论文检索

Deep Learning · Large Language Models

Rachel Lawrence, Jacqueline Maasch

Autonomous reasoning is among the most scientifically and economically motivating topics in AI today. Historically the purview of symbolic AI, recent advances have mainly emerged from deep probabilistic generative models. Despite immense interest and rapid progress, the generative AI community has not clearly converged on operational definitions for reasoning and often implicitly rejects the historical treatment of this topic in logic, verifiable automated reasoning, and symbolic methods in general. **This position contends that definitional ambiguity leaves the construct validity of reasoning evaluation unverifiable, and undermines quantifiable progress toward the collective goal of trustworthy autonomous reasoning.** We also contend that this ambiguity is addressable. To that end, we provide (1) general and extensible definitions for *valid* and *sound reasoning* based on a synthesis of the literature, which can serve as an accessible reference and a starting point for community discussion; and (2) a checklist for best practices in the communication of AI reasoning research.

Optimization · Large Scale, Parallel and Distributed

Chengjie Ma, Seungeun Oh, Jihong Park, Seong-Lyun Kim

Federated learning (FL) enables distributed model training, yet in heterogeneous deployments, Bandwidth-Constrained Clients (BCCs) often contribute inefficiently due to limited uplink bandwidth. In model-heterogeneous FL with fixed small sub-models, BCCs with sub-models may improve quickly in early rounds but become under-parameterized later, resulting in slow convergence and poor generalization. To address this challenge, we propose FedGMR, a federated learning framework centered around Gradual Model Restoration (GMR), where GMR progressively increases each client’s sub-model density during training, allowing BCCs to remain effective contributors throughout optimization. To make GMR practical under real-world heterogeneity, FedGMR is realized as an end-to-end workflow with asynchronous coordination and stable, mask-aware aggregation. We further establish convergence guarantees, showing that the aggregation error scales with the average sub-model density across clients and rounds, and that GMR provably narrows the gap toward full-model FL. Extensive experiments on FEMNIST, CIFAR-10, ImageNet-100, and StackOverflow demonstrate that FedGMR improves both convergence speed and final accuracy, especially under severe heterogeneity and non-IID data distributions.

Deep Learning · Large Language Models

Haiyu Wang, Yuanyuan Lin

Recent studies have revealed two intriguing phenomena in large language models: massive activations, characterized by a small number of activations exhibiting abnormally large magnitudes, and attention sink, where a disproportionate amount of attention is consistently allocated to specific tokens regardless of their semantic relevance. However, the co-emergence and co-existence of these two phenomena remain poorly understood. In this work, we revisit the prevailing view that massive activations are the primary mechanism responsible for concentrating attention on sink tokens, and provide a more nuanced interpretation of their relationship. Through both theoretical analysis and empirical evidence, we demonstrate that massive activations and attention sink jointly act to prevent excessive token mixing in self-attention. Specifically, attention sink suppresses mixing among non-sink tokens, whereas massive activations suppress mixing between sink tokens and non-sink tokens. Furthermore, our theory provides a principled explanation of how the location of massive activations depends on the placement of layer normalization, and why KV-biases and gating mechanisms can remove massive activations while largely preserving attention sink. We further conduct intervention analyses and find that removing the value vector of the sink token can recover attention sink even when massive activations are entirely suppressed. Overall, this work provides a mechanistic perspective on how massive activations and attention sink interact under normalization and self-attention, offering new insights into their functional roles in Transformer models.

Applications · Computer Vision

Weirong Chen, Keisuke Tateno, Hidenobu Matsuki, Michael Niemeyer, Daniel Cremers, Federico Tombari

We address the problem of 4D reconstruction from partial point cloud sequences, where observations from depth sensors are incomplete, unordered, and lack explicit point correspondence over time. Recovering coherent 4D geometry in this geometry-only setting is challenging due to missing observations and ambiguous dynamics. While recent progress has largely been driven by image-based methods, existing point-based approaches typically focus on single-object scenarios, assume relatively complete inputs, and rely on explicit correspondence. To mitigate these limitations, we propose DynaTok, a point-based framework for correspondence-free 4D reconstruction from partial point cloud sequences that operates without images. DynaTok encodes each frame into compact latent tokens, aggregates incomplete observations over time with a Transformer-based spatiotemporal encoder, and decouples geometry and motion via a residual token design within a single unified model. Conditioned on the aggregated tokens, a point flow-matching decoder reconstructs complete and temporally consistent 4D point cloud sequences using only point cloud supervision. Experiments on object-level and scene-level benchmarks demonstrate improved reconstruction quality and temporal coherence under partial point cloud observations.

