NeurIPS 2025 Papers

We study the finite-horizon Restless Multi-Armed Bandit (RMAB) problem with $N$ homogeneous arms. Prior work has shown that when an RMAB satisfies a non-degeneracy condition, Linear-Programming-based (LP-based) policies derived from the fluid approximation, which captures the mean dynamics of the system, achieve an exponentially small optimality gap. However, it is common for RMABs to be degenerate, in which case LP-based policies can result in a $\Theta(1/\sqrt{N})$ optimality gap per arm. In this paper, we propose a novel Stochastic-Programming-based (SP-based) policy that, under a uniqueness assumption, achieves an $\tilde{\mathcal{O}}(1/N)$ optimality gap for degenerate RMABs. Our approach is based on the construction of a Gaussian stochastic system that captures not only the mean but also the variance of the RMAB dynamics, resulting in a more accurate approximation than the fluid approximation. We then solve a stochastic program for this system to obtain our policy. This is the first result to establish an $\tilde{\mathcal{O}}(1/N)$ optimality gap for degenerate RMABs.

Neural networks generally prefer simple and easy-to-learn features. When these features are spuriously correlated with the labels, the network's performance can suffer, particularly for underrepresented classes or concepts. Self-supervised representation learning methods, such as contrastive learning, are especially prone to this issue, often resulting in worse performance on downstream tasks. We identify a key spectral signature of this failure: early reliance on dominant singular modes of the learned feature matrix. To mitigate this, we propose a novel framework that promotes a uniform eigenspectrum of the feature covariance matrix, encouraging diverse and semantically rich representations. Our method operates in a fully self-supervised setting, without relying on ground-truth labels or any additional information. Empirical results on SimCLR and SimSiam demonstrate consistent gains in robustness and transfer performance, suggesting broad applicability across self-supervised learning paradigms. Code: https://github.com/NaghmehGh/SpuriousCorrelation_SSRL

In this paper, we explore the universal properties underlying causal inference by formulating it in terms of a topos. More concretely, we introduce topos causal models (TCMs), a strict generalization of the popular structural causal models (SCMs). A topos category has several properties that make it attractive: a general theory for how to combine local functions that define ``independent causal mechanisms" into a consistent global function building on the theory of sheaves in a topos; a generic way to define causal interventions using a subobject classifier in a topos category; and finally, an internal logical language for causal and counterfactual reasoning that emerges from the topos itself. A striking characteristic of subobject classifiers is that they induce an intuitionistic logic, whose semantics is based on the partially ordered lattice of subobjects. We show that the underlying subobject classifier for causal inference is not Boolean in general, but forms a Heyting algebra. We define the internal Mitchell-B\'enabou language, a typed local set theory, associated with causal models, and its associated Kripke-Joyal intuitionistic semantics. We prove a universal property of TCM, namely that any causal functor mapping decomposable structure to probabilistic semantics factors uniquely through a TCM representation.

While Generative AI rapidly advances in various domains, generating truly creative, aesthetic, and counter-intuitive outputs remains a challenge. This paper presents an approach to tackle these difficulties in the domain of chess puzzles. We start by benchmarking Generative AI architectures, and then introduce an RL framework with novel rewards based on chess engine search statistics to overcome some of those shortcomings. The rewards are designed to enhance a puzzle's uniqueness, counter-intuitiveness, diversity, and realism. Our RL approach dramatically increases counter-intuitive puzzle generation by 10x, from 0.22\% (supervised) to 2.5\%, surpassing existing dataset rates (2.1\%) and the best Lichess-trained model (0.4\%). Our puzzles meet novelty and diversity benchmarks, retain aesthetic themes, and are rated by human experts as more creative, enjoyable, and counter-intuitive than composed book puzzles, even approaching classic compositions. Our final outcome is a curated booklet of these novel AI-generated puzzles, which is acknowledged for creativity by three world-renowned experts.

