论文检索

Deep Learning · Large Language Models

Hong-Yu Chen, Po-Chiao Lin, Maojiang Su, Jerry Yao-Chieh Hu, Han Liu

We study the in-context universal approximation and compositional generalization of softmax Transformers. We prove an in-context universality result: a fixed-weight softmax Transformer approximates a broad class of continuous sequence-to-sequence functions. Building on this universality, we establish a composition theorem: by concatenating prompts associated with simple ``subprograms,'' the same fixed Transformer executes their composition, and thereby synthesizes more complex programs on-the-fly. These results support a principled view of prompts as programs and fixed-weight Transformers as program interpreters. Moreover, we provide a concrete mechanism by which GPT-style models both execute and assemble algorithms in context.

Deep Learning · Theory

Yixuan Wang, Dan Guralnik, Saiedeh Akbari, Warren Dixon

This article initiates the study of a basic question about model pruning. Given a vector s of importance scores assigned to model components, how many of the scored components could be discarded without sacrificing performance? We propose Effective Model Pruning (EMP), which derives the desired sparsity directly from the score distribution using the notion of effective sample size from particle filtering, also known as the inverse Simpson index. Rather than prescribe a pruning criterion, EMP supplies a universal adaptive threshold derived from the distribution of the score $s$ over the model components: EMP maps $s$ to a number $N_{eff} = N_{eff} (s)$, called the effective sample size. The $N − N_{eff}$ lowest scoring components are discarded. A tight lower bound on the preserved mass fraction seff (the sum of retained normalized scores) in terms of $N_{eff}$ is derived. This process yields models with a provable upper bound on the loss change relative to the original dense model. Numerical experiments are performed demonstrating this phenomenon across a variety of network architectures including MLPs, CNNs, Transformers, LLMs, and KAN. It is also shown that EMP addresses a rich set of pruning criteria such as weight magnitude, attention score, KAN importance score, and even feature-level signals such as image pixels.

General Machine Learning · Unsupervised and Semi-supervised Learning

Joonhyuk Lee, Virginia L., Sarah Zhao, Yash Nair, Asher Spector, Regev Cohen, Emmanuel J Candes

Verification of model outputs is rapidly emerging as a key primitive for both training and real-world deployment of large language models (LLMs). In practice, this often involves using imperfect LLM judges and reward models since ground truth acquisition can be time-consuming and expensive. We introduce Fully Unsupervised Score Ensembling (FUSE), a method for improving verification quality by ensembling verifiers without access to ground truth correctness labels. The key idea behind FUSE is to control conditional dependencies between verifiers in a manner that improves the unsupervised performance of a class of spectral algorithms from the ensembling literature. Despite requiring zero ground truth labels, FUSE typically matches or improves upon semi-supervised alternatives in test-time scaling experiments with diverse sets of generator models, verifiers, and benchmarks. In particular, we validate our method on both conventional academic benchmarks such as GPQA Diamond and on frontier, unsaturated benchmarks such as Humanity's Last Exam and IMO Shortlist questions.

General Machine Learning · Unsupervised and Semi-supervised Learning

Jiazhen Huang, Zhiming Liu, Changhu Wang, Wei Ju, Ziyue Qiao, Xiao Luo

Many methods aim to enhance the performance of vision-language models (VLMs) at test time. Among them, transduction has emerged as a promising paradigm due to its strong compatibility and efficiency. However, realistic evaluations often involve highly imbalanced class distributions, which cause performance degradation or even collapse. In this work, we systematically revisit transduction from the perspective of penalized likelihood estimation (PLE), showing that PLE with a KL-divergence anchor term naturally yields an adaptive shrinkage behavior between prior anchors and empirical estimates. From this viewpoint, the brittleness of transductive methods can be attributed to the absence of anchoring mechanism and static modeling of the shrinkage strength. Therefore, we propose Mixture of Von Mises-Fisher Models with Dynamic Shrinkage (MOON). MOON is based on a mixture of von Mises-Fisher distributions to model feature representations on the unit hypersphere. To handle imbalance, MOON dynamically adjusts the shrinkage strength using zero-shot priors at both instance and class levels. Thus, it suppresses unreliable assignments and prevents harmful updates from outlier classes, thereby mitigating negative transfer. MOON is model-agnostic, training-free, and requires no task-specific hyperparameter tuning. Extensive experiments further validate the advantage of MOON in both performance and efficiency.

