NeurIPS 2026 论文列表

Supervised fine-tuning (SFT) on long teacher trajectories is the dominant method for instilling investigation and reasoning capabilities into open software-engineering (SWE) agents. Under SFT, every retained response is an imitation target, so the student inherits not only the trajectory's outcome but also any flaw in its intermediate steps, including ungrounded leaps and redundant loops. High-quality training data must therefore be jointly effective (each step is grounded and narrows the agent's epistemic gap to the correct fix) and efficient (each step is information-bearing rather than redundant or looping). Existing recipes filter or relabel teacher rollouts using only a binary terminal verifier, which does not directly target these axes and provides no supervision on instances where the teacher fails. Every real issue ships with a developer-authored reference patch $p^\star$ that implicitly testifies to the file paths, runtime behaviors, and conventions a fix presupposes, but the standard pipeline discards it. We propose P2T (Patches-to-Trajectories), which uses $p^\star$ as privileged information during curation, and frames trajectory construction as a bi-objective program over per-step effectiveness and trajectory length. A reverse phase distills $p^\star$ into a latent process graph $G^\star$ of contextual facts and solution milestones, encoding dense intermediate anchors in constructive ordering. A forward phase curate trajectories from blinded teacher continuations, scoring per-step progress against $G^\star$ under a leakage-blocking groundedness check and committing the shortest segments that retain effectiveness. Using only 1.8k curated SWE-Gym instances, P2T improves both axes simultaneously over outcome-filtered SFT and its tool-error-masking variant: on SWE-bench Verified, it lifts Pass@1 by up to +10.8 points while cutting per-instance inference cost by 15%, with consistent gains on SWE-bench Lite and across two teachers. A size-matched ablation and qualitative analysis further isolate per-trajectory quality from data scale.

Model-based reinforcement learning in partially observable environments requires inferring hidden states, learning latent dynamics, and planning under the resulting belief. Practical algorithms often solve one aspect of the above pipeline while relying on approximations for the others that are often poorly understood, while existing POMDP theory typically analyzes computationally inefficient theoretical algorithms far from modern deep RL practice. To fill this gap, we propose a practical end-to-end algorithm that approximately solves the hidden state filtering problem via sequential Monte Carlo (SMC), updates a latent dynamics model online, and performs approximate online planning for $d$ steps before bootstrapping leaves with a state-value critic (a QMDP approximation). We analyze each approximation and prove an end-to-end regret bound decomposing the error into model estimation error, particle filter error, Monte Carlo rollout error, and a structural bias term induced by the QMDP approximation controllable by the planning depth that we further analyze in various special cases. The resulting guarantee makes explicit how statistical sample size, particle budget, rollout budget, and planning depth affect performance. We instantiate our algorithm in an MPC-based deep RL framework, demonstrating its practicality on a variety of environments. Our results provide a practical and theoretically grounded route for bridging theory and practice within online decision-making in POMDPs.

Large Language Models (LLMs) demonstrate remarkable reasoning capabilities, yet the internal mechanisms driving these multi-step processes remain opaque. Existing mechanistic interpretability approaches often rely on superficial text-pattern co-occurrences or capture only short-term effects, struggling to trace the long-horizon, outcome-driven influence of internal components such as neurons or sparse features in multi-step reasoning. In this paper, we introduce Integrated Policy Gradient (IPG), a novel, training-free framework that brings policy-based attribution into mechanistic interpretability for LLM reasoning. Grounded in *outcome-oriented* and *sequential-influence-aware* principles, it backpropagates outcome-level, non-differentiable behavioral signals (e.g., reasoning outcome correctness) through entire inference trajectories, utilizing path integration to isolate outcome-relevant mediators at component level for reliable attribution. This provides a linear approximation of Causal Mediation Analysis (CMA) in reasoning. Empirical evaluations demonstrate that IPG achieves precise localization and reveals mechanistic insights by identifying sparse, concentrated subsets of internal components associated with specific reasoning behaviors. Our method also enables modulation of reasoning capabilities, offering a powerful and reliable framework for understanding LLM reasoning.