Reinforcement Learning from Human Feedback (RLHF) has emerged as a pivotal technique for large language model (LLM) alignment. This paper studies the setting of online RLHF and focuses on improving its sample efficiency. All existing algorithms for online RLHF, whether doing passive exploration or active exploration, suffer from a sample complexity that scales exponentially with the range of the reward function. This statistical inefficiency hinders their effectiveness in scenarios with heavily skewed preferences, e.g. questions with objectively correct answers. To address this, we introduce Self-Exploring Preference-Incentive Online Preference Optimization (SE-POPO), an online RLHF algorithm that for the first time achieves a sample complexity that scales polynomially with the reward range, answering an open problem raised by Xie et al. [2024]. Theoretically, we demonstrate that the sample complexity of SE-POPO dominates that of existing exploration algorithms. Empirically, our systematic evaluation confirms that SE-POPO is more sample-efficient than both exploratory and non-exploratory baselines, in two primary application scenarios of RLHF as well as on public benchmarks, marking a significant step forward in RLHF algorithm design.

In multi-armed bandits with network interference (MABNI), the action taken by one node can influence the rewards of others, creating complex interdependence. While existing research on MABNI largely concentrates on minimizing regret, it often overlooks the crucial concern that an excessive emphasis on the optimal arm can undermine the inference accuracy for sub-optimal arms. Although initial efforts have been made to address this trade-off in single-unit scenarios, these challenges have become more pronounced in the context of MABNI. In this paper, we establish, for the first time, a theoretical Pareto frontier characterizing the trade-off between regret minimization and inference accuracy in adversarial (design-based) MABNI. We further introduce an anytime-valid asymptotic confidence sequence along with a corresponding algorithm, $\texttt{EXP3-N-CS}$, specifically designed to balance the trade-off between regret minimization and inference accuracy in this setting.

Light field microscopy (LFM) has become an emerging tool in neuroscience for large-scale neural imaging in vivo, with XLFM (eXtended Light Field Microscopy) notable for its single-exposure volumetric imaging, broad field of view, and high temporal resolution. However, learning-based 3D reconstruction in XLFM remains underdeveloped due to two core challenges: the absence of standardized datasets and the lack of methods that can efficiently model its angular–spatial structure while remaining physically grounded. We address these challenges by introducing three key contributions. First, we construct the XLFM-Zebrafish benchmark, a large-scale dataset and evaluation suite for XLFM reconstruction. Second, we propose Masked View Modeling for Light Fields (MVM-LF), a self-supervised task that learns angular priors by predicting occluded views, improving data efficiency. Third, we formulate the Optical Rendering Consistency Loss (ORC Loss), a differentiable rendering constraint that enforces alignment between predicted volumes and their PSF-based forward projections. On the XLFM-Zebrafish benchmark, our method improves PSNR by 7.7\% over state-of-the-art baselines. Code and datasets are publicly available at: https://github.com/hefengcs/XLFM-Former.

Dependency parsing is the task of inferring natural language structure, often approached by modeling word interactions via attention through biaffine scoring. This mechanism works like self-attention in Transformers, where scores are calculated for every pair of words in a sentence. However, unlike Transformer attention, biaffine scoring does not use normalization prior to taking the softmax of the scores. In this paper, we provide theoretical evidence and empirical results revealing that a lack of normalization necessarily results in overparameterized parser models, where the extra parameters compensate for the sharp softmax outputs produced by high variance inputs to the biaffine scoring function. We argue that biaffine scoring can be made substantially more efficient by performing score normalization. We conduct experiments on semantic and syntactic dependency parsing in multiple languages, along with latent graph inference on non-linguistic data, using various settings of a $k$-hop parser. We train $N$-layer stacked BiLSTMs and evaluate the parser's performance with and without normalizing biaffine scores. Normalizing allows us to achieve state-of-the-art performance with fewer samples and trainable parameters. Code: https://github.com/paolo-gajo/EfficientSDP

Traditional reinforcement learning (RL) algorithms face significant limitations in handling long-term tasks with sparse rewards. Recent advancements have leveraged large language models (LLMs) to enhance RL by utilizing their world knowledge for task planning and reward generation. However, planning-based approaches often depend on pre-defined skill libraries and fail to optimize low-level control policies, while reward-based methods require extensive human feedback or exhaustive searching due to the complexity of tasks. In this paper, we propose the Progress Reward Model for RL (PRM4RL), a novel framework that integrates task planning and dense reward to enhance RL. For high-level planning, a complex task is decomposed into a series of simple manageable subtasks, with a subtask-oriented, fine-grained progress function designed to monitor task execution progress. For low-level reward generation, inspired by potential-based reward shaping, we use the progress function to construct a Progress Reward Model (PRM), providing theoretically grounded optimality and convergence guarantees, thereby enabling effective policy optimization. Experimental results on robotics control tasks demonstrate that our approach outperforms both LLM-based planning and reward methods, achieving state-of-the-art performance.

