant computational challenges, prompting the need for architectural advancements. Recent efforts have led to the introduction of novel architectures like the ``Mamba'' model as alternative solutions to traditional CNNs or ViTs. The Mamba model excels in the linear processing of one-dimensional data with low computational demands. However, Mamba's potential for 3D medical image analysis remains underexplored and could face significant computational challenges as the dimension increases. This manuscript presents MobileViM, a streamlined architecture for efficient segmentation of 3D medical images. In the MobileViM network, we invent a new dimension-independent mechanism and a dual-direction traversing approach to incorporate with a vision-Mamba-based framework. MobileViM also features a cross-scale bridging technique to improve efficiency and accuracy across various medical imaging modalities. With these enhancements, MobileViM achieves segmentation speeds exceeding 90 frames per second (FPS) on a single graphics processing unit (i.e., NVIDIA RTX 4090). This performance is over 24 FPS faster than the state-of-the-art deep learning models for processing 3D images with the same computational resources. In addition, experimental evaluations demonstrate that MobileViM delivers superior performance, with Dice similarity scores reaching 92.72%, 86.69%, 80.46%, and 77.43% for PENGWIN, BraTS2024, ATLAS, and Toothfairy2 datasets, respectively, which significantly surpasses existing models.
Authors: Guanzheng Chen, Xin Li, Michael Qizhe Shieh, Lidong Bing
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities through pretraining and alignment. However, superior short-context LLMs may underperform in long-context scenarios due to insufficient long-context alignment. This alignment process remains challenging due to the impracticality of human annotation for extended contexts and the difficulty in balancing short- and long-context performance. To address these challenges, we introduce LongPO, that enables short-context LLMs to self-evolve to excel on long-context tasks by internally transferring short-context capabilities. LongPO harnesses LLMs to learn from self-generated short-to-long preference data, comprising paired responses generated for identical instructions with long-context inputs and their compressed short-context counterparts, respectively. This preference reveals capabilities and potentials of LLMs cultivated during short-context alignment that may be diminished in under-aligned long-context scenarios. Additionally, LongPO incorporates a short-to-long KL constraint to mitigate short-context performance decline during long-context alignment. When applied to Mistral-7B-Instruct-v0.2 from 128K to 512K context lengths, LongPO fully retains short-context performance and largely outperforms naive SFT and DPO in both long- and short-context tasks. Specifically, LongPO-trained models can achieve results on long-context benchmarks comparable to, or even surpassing, those of superior LLMs (e.g., GPT-4-128K) that involve extensive long-context annotation and larger parameter scales. Our code is available at https://github.com/DAMO-NLP-SG/LongPO.
Authors: Amirreza Neshaei Moghaddam, Alex Olshevsky, Bahman Gharesifard
Abstract: Inspired by REINFORCE, we introduce a novel receding-horizon algorithm for the Linear Quadratic Regulator (LQR) problem with unknown parameters. Unlike prior methods, our algorithm avoids reliance on two-point gradient estimates while maintaining the same order of sample complexity. Furthermore, it eliminates the restrictive requirement of starting with a stable initial policy, broadening its applicability. Beyond these improvements, we introduce a refined analysis of error propagation through the contraction of the Riemannian distance over the Riccati operator. This refinement leads to a better sample complexity and ensures improved convergence guarantees. Numerical simulations validate the theoretical results, demonstrating the method's practical feasibility and performance in realistic scenarios.
