ICML 2026 Papers

Process rewards have been widely used in deep reinforcement learning to improve training efficiency, reduce variance, and prevent reward hacking. In LLM reasoning, existing works also explore various solutions for learning effective process reward models (PRM) with or without the help of an expert policy. However, existing methods either rely on strong assumptions about the expert policies (e.g., requiring their reward functions) or suffer intrinsic limitations (e.g., entropy collapse), resulting in weak PRMs or limited generalizability. In this paper, we introduce rePIRL, an inverse RL-inspired framework that learns effective PRMs with minimal assumptions about expert policies. Specifically, we design a dual learning process that updates the policy and the PRM interchangeably. Our learning algorithm has customized techniques to address the challenges of scaling traditional inverse RL to LLMs. We theoretically show that our proposed learning framework can unify both online and offline PRM learning methods, justifying that rePIRL can learn PRMs with minimal assumptions. Empirical evaluations on standardized math and coding reasoning datasets demonstrate the effectiveness of rePIRL over existing methods. We further show the application of our trained PRM in test-time training, test-time scaling, and providing an early signal for training hard problems. Finally, we validate our training recipe and key design choices via a detailed ablation study.

We introduce a predictive model that estimates the pre-training loss of large models from model size ($N$), batch size ($B$) and number of weight updates ($K$). This is the first loss prediction model that can handle changing batch size. The model outperforms Chinchilla's loss model, a model of the test loss using the batch size and number of tokens, in terms of projecting the loss at extrapolated compute budgets (up to 1000 folds). A natural use of the model is to find optimal $N,B,K$ configurations under explicit and compound resource constraints like time, memory and compute. In our experiments, the model-selected configurations are close to ground-truth optimal. Our work advocates for loss prediction as a better alternative to heuristic-based laws, which are growing in complexity. The implementation is available on \href{https://github.com/chuningxdy/Noisy-Quadratic-System}{GitHub}.

Segmentation based on language has been a popular topic in computer vision. While recent advances in multimodal large language models (MLLMs) have endowed segmentation systems with reasoning capabilities, these efforts remain confined by the frozen internal knowledge of MLLMs, which limits their potential for real-world scenarios that involve up-to-date information or domain-specific concepts. In this work, we propose \textbf{Seg-ReSearch}, a novel segmentation paradigm that overcomes the knowledge bottleneck of existing approaches. By enabling interleaved reasoning and external search, Seg-ReSearch empowers segmentation systems to handle dynamic, open-world queries that extend beyond the frozen knowledge of MLLMs. To effectively train this capability, we introduce a hierarchical reward design that harmonizes initial guidance with progressive incentives, mitigating the dilemma between sparse outcome signals and rigid step-wise supervision. For evaluation, we construct OK-VOS, a challenging benchmark that explicitly requires outside knowledge for video object segmentation. Experiments on OK-VOS and two existing reasoning segmentation benchmarks demonstrate that our Seg-ReSearch improves state-of-the-art approaches by a substantial margin. Code and data is available at https://github.com/iSEE-Laboratory/Seg-ReSearch.

Analogy is a central faculty of human intelligence, enabling abstract patterns discovered in one domain to be applied to another. However, the mechanisms underlying analogical reasoning in Transformers remain poorly understood. In this work, inspired by the notion of functors in category theory, we formalize analogical reasoning as the inference of correspondences between entities across categories. Based on this formulation, we introduce synthetic tasks that evaluate the emergence of analogical reasoning under controlled settings. We find that the emergence of analogical reasoning is highly sensitive to data characteristics, optimization choices, and model scale. Through mechanistic analysis, we show that analogical reasoning in Transformers decomposes into two key components: (1) geometric alignment of relational structure in the embedding space, and (2) the application of a functor within the Transformer. These mechanisms enable models to transfer relational structure from one category to another, realizing analogy. Finally, we quantify these effects and find that the same trends are observed in pretrained LLMs. In doing so, we move analogy from an abstract cognitive notion to a concrete, mechanistically grounded phenomenon in modern neural networks.

Large Language Models (LLMs) often struggle with complex logical reasoning. Existing approaches typically rely on either purely neural reasoning in natural language or offloading to formal solvers via symbolic representations. However, both paradigms face significant limitations: while LLMs exhibit strong semantic intuition they are prone to hallucinations, whereas symbolic solvers offer rigorous derivation but remain highly sensitive to minor syntactic errors. To combine the strengths of these two paradigms while mitigating their respective limitations, we introduce **LogicSAGE** (**L**ogic-informed **S**ocratic **A**gent for **G**uided **E**nhancement), a dual-process framework that integrates a robust neural reasoner (System 1) with a rigorous symbolic validator (System 2). Specifically, our framework employs a Socratic Error Correction mechanism that treats solver feedback not as terminal failures but as pedagogical signals, engaging in a dialectic loop to iteratively refine logic programs and resolve semantic ambiguities. Extensive experiments on five benchmarks show that LogicSAGE (8B) achieves a state-of-the-art 92.36% average accuracy, significantly outperforming GPT-4 baselines, which establishes that architectural innovation can supersede model scale in faithful reasoning.