Modern code generation has made significant strides in functional correctness and execution efficiency. However, these systems often overlook a critical dimension in real-world software development: \textit{maintainability}. To handle dynamic requirements with minimal rework, we propose \textbf{MaintainCoder} as a pioneering solution. It integrates the Waterfall model, design patterns, and multi-agent collaboration to systematically enhance cohesion, reduce coupling, achieving clear responsibility boundaries and better maintainability. We also introduce \textbf{MaintainBench}, a benchmark comprising requirement changes and novel dynamic metrics on maintenance efforts. Experiments demonstrate that existing code generation methods struggle to meet maintainability standards when requirements evolve. In contrast, MaintainCoder improves dynamic maintainability metrics by more than 60\% with even higher correctness of initial codes. Furthermore, while static metrics fail to accurately reflect maintainability and even contradict each other, our proposed dynamic metrics exhibit high consistency. Our work not only provides the foundation for maintainable code generation, but also highlights the need for more realistic and comprehensive code generation research. Resources: https://github.com/IAAR-Shanghai/MaintainCoder.

We investigate the concept of algorithmic replicability introduced by Impagliazzo et al.(2022) in an online setting. In our model, the input sequence received by the online learner is generated from time-varying distributions chosen by an adversary (obliviously). Our objective is to design low-regret online algorithms that, with high probability, produce the \emph{exact same sequence} of actions when run on two independently sampled input sequences generated as described above. We refer to such algorithms as adversarially replicable. Previous works explored replicability in the online setting under inputs generated independently from a fixed distribution; we term this notion as iid-replicability. Our model generalizes to capture both adversarial and iid input sequences, as well as their mixtures, which can be modeled by setting certain distributions as point-masses. We demonstrate adversarially replicable online learning algorithms for online linear optimization and the experts problem that achieve sub-linear regret. Additionally, we propose a general framework for converting an online learner into an adversarially replicable one within our setting, bounding the new regret in terms of the original algorithm’s regret. We also present a nearly optimal (in terms of regret) iid-replicable online algorithm for the experts problem, highlighting the distinction between the iid and adversarial notions of replicability. Finally, we establish lower bounds on the regret (in terms of the replicability parameter and time) that any replicable online algorithm must incur.

Current evaluation paradigms for large language models (LLMs) represent a critical blind spot in AI research—relying on opaque numerical metrics that conceal fundamental limitations in spatial reasoning while providing no intuitive understanding of model capabilities. This deficiency creates a dangerous disconnect between reported performance and practical abilities, particularly for applications requiring physical world understanding. We introduce LTD-Bench, a breakthrough benchmark that transforms LLM evaluation from abstract scores to directly observable visual outputs by requiring models to generate drawings through dot matrices or executable code. This approach makes spatial reasoning limitations immediately apparent even to non-experts, bridging the fundamental gap between statistical performance and intuitive assessment. LTD-Bench implements a comprehensive methodology with complementary generation tasks (testing spatial imagination) and recognition tasks (assessing spatial perception) across three progressively challenging difficulty levels, methodically evaluating both directions of the critical language-spatial mapping. Our extensive experiments with state-of-the-art models expose an alarming capability gap: even LLMs achieving impressive results on traditional benchmarks demonstrate profound deficiencies in establishing bidirectional mappings between language and spatial concepts—a fundamental limitation that undermines their potential as genuine world models. Furthermore, LTD-Bench's visual outputs enable powerful diagnostic analysis, offering a potential approach to investigate model similarity. Our dataset and codes are available at https://github.com/walktaster/LTD-Bench.