General Machine Learning · Causality

Kiet Vo, Gowtham Reddy Abbavaram, Julian Rodemann, Siu Lun Chau, Krikamol Muandet

We study off-policy evaluation (OPE) under strategic behavior where decision subjects (or agents) respond to a decision maker's policy by strategically modifying their covariates. Such behavior induces a policy-dependent covariate shift, breaking the standard assumption in existing methods that covariates are exogenous to the policy. Related work addresses this challenge by imposing strong assumptions such as repeated interactions or full knowledge of agents’ response behavior, substantially limiting its applicability to OPE. In contrast, we consider a one-shot OPE setting where the decision maker has only partial knowledge of the agents' response behavior. Our key insight is that disclosing local information through post-hoc explanations reveals agents’ pre-strategic covariates prior to adaptation, mitigating the information loss induced by strategic behavior. Leveraging this structure, we estimate a statistical model for the agents’ responses and construct a doubly robust estimator for policy value. By assuming that the agents' cost sensitivity follows a conditional log-normal distribution, we establish consistency of the proposed estimator and validate our approach empirically. More broadly, our results highlight how interaction design can mitigate information asymmetry by revealing otherwise hidden structure in agents' strategic responses.

Theory · Learning Theory

Julian Rodemann, Unai Fischer Abaigar, James Bailie, Krikamol Muandet

Performative predictions influence the very outcomes they aim to forecast. We study performative predictions that affect a sample (e.g., only existing users of an app) and/or the whole population (e.g., all potential app users). This raises the question of how well models generalize under performativity. For example, how well can we draw insights about new app users based on existing users when both of them react to the app's predictions? We address this question by embedding performative predictions into statistical learning theory. We prove generalization bounds under performative effects on the sample, on the population, and on both. A key intuition behind our proofs is that in the worst case, the population negates predictions, while the sample deceptively fulfills them. We cast such self-negating and self-fulfilling predictions as min-max and min-min risk functionals in Wasserstein space, respectively. Our analysis reveals both a fundamental trade-off between performatively changing the world and learning from it, as well as a surprising insight on how to improve generalization guarantees by retraining on performatively distorted samples. We illustrate our bounds using real data on prediction-informed assignments to job trainings.

Optimization · Large Scale, Parallel and Distributed

Zaifeng Pan, Yipeng Shen, Zhengding Hu, Zhuang Wang, Aninda Manocha, Zheng Wang, zhongkai yu, Yue Guan, Yufei Ding

LLM-based multi-agent simulations are increasingly adopted across application domains, but remain difficult to scale due to GPU memory pressure. Each agent maintains private GPU-resident states, including models, prefix caches, and adapters, which quickly exhaust device memory as the agent count grows. We identify two key properties of these workloads: sparse agent activation and an estimable agent invocation order. Based on an analysis of representative workload classes, we introduce invocation distance, a unified abstraction that estimates the relative order in which agents will issue future LLM requests. Leveraging this abstraction, we present ScaleSim, a memory-efficient LLM serving system for large-scale multi-agent simulations. ScaleSim enables proactive prefetching and priority-based eviction, supports diverse agent-specific memory through a modular interface, and achieves up to 1.74$\times$ speedup over SGLang on simulation benchmarks.

Deep Learning · Large Language Models

Haochun Wang, Chaofen Yang, Jiatong Liu, Jingbo Wang, Zewen Qiang, Sendong Zhao, Ting Liu, Bing Qin

In-context learning (ICL) is highly sensitive to which demonstrations appear in the prompt, but selecting them is expensive because candidate contexts must be validated with repeated LLM calls. We argue that demonstration selection is \emph{easier to judge than to find}: predicting whether a specific query--context pair $(q,D)$ will succeed is cheaper and more general than searching for an optimal $D^\star$. Based on this insight, we propose DiSP, a sample-and-judge framework that stratifies queries by difficulty. DiSP runs random demonstration trials to estimate each training query's success rate, trains a lightweight router to predict difficulty from the query, and trains level-specific judges to score sampled contexts. At inference, DiSP performs stop-on-acceptance judging under an explicit budget and typically makes a single LLM call, emitting diagnostic risk tags when no suitable context is found. Across five classification datasets with Llama 3–8B and Qwen 2.5–7B, DiSP achieves the best average accuracy, improving over strong learned selection baselines by up to 3.4%, while achieving up to 23× end-to-end wall-clock speedup.