Diffusion language models (DLMs) can generate multiple tokens in parallel, but training large DLMs from scratch remains expensive. A practical alternative is to adapt off-the-shelf autoregressive (AR) checkpoints into diffusion models, reusing their linguistic and reasoning capabilities. Existing adaptation recipes either modify logits and grow attention masks toward full-sequence diffusion, or directly fine-tune ARweights under a block-diffusion objective, leaving two questions underexplored: what diffusion paradigm should AR-to-DLM adaptation target, and what transition path preserves AR knowledge most effectively? We argue that Block-Diffusion is a natural destination because AR decoding corresponds to block size one at the level of attention and generation order, while larger blocks introduce controlled intra-block bidirectionality and parallel generation. Based on this view, we propose a context-causal adaptation path that keeps committed context strictly causal, a one-pass parallel training formulation with auxiliary AR guidance, and a gradual block-size curriculum. Across several AR initializations and model scales, these components improve average adaptation performance over random mask annealing and direct fine-tuning. Scaling the recipe yields NBDIFF-7B, which supports 32K-token contexts and achieves the strongest average performance among the compared diffusion LLM baselines on general, math, and code benchmarks. Code and checkpoints will be released upon publication.

Importance weights are essential in domain adaptation under label shift, yet their utility is often undermined by the finite sample uncertainty associated with their estimation. Existing methods typically analyze this uncertainty through Gaussian elimination on interval-valued linear systems, which leads to overly conservative confidence regions and inefficient downstream applications. We propose a paradigm shift from inversion-based inference to a direct matrix constraint framework. We use this framework to define a joint confidence region and extract marginal intervals via linear programming, deriving provably tighter bounds for importance weights while maintaining exact finite-sample validity. Furthermore, we analyze the confidence region's geometry and provide the theoretical results for its diameter bounds. Evaluated across text, image, multimodal benchmarks, including AGNews, MNIST, CIFAR-10, N24News, and a real-world autonomous driving dataset, nuImages, our approach consistently yields shorter confidence intervals and smaller prediction sets than inversion-based methods.

Large language models (LLMs) are driving autonomous agents toward real-world environments, with the Web emerging as one of the most representative interaction scenarios. However, training Web agents with naive exploration remains challenging: successful trajectories are rare, and the sparse terminal rewards make it difficult for agents to obtain useful feedback from failed trajectories. Inspired by skill acquisition theory, we propose Skill-Composed Progressive Exploration (SCOPE), a short-to-long learning framework for long-horizon Web agents that decomposes complex task exploration into a gradual expansion process from local skills to complete task-solving capability. Specifically, SCOPE iteratively explores around local skills: Skill Discovery extracts reusable local skills and failure experiences from historical rollouts; Skill Orchestration composes discovered skills into high-level skill chains; andSkill Expansion leverages failure experiences aligned with the current skill chain to guide the agent toward exploring new subskills. Through this process, SCOPE progressively constructs task-completing skill chains across rollouts and internalizes them into the policy parameters through agentic on-policy distillation. Extensive experiments on WebArena, WebVoyager, and WebShop show that SCOPE consistently improves exploration dynamics and achieves 3.9\%--18.6\% relative gains over the strongest baselines in each setting. Our code is publicly available at https://anonymous.4open.science/r/SCOPE-14D3.

Link prediction is a standard pretraining objective for graph neural networks: hide edges, train an encoder--decoder to predict them, and reuse the node embeddings for downstream node tasks. Despite its empirical success, there is little quantitative theory explaining why an edge-level objective should produce embeddings that are linearly useful for node classification.We study this question in the finite-sample stochastic block model (SBM). For any bounded GNN encoder paired with a symmetric bilinear decoder, we prove a conditional reduction: small population held-out LP excess risk implies small community-classification error for a transductive linear classifier on the learned embeddings.The bound decomposes into an optimization residual, an architectural approximation residual, and explicit statistical terms of order $\widetilde{O}(\sqrt{d/n})$ in bounded dense regimes. We further give a constructive few-label classifier: nearest-centroid classification under the PSD quadratic form $G = W Z^\top Z W$, whose label error is coupled to LP quality.The framework is architecture-agnostic through the approximation residual, and we instantiate this residual as $O(\log n / n)$ for a spectral-positional-encoding GIN with a bilinear decoder. Synthetic SBM experiments validate the theorem chain from LP risk to score error to classification error, and show when the worst-case constants become conservative.