RL methods for scaling large reasoning models stall on datasets with low initial success rates, and thus little training signal. We investigate a fundamental question:Can a pretrained LLM leverage latent knowledge to generate an automated curriculum for problems it cannot solve? We explore this with SOAR: An asymmetric self-play framework that uses meta-RL to surface these pedagogical signals. A teacher model proposes synthetic problems for a student model, and is rewarded with its improvement on a subset of hard problems, thus grounding the curriculum in real student progress rather than intrinsic proxy rewards. Our study on the hardest subsets of math benchmarks (0/128 success) reveals three core findings. First, it is possible to realize bilevel meta-RL that unlocks learning under sparse, binary rewards by sharpening a latent capacity of pretrained models to generate useful problems. Second, grounded rewards outperform intrinsic learnability rewards used in prior LLM self-play, reliably avoiding typical instability and diversity collapse modes. Third, the structure and well-posedness of questions are more critical for learning progress than solution correctness. Our results suggest that the ability to generate useful stepping stones does not require the preexisting ability to solve the hard problems, paving a principled path to escape reasoning plateaus without additional curated data.

Printed-Circuit-board (PCB) footprint geometry labeling of integrated circuits (IC) is essential in defining the physical interface between components and the PCB layout, requiring precise visual perception. However, the unstructured nature of footprint drawings and abstract diagram annotations prevents direct IC footprint parsing and automated package geometry labeling methods from developing. Existing Large Multimodal Models (LMMs) struggle with inaccurate geometric perception, limiting their effectiveness in this task. To address these challenges, we propose LMM4-IC4K, a novel framework that treats IC mechanical drawings as images and leverages LMMs for structured geometric interpretation. To support such a framework, we introduce ICGeo8K, a multi-modal dataset with 8,608 labeled samples, including 4138 real-world IC footprint samples and 4470 synthetically generated samples. We further present a two-stage training framework to fine-tune LMMs for IC footprint labeling. Extensive experiments demonstrate that our model outperforms state-of-the-art LMMs on the proposed benchmark. The accurate translation of footprint diagrams enabled by LMM4-IC4K contributes to advancing automation and standardization within the PCB industry.

Distilling long-form reasoning from teacher models into smaller students requires selecting which candidate solutions to train on. Recent work argues that one should select responses the student model assigns highest probability, i.e., favoring solutions ``natural'' to the student. However, we find that this approach works within a single teacher but fails when scaling to long reasoning traces from multiple diverse teachers. We identify a key cause: this approach scores entire solutions, but students generalize by recombining familiar reasoning steps, not by memorizing complete solutions. Full-trajectory scoring optimizes the wrong target; it rewards global fluency while the transferable signal lies in local step transitions. We propose Local Average Log Probability (LALP), which scores each reasoning step using only a small window of preceding context, measuring whether each step is justified by its immediate premises rather than whether the full response looks natural to the student. LALP enables two practical use cases: selecting the best teacher before fine-tuning and curating training data from diverse teacher pools. Across math, coding, and science reasoning tasks, LALP consistently improves accuracy when selecting the most natural solutions by a large margin.

Math reasoning has become the poster child of progress in large language models (LLMs), with new models rapidly surpassing human-level performance on benchmarks like MATH and AIME. But as math leaderboards improve week by week, it is worth asking: do these gains reflect broader problem-solving ability or just narrow overfitting? To answer this question, we evaluate over 20 open-weight reasoning-tuned models across a broad suite of tasks, including math, scientific QA, agent planning, coding, and standard instruction-following. We surprisingly find that most models that succeed in math fail to transfer their gains to other domains. To rigorously study this phenomenon, we conduct controlled experiments using math-only data with two widely-used methods: Reinforcement Learning (RL) and Supervised Finetuning (SFT) with detailed ablations. On top of the observation that RL-tuned models transfer better than SFT-tuned models, we identify on-policy fine-tuning as the key mechanism underlying cross-domain transfer, regardless of whether the training signal comes from RL or supervised learning. Latent-space representation and token-space distribution shift analyses reveal that off-policy SFT induces substantial representation and output drift, while on-policy RL preserves general-domain structure. Our results suggest a need to rethink the post-training recipes, particularly the reliance on off-policy SFT-distilled data to advance reasoning models.

The integration of Large Language Models (LLMs) with automated theorem proving has shown immense promise, yet is constrained by challenges in scaling up both training-time reinforcement learning (RL) and inference-time compute. This paper introduces BFS-Prover-V2, a step-level theorem proving system designed to address this dual scaling problem. We present two primary innovations. The first is a novel multi-turn off-policy RL framework for continually improving the performance of the LLM step-prover at training time. This framework, inspired by the principles of AlphaZero, utilizes a multi-stage expert iteration pipeline featuring adaptive tactic-level data filtering and periodic retraining to surmount the performance plateaus that typically curtail long-term RL in LLM-based agents. The second innovation is a planner-enhanced multi-agent system that scales reasoning capabilities at inference time. This architecture employs a general reasoning model as a high-level planner to iteratively decompose complex theorems into a sequence of simpler subgoals. This hierarchical approach substantially reduces the search space, enabling a team of parallel prover agents to collaborate efficiently by leveraging a shared proof cache. We demonstrate that this dual approach to scaling yields state-of-the-art results on established formal mathematics benchmarks. BFS-Prover-V2 achieves 95.08% and 41.4% on the miniF2F and ProofNet test sets respectively. While demonstrated in the domain of formal mathematics, the RL and inference techniques presented in this work are of broader interest and may be applied to other domains requiring long-horizon multi-turn reasoning and complex search.