Applications · Chemistry, Physics, and Earth Sciences

Gnankan Landry Regis N'guessan, Bum Jun Kim

Physical systems near singularities, interfaces, and critical points exhibit power-law scaling, yet standard neural networks leave the governing exponents implicit. We introduce physics-informed M\"untz-Sz\'asz Networks (MSN-PINN), a power-law basis network that treats scaling exponents as trainable parameters. The model outputs both the solution and its scaling structure. We prove identifiability, or unique recovery, and show that, under these conditions, the squared error between learned and true exponents scales as $O(|\mu - \alpha|^2)$. Across experiments, MSN-PINN achieves single-exponent recovery with 1--5\% error under noise and sparse sampling. It recovers corner singularity exponents for the two-dimensional Laplace equation with 0.009\% error, matches the classical result of Kondrat'ev (1967), and recovers forcing-induced exponents in singular Poisson problems with 0.03\% and 0.05\% errors. On a 40-configuration wedge benchmark, it reaches a 100\% success rate with 0.022\% mean error. Constraint-aware training encodes physical requirements such as boundary condition compatibility and improves accuracy by three orders of magnitude over naive training. By combining the expressiveness of neural networks with the interpretability of asymptotic analysis, MSN-PINN produces learned parameters with direct physical meaning.

Reinforcement Learning · Policy Search

Jiaqi Wang, Haoge Deng, Ting Pan, Yang Liu, Chengyuan Wang, Fan Zhang, Yonggang Qi, Xinlong Wang

Uniform Discrete Diffusion (UDM) has recently emerged as a promising paradigm for discrete generative modeling; however, its integration with reinforcement learning remains largely unexplored. We observe that naively adapting GRPO to UDM leads to unstable training and marginal performance. To address this, we propose \Ours, the first framework that integrates UDM with RL. Our method is guided by two key insights: (i) treating the final clean sample, rather than intermediate predicted sample, as the action provides more accurate and stable optimization signals; and (ii) adopting the forward process to reconstruct the training trajectories helps the model learn probability paths that are more consistent with pretraining. For efficiency, we introduce Reduction-Step and CFG-Free training strategies. \Ours significantly improves the performance of the base model across multiple T2I tasks. Notably, GenEval accuracy improves from $69\%$ to $96\%$ and PickScore increases from $20.46$ to $23.81$, achieving state-of-the-art performance in both continuous and discrete settings. On the OCR benchmark, accuracy improves from $4\%$ to $57\%$, further validating the effectiveness and generalization capability of our method.

General Machine Learning · Sequential, Network, and Time Series Modeling

Nitin Bisht, Xiuwen Gong, Huan Huo, Guandong Xu

Modern recommender systems are typically trained offline and deployed with parameters held fixed between periodic refreshes, yet user behavior can evolve substantially during deployment. This can cause ranking utility to degrade over time and makes it difficult to provide formal guarantees about recommendation quality. We propose **CARE**, an adaptive calibration framework that wraps an arbitrary backbone recommender and outputs variable-size recommendation sets with finite-sample performance guarantees over interaction streams. CARE combines (i) a loss-based monitoring module that localizes behavioral changes and triggers threshold recalibration, and (ii) an online aggregation rule that promotes compact recommendation sets by dynamically reweighting candidate set predictors. We provide theoretical results establishing finite-sample guarantees for utility-based risk control and bounds on the expected set size relative to the best constituent predictor. Experiments across multiple datasets and backbone models demonstrate that CARE improves robustness and maintains compact recommendation sets while preserving the desired statistical guarantees. The code and implementation are available in https://anonymous.4open.science/r/CARE-FCBD.

General Machine Learning · Online Learning, Active Learning and Bandits

Nitin Bisht, Linjiang Guo, Xiuwen Gong, Huan Huo, Guandong Xu

Recommender systems (RS) may suffer from feedback-driven exposure concentration, where repeated engagement optimization collapses exposure onto a narrow set of categories, reducing catalog coverage and degrading long-horizon learning. Existing methods are often post hoc and typically lack principled uncertainty-aware risk estimates for regulating exposure under endogenous feedback. We therefore propose **CORAL**, a model-agnostic, uncertainty-aware framework that formulates exposure regulation as a constrained sequential decision problem. Specifically, we model self-reinforcing interactions to construct an exposure-saturation state, then derive an upper confidence bound on category-conditioned violation risk from observed history and incorporate it through a state-dependent penalty for adaptive intervention near saturation. Moreover, we provide theoretical guarantees for risk bounds, finite-time recovery, and efficient long-term performance. Extensive experiments on real-world datasets and controlled simulations validate the effectiveness of the proposed framework, which aligns with our theoretical analysis. Our code is available at: https://anonymous.4open.science/r/Coral_Rec-8400.