Multimodal Large Language Models (MLLMs) face significant scalability challenges in long-video understanding due to quadratic attention complexity and context window constraints. Existing retrieval-based approaches typically using supervised learning on static keyframe annotations. However, these label priors often fail to align with the actual information required by the downstream MLLM for complex reasoning, creating a fundamental misalignment between the training target and inference needs. To bridge this gap, we propose a probabilistic framework that reformulates frame selection as a Bayesian inference process. We first initialize a policy using supervised learning to capture general saliency. Then, we treat the frozen MLLM as a task environment and employ its negative log-likelihood (NLL) as dense reward evidence to update the policy via Bayesian Group Relative Policy Optimization (GRPO). This process explicitly transits the selection principle from label priors to task evidence. Extensive experiments on Video-MME, MLVU, and LongVideoBench demonstrate that our approach matches the accuracy of heavy MLLM scorers while maintaining the efficiency of lightweight models.

Rough path signatures provide a universal feature map for continuous paths and, via the expected signature, a principled characterisation of path distributions. These results do not directly extend to \clag paths of Temporal Point Processes (TPPs), limiting the use of signature methods for event sequences. Furthermore, neural TPP models, including recent generative approaches, optimise per-event objectives with no global sequence-level loss, while evaluation of variable-length event sequences lacks distributional discrepancy measures. This paper proposes a common pathwise framework for addressing these limitations. We introduce the interarrival embedding, a stable (homeomorphic) lift from jump paths to continuous paths of bounded variation, enabling signature methods to discrete event sequences. Our theoretical contributions give rise to \sigTPP, the first signature-based generative model for TPPs, trained using a path-level loss on complete trajectories. We further analyse the space of counting paths and derive three distributional discrepancies, providing mathematically justified tools for evaluating generative TPP models. Across synthetic and real-world datasets, \sigTPP~achieves the best average rank based on 8 complementary metrics, outperforms or is within one standard error of the strongest baseline in $64\%$ of the dataset-metric pairs, and according to a relative score, improves against every baseline by at least $19\%$ on average.

How can we understand what a music foundation model has learned \textit{internally}? Most interpretability approaches, such as probing and Sparse Autoencoders (SAEs), focus on identifying individual features with minimal structural assumptions. We argue that many concepts are better understood as \textit{structured relations} rather than isolated features. This is especially prominent in music, where tonal structures are organized in the space of pitch and time. For example, concepts such as chords or keys are naturally expressed as structured sets (e.g., the 12 transpositions of a chord or the diatonic system within a key), rather than isolated features. In this study, \textbf{we shift from feature identification to structure-based analysis}, asking whether the learned inner representations of music language model emerge as organized structures over features. To this end, we introduce a framework that uses pitch transposition as an inductive bias to induce ordered orbits via multi-view SAE alignment. Concretely, we generate pitch-shifted input pairs and align their SAE representations to discover structured groups of pitch-related features. Experimental results show that this approach recovers orbit structures corresponding to chords, keys, and melodic patterns across two state-of-the-art music foundation models, while requiring only minimal grounding (e.g., a few anchor examples) to interpret entire concept families. Beyond interpretability, our results suggest a new paradigm for representation analysis: evaluating whether coherent relational structures emerge in learned representations as a proxy for model understanding.

Outcome homogenization---models producing similar outputs---has drawn increasing attention in the large language model (LLM) community as researchers and practitioners have come to recognize that alignment practices can reduce model diversity and reproduce dominant narratives. However, most approaches address outcome homogenization exclusively through a technical lens. In this position paper, we argue that outcome homogenization is best understood as a value-chain phenomenon shaped by stakeholders, governance structures, resource allocation, model development, and organizational practices. We identify key drivers of homogenization in the AI ecosystem, showing how current practices reinforce dominant paradigms while marginalizing alternative perspectives. We argue that addressing homogenization requires shifting attention from model-level decisions to the broader power dynamics and organizational practices. Through a case study of LLM vendors’ safety alignment practices, we illustrate how these ecosystem-level forces shape model development and deployment. We conclude with a call to action, offering targeted recommendations for the research community.