Authors: Hamid Moradi-Kamali, Mohammad-Hossein Rajabi-Ghozlou, Mahdi Ghazavi, Ali Soltani, Amirreza Sattarzadeh, Reza Entezari-Maleki
Abstract: Financial Sentiment Analysis (FSA) traditionally relies on human-annotated sentiment labels to infer investor sentiment and forecast market movements. However, inferring the potential market impact of words based on their human-perceived intentions is inherently challenging. We hypothesize that the historical market reactions to words, offer a more reliable indicator of their potential impact on markets than subjective sentiment interpretations by human annotators. To test this hypothesis, a market-derived labeling approach is proposed to assign tweet labels based on ensuing short-term price trends, enabling the language model to capture the relationship between textual signals and market dynamics directly. A domain-specific language model was fine-tuned on these labels, achieving up to an 11% improvement in short-term trend prediction accuracy over traditional sentiment-based benchmarks. Moreover, by incorporating market and temporal context through prompt-tuning, the proposed context-aware language model demonstrated an accuracy of 89.6% on a curated dataset of 227 impactful Bitcoin-related news events with significant market impacts. Aggregating daily tweet predictions into trading signals, our method outperformed traditional fusion models (which combine sentiment-based and price-based predictions). It challenged the assumption that sentiment-based signals are inferior to price-based predictions in forecasting market movements. Backtesting these signals across three distinct market regimes yielded robust Sharpe ratios of up to 5.07 in trending markets and 3.73 in neutral markets. Our findings demonstrate that language models can serve as effective short-term market predictors. This paradigm shift underscores the untapped capabilities of language models in financial decision-making and opens new avenues for market prediction applications.
Authors: Giuseppe Paolo, Abdelhakim Benechehab, Hamza Cherkaoui, Albert Thomas, Bal\'azs K\'egl
Abstract: Hierarchical organization is fundamental to biological systems and human societies, yet artificial intelligence systems often rely on monolithic architectures that limit adaptability and scalability. Current hierarchical reinforcement learning (HRL) approaches typically restrict hierarchies to two levels or require centralized training, which limits their practical applicability. We introduce TAME Agent Framework (TAG), a framework for constructing fully decentralized hierarchical multi-agent systems.TAG enables hierarchies of arbitrary depth through a novel LevelEnv concept, which abstracts each hierarchy level as the environment for the agents above it. This approach standardizes information flow between levels while preserving loose coupling, allowing for seamless integration of diverse agent types. We demonstrate the effectiveness of TAG by implementing hierarchical architectures that combine different RL agents across multiple levels, achieving improved performance over classical multi-agent RL baselines on standard benchmarks. Our results show that decentralized hierarchical organization enhances both learning speed and final performance, positioning TAG as a promising direction for scalable multi-agent systems.
Authors: Yepeng Weng, Dianwen Mei, Huishi Qiu, Xujie Chen, Li Liu, Jiang Tian, Zhongchao Shi
Abstract: Speculative decoding is a powerful technique that accelerates Large Language Model (LLM) inference by leveraging a lightweight speculative draft model. However, existing designs suffers in performance due to misalignment between training and inference. Recent methods have tried to solve this issue by adopting a multi-step training strategy, but the complex inputs of different training steps make it harder for the draft model to converge. To address this, we propose CORAL, a novel framework that improves both accuracy and efficiency in speculative drafting. CORAL introduces Cross-Step Representation Alignment, a method that enhances consistency across multiple training steps, significantly improving speculative drafting performance. Additionally, we identify the LM head as a major bottleneck in the inference speed of the draft model. We introduce a weight-grouping mechanism that selectively activates a subset of LM head parameters during inference, substantially reducing the latency of the draft model. We evaluate CORAL on three LLM families and three benchmark datasets, achieving speedup ratios of 2.50x-4.07x, outperforming state-of-the-art methods such as EAGLE-2 and HASS. Our results demonstrate that CORAL effectively mitigates training-inference misalignment and delivers significant speedup for modern LLMs with large vocabularies.
Authors: Minyeong Hwang, Ziseok Lee, Kwang-Soo Kim, Kyungsu Kim, Eunho Yang
Abstract: Linker generation is critical in drug discovery applications such as lead optimization and PROTAC design, where molecular fragments are assembled into diverse drug candidates. Existing methods fall into PC-Free and PC-Aware categories based on their use of 3D point clouds (PC). PC-Free models prioritize diversity but suffer from lower validity due to overlooking PC constraints, while PC-Aware models ensure higher validity but restrict diversity by enforcing strict PC constraints. To overcome these trade-offs without additional training, we propose HybridLinker, a framework that enhances PC-Aware inference by providing diverse bonding topologies from a pretrained PC-Free model as guidance. At its core, we propose LinkerDPS, the first diffusion posterior sampling (DPS) method operating across PC-Free and PC-Aware spaces, bridging molecular topology with 3D point clouds via an energy-inspired function. By transferring the diverse sampling distribution of PC-Free models into the PC-Aware distribution, HybridLinker significantly and consistently surpasses baselines, improving both validity and diversity in foundational molecular design and applied property optimization tasks, establishing a new DPS framework in the molecular and graph domains beyond imaging.