Zhen Wang, Fan Bai, Zhongyan Luo, Jinyan Su, Kaiser Sun, Xinle Yu, Jieyuan Liu, Kun Zhou, Claire Cardie, Mark Dredze 等

Autonomous agents powered by large language models (LLMs) promise to accelerate scientific discovery, but rigorously evaluating their capacity for verifiable discovery remains a central challenge. Existing benchmarks face a trade-off: they either rely on LLM-as-judge evaluations of automatically generated papers, or optimize isolated performance metrics that provide only coarse proxies for scientific insight. To address this, we introduce FIRE-Bench (Full-cycle Insight Rediscovery Evaluation), a benchmark that evaluates agents through the rediscovery of established findings from recent, high-impact machine learning research. Agents are given only a high-level research question from a published study and must autonomously design experiments, implement code, execute their plans, and derive conclusions supported by empirical evidence. We evaluate a range of state-of-the-art agents with frontier model backbones, such as gpt-5, on FIRE-Bench. Our results show that full-cycle scientific research remains challenging for current agent systems: even the strongest agents achieve limited rediscovery success, exhibit high variance across runs, and display recurring failure modes in experimental design, execution, and evidence-based reasoning. Overall, FIRE-Bench provides a rigorous and diagnostic framework for measuring progress toward reliable agent-driven scientific discovery.

Deep Learning · Generative Models and Autoencoders

Zhao zuopeng, Ying Liu, xiaoyu Li, SuLuo, Lu Li, Wen L

Existing diffusion models have made significant progress in generating realistic images. However, their direct adaptation to remote sensing imagery often disregards intrinsic physical laws. This oversight frequently leads to spectral distortion and radiometric inconsistency, severely limiting the scientific utility of generated data. To address this issue, this paper introduces AnyBand-Diff, a novel spectral-prior-guided diffusion framework tailored for robust spectral reconstruction. Specifically, we design a Masked Conditional Diffusion backbone integrated with a dual stochastic masking strategy, empowering the model to recover complete spectral information from arbitrary band subsets. Subsequently, to ensure radiometric fidelity, a Physics-Guided Sampling mechanism is proposed, leveraging gradients from a differentiable physical model to explicitly steer the denoising trajectory toward the manifold of physically plausible solutions. Furthermore, a Multi-Scale Physical Loss is formulated to enforce rigorous constraints across pixel, region, and global levels in a joint manner. Extensive experiments confirm the effectiveness of AnyBand-Diff in generating reliable imagery and achieving accurate spectral reconstruction, contributing to the advancement of physics-aware generative methods for Earth observation.

Reinforcement Learning · Multi-agent

Xin Yang, Letian Li, Zimo Ji, Terry Zhang, Wenyuan Jiang

LLM-based multi-agent systems (MAS) promise scalable collaboration, yet adding agents often *reduces* reliability. This position paper argues that many MAS failures are fundamentally **concurrency control problems**: agents concurrently read and write shared state, and long LLM inference windows amplify the risk of stale reads, lost updates, and inconsistent outcomes. Failure modes commonly attributed to "coordination" or "communication" breakdowns can be mapped directly onto classical concurrency anomalies. Rather than treating these as emergent behaviors to be solved by better prompting or more capable models, we contend that MAS frameworks should incorporate explicit concurrency control mechanisms: conflict detection, isolation guarantees, and structured access to shared resources. Concurrency control should be a first-class design concern, not an afterthought.

Deep Learning · Large Language Models

Ru Zhang, Renda Li, Ziyu Ma, Weijie Qiu, Chongyang Tao, Yong Wang, Xiangxiang Chu

Reinforcement learning (RL) has demonstrated potential for enhancing reasoning in large language models (LLMs). However, effective RL training, which requires medium-difficulty training samples, faces two fundamental challenges: Effective Data Scarcity and Dynamic Difficulty Shifts, where medium-difficulty samples are scarce and become trivial as models improve. Existing methods mitigate this scarcity to some extent by generating training samples. However, these approaches suffer from anchor-free generation, ignoring co-evolution, and difficulty mismatch. To address these issues, we propose D²Evo, a Dual Difficulty-aware self-Evolution RL framework. In each iteration, our method mines medium-difficulty anchors based on the current Solver's capability, trains the Questioner to generate diverse questions at appropriate difficulty levels, and jointly optimizes both components to enable progressive reasoning gains. Extensive experiments demonstrate that D²Evo outperforms existing methods on mathematical reasoning benchmarks with fewer than 2K real mathematical samples, and exhibits strong generalization on general reasoning benchmarks.