Social Aspects · Accountability, Transparency, and Interpretability

Guanzheng Qin, Chenghao Sun, Zhining Xie, Xinmei Tian

Large Language Models (LLMs) demonstrate remarkable potential in role-playing tasks but frequently suffer from personality decay—termed "Out-of-Character" (OOC) behavior—during prolonged interactions. While heuristic strategies exist to align model behaviors, the internal computational dynamics driving personality expression remain opaque. A fundamental barrier to decoding these mechanisms is a *metric gap*: while standard causal attribution paradigms target atomic, single-token outcomes, personality manifests as a holistic, multi-token behavioral tendency. We bridge this gap via the *Latent Persona Vector*, a differentiable proxy enabling the first fine-grained causal tracing of personality circuits. This metric reveals a structured "Preparation-Establishment-Expression" dynamic and identifies the mechanistic root of OOC behavior not as knowledge erasure, but as *generic prior dominance*. Specifically, we find that intrinsic assistant priors suppress emergent persona intents during the critical "Establishment" phase. Guided by this diagnosis, we propose surgically recalibrating the signal magnitude in fewer than 5% of attention heads. This targeted intervention effectively counteracts prior suppression, significantly restoring character consistency while preserving general reasoning capabilities.

Theory · Optimization

Yi Feng, Weiming Ou, Xiao Wang

The remarkable success of the Adam in training neural networks has naturally led to the widespread use of its descent-ascent counterpart, Adam-DA, for solving zero-sum games. Despite its popularity in practice, a rigorous theoretical understanding of Adam-DA still lags behind. In this paper, we derive ordinary differential equations (ODEs) that serve as continuous-time limits of the Adam-DA. These ODEs closely approximate the discrete-time dynamics of Adam-DA, providing a tractable analytical framework for understanding its behavior in zero-sum games. Using this ODE approach, we investigate two fundamental aspects of Adam-DA: local convergence and implicit gradient regularization. Our analysis reveals that the roles of the first- and second-order momentum parameters in zero-sum games are exactly the opposite of their well-documented effects in minimization problems. We validate these predictions through GAN experiments across multiple architectures and datasets, demonstrating the practical implications of this reversed momentum effect.

Deep Learning · Large Language Models

Gal Yona, Mor Geva, Yossi Matias

Despite significant improvements in factuality, confident errors continue to reappear as benchmarks probe more niche knowledge that models lack. We argue that most gains have come from expanding the model's knowledge boundary (encoding more facts) rather than improving awareness of that boundary (distinguishing known from unknown). We conjecture that this stems from the fact that the latter is inherently difficult: in the absence of strong ability to separate correct from incorrect answers (discrimination), fully eliminating hallucinations requires aggressive abstention, imposing a significant utility tax. Given this limitation, we propose complementing knowledge expansion with faithful uncertainty -- honestly conveying whatever uncertainty remains. This metacognitive capability becomes even more critical for tool-augmented models, where it serves as the control layer that determines when to search and how to weigh conflicting information. We conclude by highlighting the key challenges and open problems that must be tackled to make progress toward this objective.

Theory · Learning Theory

Mehryar Mohri, Yutao Zhong

The choice of loss function in classification involves a fundamental trade-off: smooth losses (like Cross-Entropy) enable fast optimization rates but yield slow square-root consistency bounds, while piecewise-linear losses (like Hinge) offer fast linear consistency rates but suffer from non-differentiability. We propose *Linear-Core (LC) Surrogates*, a new family of convex loss functions that resolve this tension by stitching a linear core to a smooth tail. We prove that these surrogates are differentiable everywhere while retaining strict linear $H$-consistency bounds, effectively combining the optimization benefits of smoothness with the statistical efficiency of margin-based losses. In the structured prediction setting, we show that this smoothness unlocks a massive computational and energy advantage: it allows for an unbiased stochastic gradient estimator that bypasses the quadratic complexity $O(|\mathcal{Y}|^2)$ of exact inference (e.g., Viterbi). Empirically, our method achieves a *23$\times$ speedup* over Structured SVMs on large-vocabulary sequence tagging tasks and demonstrates superior robustness to instance-dependent label noise, outperforming Cross-Entropy by *2.6\%* on corrupted CIFAR-10.