Authors: Mingyuan Sun, Zheng Fang, Jiaxu Wang, Junjie Jiang, Delei Kong, Chenming Hu, Yuetong Fang, Renjing Xu
Abstract: The increasing complexity and parameter count of Convolutional Neural Networks (CNNs) and Transformers pose challenges in terms of computational efficiency and resource demands. Pruning has been identified as an effective strategy to address these challenges by removing redundant elements such as neurons, channels, or connections, thereby enhancing computational efficiency without heavily compromising performance. This paper builds on the foundational work of Optimal Brain Damage (OBD) by advancing the methodology of parameter importance estimation using the Hessian matrix. Unlike previous approaches that rely on approximations, we introduce Optimal Brain Apoptosis (OBA), a novel pruning method that calculates the Hessian-vector product value directly for each parameter. By decomposing the Hessian matrix across network layers and identifying conditions under which inter-layer Hessian submatrices are non-zero, we propose a highly efficient technique for computing the second-order Taylor expansion of parameters. This approach allows for a more precise pruning process, particularly in the context of CNNs and Transformers, as validated in our experiments including VGG19, ResNet32, ResNet50, and ViT-B/16 on CIFAR10, CIFAR100 and Imagenet datasets. Our code is available at https://github.com/NEU-REAL/OBA.
Authors: Yi Feng, Xiao Wang, Tian Xie
Abstract: We consider nonconvex optimization problem over simplex, and more generally, a product of simplices. We provide an algorithm, Langevin Multiplicative Weights Update (LMWU) for solving global optimization problems by adding a noise scaling with the non-Euclidean geometry in the simplex. Non-convex optimization has been extensively studied by machine learning community due to its application in various scenarios such as neural network approximation and finding Nash equilibrium. Despite recent progresses on provable guarantee of escaping and avoiding saddle point (convergence to local minima) and global convergence of Langevin gradient based method without constraints, the global optimization with constraints is less studied. We show that LMWU algorithm is provably convergent to interior global minima with a non-asymptotic convergence analysis. We verify the efficiency of the proposed algorithm in real data set from polynomial portfolio management, where optimization of a highly non-linear objective function plays a crucial role.
Authors: Shulai Zhang, Ningxin Zheng, Haibin Lin, Ziheng Jiang, Wenlei Bao, Chengquan Jiang, Qi Hou, Weihao Cui, Size Zheng, Li-Wen Chang, Quan Chen, Xin Liu
Abstract: Mixture-of-experts (MoE) has been extensively employed to scale large language models to trillion-plus parameters while maintaining a fixed computational cost. The development of large MoE models in the distributed scenario encounters the problem of large communication overhead. The inter-device communication of a MoE layer can occupy 47% time of the entire model execution with popular models and frameworks. Therefore, existing methods suggest the communication in a MoE layer to be pipelined with the computation for overlapping. However, these coarse grained overlapping schemes introduce a notable impairment of computational efficiency and the latency concealing is sub-optimal. To this end, we present COMET, an optimized MoE system with fine-grained communication-computation overlapping. Leveraging data dependency analysis and task rescheduling, COMET achieves precise fine-grained overlapping of communication and computation. Through adaptive workload assignment, COMET effectively eliminates fine-grained communication bottlenecks and enhances its adaptability across various scenarios. Our evaluation shows that COMET accelerates the execution of a single MoE layer by $1.96\times$ and for end-to-end execution, COMET delivers a $1.71\times$ speedup on average. COMET has been adopted in the production environment of clusters with ten-thousand-scale of GPUs, achieving savings of millions of GPU hours.