Deep Learning · Sequential Models, Time series

Dai Shi, Lequan Lin, Andi Han, Luke Thompson, Jose Miguel Hernandez-Lobato, Zhiyong Wang, Junbin Gao

Stochastic differential equations (SDEs) and stochastic partial differential equations (SPDEs) are fundamental for modeling stochastic dynamics across the natural sciences and modern machine learning. Learning their solution operators with deep learning models promises fast solvers and new perspectives on classical learning tasks. In this work, we build on Wiener–chaos expansions (WCE) to design neural operator (NO) architectures for SDEs and SPDEs: we project driving noise paths onto orthonormal Wick–Hermite features and use NOs to parameterize the resulting chaos coefficients, enabling reconstruction of full trajectories from noise in a single forward pass. We also make the underlying WCE structure explicit for multi-dimensional SDEs and semilinear SPDEs by showing the coupled deterministic ODE/PDE systems governing these coefficients. Empirically, we achieve competitive accuracy across several tasks, including standard SPDE benchmarks and SDE-based diffusion one-step image sampling, topological graph interpolation, financial extrapolation, parameter estimation, and manifold SDE flood forecasting. These results suggest WCE-based neural operators are a practical and scalable approach to learning SDE/SPDE solution operators across domains.

Deep Learning · Graph Neural Networks

Dai Shi, Linhan Luo, Luke Thompson, Lequan Lin, Andi Han, Junbin Gao, Jose Miguel Hernandez-Lobato

Message-passing neural networks (MPNNs) often suffer from an information bottleneck when capturing long-range dependencies, leading to the oversquashing (OSQ) phenomenon. Alongside spatial connectivity enrichment (e.g., rewiring), recent studies have shown that spectral filtering can yield strong long-range learning outcomes, as spectral operators enable global information mixing that alleviates OSQ. These approaches achieve this either by stabilizing the Jacobian energies in deep propagation or by guaranteeing OSQ mitigation under strong theoretical assumptions. We revisit these conclusions and show that the associated Jacobian sensitivity lower bound is generally difficult to achieve in practice. We then propose S$^3$GNN, which mitigates OSQ without such restrictive assumptions by lightweightly reintroducing omitted components with substantially lower computational complexity, while standard stability constraints on feature transformations remain effective under our new dynamics. Extensive experiments across diverse domains (e.g., long-range benchmarks, KGQA, and mesh-based fluid dynamics) demonstrate that S$^3$GNN achieves up to an order-of-magnitude error reduction with up to 50\% fewer parameters. Our code can be found in https://anonymous.4open.science/r/S3GNN-2359.

Deep Learning · Large Language Models

Qiuyang Mang, Wenhao Chai, Zhifei Li, Huanzhi Mao, Shang Zhou, Alexander Du, Hanchen Li, Shu Liu, Edwin Chen, Yichuan Wang 等

We introduce FrontierCS, a benchmark of 240 open-ended problems across diverse areas of computer science, designed and reviewed by experts, including CS PhDs and top-tier competitive programming participants and problem setters. Unlike existing benchmarks that focus on tasks with known optimal solutions, FrontierCS targets problems where the optimal solution is unknown, but the quality of a solution can be objectively evaluated. Models solve these tasks by implementing executable programs rather than outputting a direct answer. FrontierCS includes algorithmic problems, which are often NP-hard variants of competitive programming problems with objective partial scoring, and research problems with the same property. For each problem, we provide an expert reference solution and an automatic evaluator. Combining open-ended design, measurable progress, and expert curation, FrontierCS provides a benchmark at the frontier of computer-science difficulty. Empirically, we find that frontier reasoning models still lag far behind human experts, and that simply increasing reasoning budgets does not close this gap on open-ended challenges. Moreover, these models struggle to identify internal equivalence classes, and existing agentic frameworks also exhibit brittleness on such problems due to overfitting. FrontierCS thus offers a new lens into model capabilities on real frontier computer science problems.

Deep Learning · Large Language Models

Julianna Piskorz, Cristina Pinneri, Alvaro Correia, Motasem Alfarra, Risheek Garrepalli, Christos Louizos

Masked Diffusion Language Models (MDLMs) have recently emerged as a promising alternative to Autoregressive Language Models (ARLMs), leveraging a denoising objective that, in principle, should enable more uniform context utilisation. In this work, we examine the context comprehension abilities of MDLMs and uncover two key limitations. First, despite their more global training objective and bidirectional attention mechanism, similarly to ARLMS, **MDLMs exhibit a strong locality bias**: performance is highly sensitive to the position of relevant information within the input, favouring local over distant context. Second, appending a large number of **mask tokens—required for generation—can significantly degrade context comprehension** in models trained from scratch. Through systematic ablations, we find that these masks **act as distractors**, reducing the model's ability to process relevant information. To address and further study this undesirable behaviour, we introduce the mask-agnostic loss function that encourages predictions to remain invariant to the number of appended masks. Fine-tuning with this objective substantially mitigates the distracting effect of masks, improving robustness of MDLMs. Overall, our findings reveal critical limitations of the current MDLM training paradigm, with implications for training, evaluation and deployment.