Theory · Learning Theory

Mingyue Xu, Gal Vardi, Itay Safran

We study *grokking* - the onset of generalization long after overfitting - in a classical ridge regression setting. We prove end-to-end grokking results for learning over-parameterized linear regression models using gradient descent with weight decay. Specifically, we prove that the following stages occur: (i) the model overfits the training data early during training; (ii) poor generalization persists long after overfitting has manifested; and (iii) the generalization error eventually becomes arbitrarily small. Moreover, we show, both theoretically and empirically, that grokking can be amplified or eliminated in a principled manner through proper hyperparameter tuning. To the best of our knowledge, these are the first rigorous quantitative bounds on the generalization delay (which we refer to as the "grokking time") in terms of training hyperparameters. Lastly, going beyond the linear setting, we empirically demonstrate that our quantitative bounds also capture the behavior of grokking on non-linear neural networks. Our results suggest that grokking is not an inherent failure mode of deep learning, but rather a consequence of specific training conditions, and thus does not require fundamental changes to the model architecture or learning algorithm to avoid.

Theory · Game Theory

Qichuan Yin, Ziwei Su, Shuangning Li

Human labeling increasingly relies on AI assistance, raising incentive challenges when annotators’ effort is unobserved. Recent work by Bastani & Cachon (2025); Sambasivan et al. (2021) shows that accuracy-based payment schemes suffer from incentive collapse: as AI accuracy improves, sustaining positive human effort requires unbounded payments. We study this problem in a budget-constrained setting with strategic annotators whose labeling accuracy depends on unobserved effort. We propose a sentinel-auditing payment mechanism that enforces a strictly positive and controllable level of human effort at finite cost, independent of AI accuracy. Building on this incentive-robust foundation, we develop an incentive-aware active statistical inference framework that jointly optimizes (i) the auditing rate and (ii) active sampling and budget allocation across instances of varying difficulty to minimize the final statistical loss under a single budget. Experiments demonstrate improved cost–error tradeoffs relative to standard active learning and auditing-only baselines.

Kai Hu, Akash Bharadwaj, Weichen Yu, Matt Fredrikson

We present a black-box model-stealing attack that recovers private vision-tokenizer configurations of deployed vision-language models (VLMs), including the visual patch size and input preprocessing pipeline. The key idea is a task-level side channel induced by ViT-style patchification: when a synthetic grid image is aligned with the hidden patch grid, boundary cues are erased at tokenization, causing periodic accuracy drop. By sweeping the grid cell size and measuring these collapses, we infer the patch size; by introducing padding and a consistency-check test, we further identify whether preprocessing is dynamic- or fixed-resolution and recover the target resize resolution. Across open-source Qwen-VL variants and proprietary models including GPT and Claude, we reliably recover tokenizer-related parameters. Finally, we show that such leakage enables preprocessing-aware transfer attacks and model-targeted adversarial manipulation.

Deep Learning · Large Language Models

beiya dai, Yuliang Liu, Daozheng Xue, Yunchong Song, Qipeng Guo, Kai Chen, Xinbing Wang, Bowen Zhou, Zhouhan Lin

We propose ContextLM, a framework that implicitly learns multi-token prediction by augmenting standard pretraining with an intrinsic next-context prediction objective. ContextLM builds a language model on top of context embeddings that span multiple tokens, enabling better next-token prediction by predicting the next context. Our model is fully compatible with standard autoregressive, token-by-token evaluation paradigms (e.g., perplexity). Extensive experiments with GPT-2 and Pythia backbones (up to 1.5B parameters and 300B training tokens) reveal that ContextLM shifts the Pareto frontier of scaling laws, exhibiting superior efficiency in parameters, training tokens, and FLOPs. Our results show that ContextLM could already achieve the baseline perplexity using 39\% fewer parameters and demonstrates robust generalization improvements on extensive downstream tasks under equivalent parameter counts.

Theory · Game Theory

ARNAB MAITI, Junyan Liu, Kevin Jamieson, Lillian Ratliff

We study the discrete Bertrand pricing game with a non-increasing demand function. The game has $n \ge 2$ players who simultaneously choose prices from the set {$1/k, 2/k, \ldots, 1$}, where $k\in\mathbb{N}$. The player who sets the lowest price captures the entire demand; if multiple players tie for the lowest price, they split the demand equally. We study the Bertrand paradox, where classical theory predicts low prices, yet real markets often sustain high prices. To understand this gap, we analyze a repeated-game model in which firms set prices using no-regret learners. Our goal is to characterize the equilibrium outcomes that can arise under different no-regret learning guarantees. We are particularly interested in questions such as whether no-external-regret learners can converge to undesirable high-price outcomes, and how stronger guarantees such as no-swap regret shape the emergence of competitive low-price behavior. We address these and related questions through a theoretical analysis, complemented by experiments that support the theory and reveal surprising phenomena for no-swap regret learners.