Distributed learning algorithms are vulnerable to adversarial nodes, a.k.a. Byzantine failures. To solve this issue, robust algorithms have been developed, which typically replace parameter averaging by robust aggregations. While generic conditions on these aggregations exist to guarantee the convergence of (Stochastic) Gradient Descent (SGD), the analyses remain rather ad-hoc. This hinders the development of more complex robust algorithms, such as accelerated ones. In this work, we show that Byzantine-robust distributed optimization can, under standard generic assumptions, be cast as a general optimization with inexact gradient oracles, an active field of research. This allows to obtain state-of-the-art results for Byzantine-robust optimization from general inexact first-order analyses. We first show that inexact GD on top of standard robust aggregation procedures obtains optimal asymptotic error in the Byzantine setting. Going further, we study an algorithm for Optimization under Similarity, in which the server leverages an auxiliary loss function that approximates the global loss. Then, we introduce an Accelerated Extra-gradient method, that yields acceleration in both the standard and similarity settings. We first give new general convergence results for these inexact schemes, and then instantiate these results in the Byzantine setting through our reduction. Both algorithms drastically reduce the communication complexity compared to previous methods, as we show theoretically and empirically.

Large Language Model-based Automated Algorithm Design (LLM-AAD) has shown promising results across diverse domains by coupling LLMs with iterative search. Some recent efforts have begun to fine-tune LLMs for algorithm design through reinforcement learning. However, these attempts still rely on outcome-level supervision, leaving intermediate ideas and plans without direct credit, even though they determine the conceptual direction of algorithm design. In this paper, we propose Algorithm Tree Policy Optimization (ATPO), which utilizes tree-structured rollouts to decouple the generation of conceptual ideas from their specific code implementations. By applying group-relative advantage estimation, ATPO assigns credit to both strategic ideas and algorithmic implementations, thereby optimizing the policy across multiple levels of abstraction. Integration with FunSearch, EoH, and OpenEvolve demonstrates improvements over their original counterparts. Notably, the learned policy transfers effectively to unseen problem instances and alternative search methods, indicating that it can improve the generalization of algorithm design policies.

The automatic synthesis of a program from any form of specification is regarded as a holy grail of computer science. Fueled by LLMs, NL2Code has achieved tremendous success, yet the fundamentally more challenging task of synthesizing programs from input-output behavior, which we refer to as IO2Code, remains largely unsolved. Whereas NL2Code can exploit the semantic alignment between natural language and code acquired during pretraining, IO2Code requires recovering underlying principles from concrete computational behavior, navigating a vast and underspecified hypothesis space. To address this, we propose DIO-Agent, a discovery agent for IO2Code. Our method frames IO2Code as an evolutionary search over discrete program space, in which an LLM serves as the mutation operator and concrete error signals from execution guide each mutation. To prevent the search from wandering into structurally complex yet incorrect dead ends, we introduce the Transformation Priority Premise as a mutation prior that biases the LLM toward the simplest hypothesis consistent with current evidence, progressively escalating from constants to conditionals to iteration only when simpler constructs are insufficient. To facilitate systematic study, we further construct an IO2Code benchmark spanning multiple difficulty levels. Extensive experiments show that DIO-Agent consistently outperforms both traditional program-by-example methods and evolutionary search baselines across all difficulty levels and various LLMs, while substantially surpassing test-time scaling strategies with equivalent sampling budgets. Our code and dataset are available at \url{https://anonymous.4open.science/r/IO2Code}.

Large language model (LLM) agents are increasingly used to operate browsers, files, code and tools, making personal assistants a natural deployment target. Yet personal agents face a privacy-cost-capability tension: cloud models execute multi-step workflows well but expose sensitive intermediate context to external APIs, while local models preserve privacy but remain less reliable. Both settings also pay repeatedly for long skill prompts and growing histories. We propose constant-context skill learning, a context-to-weights framework for recurring agent workflows: reusable procedures are learned in lightweight task-family modules, while inference conditions only on the current observation and a compact state block. A deterministic tracker renders this state block from task progress and supplies aligned subgoal rewards, so each module can be trained with step-level SFT and refined through online RL. Across ALFWorld, WebShop, and SciWorld, our agents achieve strong performance across Qwen3-4B, Qwen3-8B and Llama-3.1-8B. With Qwen3-8B, SFT+RL reaches 89.6\% unseen success on ALFWorld, 76.8\% success on WebShop, and 66.4\% unseen success on SciWorld. They match or exceed strong published agent-training results while reducing prompt tokens per turn by 2-7$\times$ relative to controlled ReAct prompting baselines, showing that procedural context can be moved from prompts into weights.