Social Aspects · Accountability, Transparency, and Interpretability

Pierre Fernandez, Tom Sander, Hady Elsahar, Hongyan Chang, Tomáš Souček, Valeriu Lacatusu, Tuan Tran, Sylvestre-Alvise Rebuffi, Alexandre Mourachko

Generation-time text watermarking embeds statistical signals into text for traceability of AI-generated content. We explore post-hoc watermarking where an LLM rewrites existing text while applying generation-time watermarking, to protect copyrighted documents, or detect their use in training or RAG via watermark radioactivity. Unlike generation-time approaches which are constrained by how LLMs are served, this setting offers additional degrees of freedom for both generation and detection. We thus investigate how allocating compute (through larger rephrasing models, beam search, multi-candidate generation, or entropy filtering at detection) affects the quality-detectability trade-off. Among our findings, the simple Gumbel-max scheme surprisingly outperforms more recent alternatives under nucleus sampling, and achieves strong detectability and semantic fidelity on open-ended text such as books. Moreover, most methods benefit significantly from beam search, and we counterintuitively find that smaller models outperform larger ones. However, our solutions struggles when watermarking verifiable text such as code. This study reveals both the potential and limitations of post-hoc watermarking, laying groundwork for practical applications and future research.

Social Aspects · Accountability, Transparency, and Interpretability

Sylvestre-Alvise Rebuffi, Tuan Tran, Valeriu Lacatusu, Pierre Fernandez, Tomáš Souček, Nikola Jovanović, Tom Sander, Hady Elsahar, Alexandre Mourachko

Existing approaches for watermarking AI-generated images often rely on post-hoc methods applied in pixel space, introducing computational overhead and potential visual artifacts. In this work, we explore latent space watermarking and introduce DistSeal, a unified approach for latent watermarking that works across both diffusion and autoregressive models. Our approach works by training post-hoc watermarking models in the latent space of generative models. We demonstrate that these latent watermarkers can be effectively distilled either into the generative model itself or into the latent decoder, enabling in-model watermarking. The resulting latent watermarks achieve competitive robustness while offering similar imperceptibility and up to 20x speedup compared to pixel-space baselines. Our experiments further reveal that distilling latent watermarkers outperforms distilling pixel-space ones, providing a solution that is both more efficient and more robust.

Deep Learning · Sequential Models, Time series

Luke Thompson, Dai Shi, Lequan Lin, Junbin Gao, Andi Han

Neural rough differential equations (NRDEs) learn continuous-time dynamics from irregularly sampled sequences by encoding the input path with signature features, providing robustness to discretisation and sampling irregularity. However, existing NRDEs implicitly rely on algebraic identities that can fail in two important settings: *stochastic dynamics* interpreted in the Itô sense, and *dynamics evolving on manifolds* where curvature renders the effect of repeated derivatives order-dependent. In this work, we propose Branched Neural Rough Differential Equations (B-NRDEs), a unified framework that replaces geometric signature features with tree-based (branched) rough-path lifts, yielding models that remain well-defined under Itô noise and on manifolds. Building on these branched lifts, an Itô-consistent training objective is introduced via the branched signature kernel. We provide an efficient, autodifferentiable package *Stochastax* for computing branched (log-)signatures and solving (manifold) RDEs. Across various applications, including rough Bergomi volatility modelling, sim-to-real $\mathrm{SO}(3)$ dynamics forecasting, and SPD covariance dynamics, B-NRDE shows consistently strong results.

Deep Learning · Large Language Models

Baturay Saglam, Dionysios Kalogerias

Large language models can produce toxic or inappropriate text even for benign inputs, creating risks when deployed at scale. Detoxification is therefore important for safety and user trust, particularly when we want to reduce harmful content without sacrificing the model’s generation quality. Many existing approaches rely on model retraining, gradients, or learned auxiliary components, which can be costly and may not transfer across model families or to truly black-box settings. We introduce a test-time procedure that approximates the gradient of completion toxicity with respect to the input embeddings and uses a small number of descent steps to steer generation toward less toxic continuations. This is achieved with zeroth-order optimization that requires only access to input embeddings, a toxicity scoring function, and forward evaluations of the model. Empirically, the approach delivers robust toxicity reductions across models and prompts and, in most settings, achieves the best overall toxicity–quality trade-off. More broadly, our work positions word embeddings as effective control variables and encourages wider use of black-box optimization to guide autoregressive language models toward scalable, safer text generation, without requiring any training or access to intermediate computations.