Machine unlearning has emerged as a prevalent technical solution for selectively removing unwanted knowledge absorbed during pre-training, without requiring full retraining. While recent unlearning techniques can effectively remove undesirable content without severely compromising performance on standard benchmarks, we find that they may inadvertently create ``knowledge holes''---unintended losses of benign knowledge that standard benchmarks fail to capture. To probe where unlearned models reveal knowledge holes, we propose a test case generation framework that explores both immediate neighbors of unlearned content and broader areas of potential failures. Our evaluation demonstrates significant hidden costs of unlearning: up to 98.7\% of the test cases yield irrelevant or nonsensical responses from unlearned models, despite being answerable by the pretrained model. These findings necessitate rethinking the conventional approach to evaluating knowledge preservation in unlearning, moving beyond standard, static benchmarks.

In this paper, we focus on Novel Class Discovery for Point Cloud Segmentation (3D-NCD), aiming to learn a model that can segment unlabeled (novel) 3D classes using only the supervision from labeled (base) 3D classes. The key to this task is to setup the exact correlations between the point representations and their base class labels, as well as the representation correlations between the points from base and novel classes. A coarse or statistical correlation learning may lead to the confusion in novel class inference. lf we impose a causal relationship as a strong correlated constraint upon the learning process, the essential point cloud representations that accurately correspond to the classes should be uncovered. To this end, we introduce a structural causal model (SCM) to re-formalize the 3D-NCD problem and propose a new method, i.e., Joint Learning of Causal Representation and Reasoning. Specifically, we first analyze hidden confounders in the base class representations and the causal relationships between the base and novel classes through SCM. We devise a causal representation prototype that eliminates confounders to capture the causal representations of base classes. A graph structure is then used to model the causal relationships between the base classes' causal representation prototypes and the novel class prototypes, enabling causal reasoning from base to novel classes. Extensive experiments and visualization results on 3D and 2D NCD semantic segmentation demonstrate the superiorities of our method.

Despite their consistent performance improvements, cross-modal retrieval models (e.g., CLIP) show degraded performances with retrieving keys composed of fused image-text modality (e.g., Wikipedia pages with both images and text). To address this critical challenge, multimodal retrieval has been recently explored to develop a unified single retrieval model capable of retrieving keys across diverse modality combinations. A common approach involves constructing new composed sets of image-text triplets (e.g., retrieving a pair of image and text given a query image). However, such an approach requires careful curation to ensure the dataset quality and fails to generalize to unseen modality combinations. To overcome these limitations, this paper proposes Generalized Contrastive Learning (GCL), a novel loss formulation that improves multimodal retrieval performance without the burdensome need for new dataset curation. Specifically, GCL operates by enforcing contrastive learning across all modalities within a mini-batch, utilizing existing image-caption paired datasets to learn a unified representation space. We demonstrate the effectiveness of GCL by showing consistent performance improvements on off-the-shelf multimodal retrieval models (e.g., VISTA, CLIP, and TinyCLIP) using the M-BEIR, MMEB, and CoVR benchmarks.

Reverse-Kullback-Leibler (KL) regularization has emerged to be a predominant technique to enhance policy optimization in reinforcement learning (RL) and reinforcement learning from human feedback (RLHF), which forces the learned policy to stay close to a reference policy. While the effectiveness of KL-regularization has been empirically demonstrated in various practical scenarios, current theoretical analyses of KL-regularized RLHF still yield the same $\mathcal{O}(1 / \epsilon^2)$ sample complexity as ones without KL-regularization. To understand the fundamental distinction between objectives with KL-regularization and ones without KL-regularization, we are the first to theoretically demonstrate the power of KL-regularization by providing a sharp analysis for KL-regularized contextual bandits and RLHF, revealing an $\mathcal{O}(1 / \epsilon)$ sample complexity when $\epsilon$ is sufficiently small. We also prove matching lower bounds for both settings. More specifically, we study how the coverage of the reference policy affects the sample complexity of KL-regularized online contextual bandits and RLHF. We show that with sufficient coverage from the reference policy, a simple two-stage mixed sampling algorithm can achieve an $\mathcal{O}(1 / \epsilon)$ sample complexity with only an additive dependence on the coverage coefficient, thus proving the benefits of online data even without explicit exploration. Our results provide a comprehensive understanding of the roles of KL-regularization and data coverage in online decision making, shedding light on the design of more efficient algorithms.