The electrocardiogram (ECG) is a time series that records the electrical activity driving cardiac motion through electromechanical coupling (EMC), where the depolarization captured by ECG triggers each heartbeat. The temporal delay and low spatial resolution of such multichannel dynamic data often cause temporal and spatial misalignments along the physical motion, limiting the application of ECG data to event/segment-based downstream clinic practice. Although many efforts have been made to recover spatiotemporal trajectories conditioned on ECG waveforms, the intrinsic principle of EMC that results in various temporal delays remains underexplored in current generative models. To address this limitation, we propose Time2Track, a mechanics foundation model that generates individualized 4D heart motion from same-day ECG recordings by encoding electrophysiology as delay-aware token embeddings. Specifically, a learnable wavelet decomposition extracts the electromechanical delay, which is then formulated as a physics-based conditioning signal for motion diffusion. Time2Track generates 4D motion with individual-specific kinematics, avoiding the collapse to a population average, i.e., a common failure scenario when diffusing motions whose inter-subject differences are subtle. On $n=40,577$ subjects from the UK Biobank, our Time2Track accurately synthesizes 4D cardiac trajectories from ECG and preserves inter-subject diversity, a regime in which generic motion diffusion baselines collapse to a population average.

Prior work has shown that grokked Transformers implement discrete Fourier transforms or group character representations for modular arithmetic. We unify these findings under the Wedderburn--Artin decomposition: using interchange intervention accuracy (IIA), a causal method stronger than ablation, we show that grokked models decompose computation along the Wedderburn components of the target algebra. Across 11 commutative algebras over $\F_p$ ($p = 2,3,5,7$)---including 6 with nontrivial Jacobson radical---all grokked models achieve raw IIA $\geq 0.97$. When models predict incorrectly, errors remain compartmentalized by component ($3.8\times$ chance), confirming that independence is structural, not a byproduct of accuracy. Training dynamics reveal that Wedderburn alignment emerges synchronously with grokking. For non-commutative groups, models exhibit \emph{selective} Wedderburn alignment; non-commutativity raises a capacity threshold that larger models overcome.

Online Continual Learning (OCL) requires models to learn from non-stationary data streams under a strict single-pass constraint, making them highly susceptible to catastrophic forgetting. While existing studies have explored various strategies from data, architecture, and optimization perspectives, the optimizer often directly uses generic approaches such as SGD and AdamW. In this work, we reveal a new link between optimization dynamics and OCL needs by recasting Polyak-Ruppert averaging as the engine of "plasticity" and Primal averaging as the anchor of "stability". This inspires our novel optimizer, SPIN (**S**tability-**P**lasticity **IN**terpolation), which explicitly decouples plasticity and stability by interpolating between a fast-moving "plasticity" sequence and an adaptive "stability" sequence. We theoretically demonstrate that SPIN implicitly employs an inverse Hessian approximation, providing crucial Tikhonov regularization to damp noisy, single-sample Hessian estimates and mitigate forgetting. More importantly, SPIN can be seamlessly integrated existing OCL methods by taking resulted gradients as input and replacing standard optimizers. Extensive experiments show that using SPIN instead of standard optimizers yields consistent and significant performance gains across multiple OCL approaches, including the challenging rehearsal-free setting and under strict memory constraints.

Neural decoders have shown strong potential for error correction in short- and moderate-length regimes, yet a fundamental tension remains between decoding accuracy and inference latency. Recent attempts to leverage diffusion probabilistic models for channel decoding typically adopt a fully generative paradigm, initializing the reverse process from an observation-agnostic prior, which overlooks a key structural property of channel decoding: the received signal already contains substantial information about the target codeword. In this work, we propose Channel Residual Diffusion Model (ChRes-DM), a principled diffusion-based decoding framework that reinterprets channel decoding as a directed restoration process anchored at the noisy observation. Instead of sampling from a generic prior, ChRes-DM constructs a geometric residual bridge between the received signal and the clean codeword, explicitly modeling the conditional transport induced by the channel and leading to a deterministic Probability Flow Ordinary Differential Equation (PF-ODE) that governs the decoding dynamics. By eliminating redundant stochastic sampling inherent in conventional diffusion models, ChRes-DM enables efficient iterative decoding with flexible step-skipping, offering fine-grained control over the accuracy–latency trade-off. Extensive experiments across representative channel coding benchmarks demonstrate that ChRes-DM achieves competitive or superior decoding performance compared to existing neural decoders while significantly reducing inference iterations, highlighting diffusion-based residual transport as a promising and scalable paradigm for neural channel decoding.