Theory · Probabilistic Methods

Qiaoyu Liang, Haohua Chen, Zihan Zhu, Michael Evans

From a statistical evidence perspective, we establish some asymptotic optimality properties of certain multiple testing rules based on the relative belief ratio (Evans, 2015). Under the two-groups model with an additive 0-1 loss and within a Bayesian decision theoretic asymptotic framework of Bogdan et al. (2011), we show that relative belief multiple testing rules induced by a simple one-group light-tailed normal prior with a single hyperparameter achieve the same asymptotic Bayes risk as the Bayes oracle benchmark. This risk is the minimum achievable in this asymptotic framework. Despite originating from a different starting point, the evidential relative belief approach enjoys oracle properties. The relative belief multiple testing approach is fundamentally different from existing Bayesian multiple testing procedures, virtually all induced by more complex heavy-tailed one-group global-local shrinkage priors using purely posterior-based inferences (Datta & Ghosh, 2013; Ghosh et al., 2016; Bhadra et al., 2017; Ghosh & Chakrabarti, 2017; Qin & Ghosh, 2025). By measuring statistical evidence via both the prior and posterior, the relative belief approach reveals an alternative new inferential paradigm for attaining asymptotic Bayes optimality under sparsity, one that does not rely on developing increasingly elaborate priors.

Theory · Domain Adaptation and Transfer Learning

Song Lai, Haohan Zhao, Rong Feng, Changyi Ma, Wenzhuo Liu, Hongbo Zhao, Xi Lin, Dong Yi, Qingfu Zhang, Hongbin Liu 等

Continual post-training (CPT) is a popular and effective technique for adapting foundation models like multimodal large language models to ever-evolving downstream tasks. While existing research primarily focus on methods like data replay, model expansion, or parameter regularization, the fundamental role of the learning paradigm remains largely unexplored. This paper presents a comparative analysis of two core post-training paradigms: supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT), investigating their respective impacts on knowledge retention during CPT. Our experiments are conducted across multiple multimodal tasks, utilizing Qwen2.5-VL-7B-Instruct as the base model. The investigation yields two significant findings: (1) When continuously learning on downstream tasks, SFT leads to catastrophic forgetting of previously learned tasks. In contrast, RFT inherently preserves prior knowledge and achieves performance comparable to multi-task training. (2) RFT successfully protects and even enhances the model's general knowledge on standard benchmarks, while SFT degrades general model capabilities severely. Further analysis reveals that this stability is not primarily due to explicit mechanisms like KL penalty or chain-of-thought reasoning. We investigate RFT's learning dynamics and find that its selective update mechanism inherently prevents interference with established knowledge. Based on this insight, we propose a rollout-based instance filtering algorithm (RIF-RFT) that enhances the training efficiency of RFT by focusing on learnable samples. Our comprehensive study demonstrates the superiority of RFT as a robust paradigm for continual post-training

Deep Learning · Foundation Models

Xiangyu Zeng, Zhiqiu Zhang, Yuhan Zhu, Xinhao Li, Zikang Wang, Changlian Ma, Qingyu Zhang, Zizheng Huang, Kun Ouyang, Tianxiang Jiang 等

Existing multimodal large language models for long-video understanding predominantly rely on uniform sampling and single-turn inference, limiting their ability to identify sparse yet critical evidence amid extensive redundancy. We introduce VideoSeeker, a novel framework that supports iterative discovery of salient visual clues, fine-grained inspection of key segments, and adaptive termination once sufficient evidence is acquired. Technically, we address two core challenges in interleaved tool invocation. First, to mitigate attention dispersion induced by the heterogeneity of reasoning and tool-calling, we propose Task-Decoupled Attention Masking, which isolates per-step concentration while preserving shared global context. Second, to control context length growth in multi-turn interactions, we introduce a Verifiable Trajectory-Guided Reward that balances exploration coverage with reasoning efficiency. To support training at scale, we further develop a data synthesis pipeline and construct Seeker-173K, comprising 173K high-quality tool-interaction trajectories for effective supervised and reinforcement learning. Extensive experiments show that VideoSeeker substantially outperforms state-of-the-art methods, achieving 72.1% accuracy on MLVU and 46.5% on Video-Holmes. These results demonstrate VideoSeeker's strong multi-hop evidence-seeking and reasoning capabilities, and validate the effectiveness of native tool invocation in long-video scenarios.