Efficient optimization is essential for training large language models. Although intra-layer selective updates have been explored, a general mechanism that enables fine-grained control while ensuring convergence guarantees is still lacking. To bridge this gap, we propose \textbf{MGUP}, a novel mechanism for selective updates. \textbf{MGUP} augments standard momentum-based optimizers by applying larger step-sizes to a selected fixed proportion of parameters in each iteration, while applying smaller, non-zero step-sizes to the rest. As a nearly {plug-and-play} module, \textbf{MGUP} seamlessly integrates with optimizers such as AdamW, Lion, and Muon. This yields powerful variants such as \textbf{MGUP-AdamW}, \textbf{MGUP-Lion}, and \textbf{MGUP-Muon}. Under standard assumptions, we provide theoretical convergence guarantees for \textbf{MGUP-AdamW} (without weight decay) in stochastic optimization. Extensive experiments across diverse tasks, including MAE pretraining, LLM pretraining, and downstream fine-tuning, demonstrate that our \textbf{MGUP}-enhanced optimizers achieve superior or more stable performance compared to their original base optimizers. We offer a principled, versatile, and theoretically grounded strategy for efficient intra-layer selective updates, accelerating and stabilizing the training of large-scale models. The code is publicly available at https://github.com/MaeChd/MGUP.

Large Language Models (LLMs) have emerged as powerful tools for diverse applications. However, their uniform token processing paradigm introduces critical vulnerabilities in instruction handling, particularly when exposed to adversarial scenarios. In this work, we identify and propose a novel class of vulnerabilities, termed Tool-Completion Attack (TCA), which exploits function-calling mechanisms to subvert model behavior. To evaluate LLM robustness against such threats, we introduce the Tool-Completion benchmark, a comprehensive security assessment framework, which reveals that even state-of-the-art models remain susceptible to TCA, with surprisingly high attack success rates. To address these vulnerabilities, we introduce Context-Aware Hierarchical Learning (CAHL), a sophisticated mechanism that dynamically equilibrates semantic comprehension with role-specific instruction constraints. CAHL leverages the contextual correlations between different instruction segments to establish a robust, context-aware instruction hierarchy. Extensive experiments demonstrate that CAHL significantly enhances LLM robustness against both conventional attacks and the proposed TCA, exhibiting strong generalization capabilities in zero-shot evaluations while still preserving model performance on generic tasks. Our code is available at https://github.com/S2AILab/CAHL.

Intracortical brain-machine interfaces (iBMIs) have enabled movement and speech in people living with paralysis by using neural data to decode behaviors in real-time. However, intracortical neural recordings exhibit significant instabilities over time, which poses problems for iBMIs, neuroscience, and machine learning. For iBMIs, neural instabilities require frequent decoder recalibration to maintain high performance, a critical bottleneck for real-world translation. Several approaches have been developed to address this issue, and the field has recognized the need for standardized datasets on which to compare them, but no standard dataset exists for evaluation over year-long timescales. In neuroscience, a growing body of research attempts to elucidate the latent computations performed by populations of neurons. Nonstationarity in neural recordings imposes significant challenges to the design of these studies, so a dataset containing recordings over large time spans would improve methods to account for instabilities. In machine learning, continuous domain adaptation of temporal data is an area of active research, and a dataset containing shift distributions on long time scales would be beneficial to researchers. To address these gaps, we present the LINK Dataset (Long-term Intracortical Neural activity and Kinematics), which contains intracortical spiking activity and kinematic data from 312 sessions of a non-human primate performing a dexterous, 2 degree-of-freedom finger movement task, spanning 1,242 days. We also present longitudinal analyses of the dataset’s neural spiking activity and its relationship to kinematics, as well as overall decoding performance using linear and neural network models. The LINK dataset (https://dandiarchive.org/dandiset/001201) and code (https://github.com/chesteklab/LINK_dataset) are freely available to the public.