Large Language Model (LLM) agents have shown significant potential in automated optimization modeling for mathematical problems. However, real-world problems are still challenging due to their knowledge-intensive nature. Existing methods, constrained by static parametric knowledge, often lack the domain expertise required to comprehend complex scenarios and apply appropriate mathematical techniques, leading to errors. To address this challenge, we propose the Opt-Miner framework, where the agent learns to identify missing knowledge, retrieve technical documents on the web, and ground its mathematical models for improved modeling performance. The core of Opt-Miner is a novel tree-guided data synthesis pipeline coupled with a retrieval-based group relative policy optimization (R-GRPO) algorithm, designed to foster the agent’s information-seeking capabilities. Specifically, we first formulate each problem into a tree structure, with its scenario contexts and mathematical techniques embedded in subtrees. We then employ subtree union, transfer, and knowledge fogging to synthesize complex, multi-domain problems that incorporate knowledge gaps, thereby necessitating active information seeking to solve these problems. Based on synthesized data, we propose R-GRPO for agent reinforcement learning. Experiments demonstrate that Opt-Miner-Qwen3-8B achieves performance comparable to 32B state-of-the-art specialized agents and commercial reasoning models.

Learning discrete neural samplers is challenging due to the lack of gradients and combinatorial complexity. While stochastic optimal control (SOC) and Schrödinger bridge (SB) provide principled solutions, efficient SOC solvers like adjoint matching (AM), which excel in continuous domains, remain unexplored for discrete spaces. We bridge this gap by revealing that the core mechanism of AM is *state-space agnostic*, and introduce **discrete ASBS**, a unified framework that extends AM and adjoint Schrödinger bridge sampler (ASBS) to discrete spaces. Theoretically, we analyze the optimality conditions of the discrete SB problem and its connection to SOC, identifying a necessary cyclic group structure on the state space to enable this extension. Empirically, discrete ASBS achieves competitive sample quality with significant advantages in training efficiency and scalability. Our code is available at https://github.com/AlexandreGUO2001/DASBS.

Efficient distillation is a key pathway for converting expensive reasoning capability into deployable efficiency, yet in the frontier regime where the student already has strong reasoning ability, naive continual distillation often yields limited gains or even degradation. We observe a characteristic training phenomenon: even as loss decreases monotonically, all performance metrics can drop sharply at almost the same bottle-neck, before gradually recovering. We further uncover a token-level mechanism: confidence bifurcates into steadily increasing Imitation-Anchor Tokens that quickly anchor optimization and other yet-to-learn tokens whose confidence is suppressed until after the bottleneck. And the characteristic that these two types of tokens cannot coexist is the root cause of the failure in continual distillation. To this end, we propose Training-Trajectory-Aware Token Selection (T3S) to reconstruct the training objective at the token level, clearing the optimization path for yet-to-learn tokens. T3S yields consistent gains in both AR and dLLM settings: with only hundreds of examples, Qwen3-8B surpasses DeepSeek-R1 on competitive reasoning benchmarks, Qwen3-32B approaches Qwen3-235B, and T3-trained LLaDA-2.0-Mini exceeds its AR baseline, achieving state-of-the-art performance among all of 16B-scale no-think models.

Scaling test-time compute with multi-path chain-of-thought can improve reasoning accuracy, but its gains hinge on an effective exploration–exploitation trade-off. Existing methods handle this trade-off in rigid ways: tree-structured search hard-codes exploration via brittle expansion rules that disrupt post-trained reasoning, while parallel reasoning over-explores redundant hypothesis paths and relies on a weak answer selection strategy. Driven by the insight that the optimal balance is *phase-dependent* and that correct vs. incorrect paths often *diverge only at late stages*, we reconceptualize test-time scaling as a dynamic *expand–reduce* control problem over a pool of hypothesis paths. We introduce **HyPER**, a *training-free online control policy* for MoE multi-path decoding that reallocates compute under a fixed budget using lightweight path statistics. HyPER features (i) an *online controller* that shifts from exploration to exploitation as the hypothesis pool evolves, (ii) an MoE-based token-level refinement primitive for efficient *generation-time exploitation* without full-path resampling, and (iii) a length- and confidence-aware aggregation rule to bridge the existence–selection gap for reliable *answer-time exploitation*. Extensive experimental results across four MoE models and diverse benchmarks demonstrate HyPER consistently achieves the accuracy–compute Pareto frontier, outperforming prior-art methods by 8-10% while reducing token consumption by 25-40%.

With the rise of reasoning language models and test-time scaling methods as a paradigm for improving model performance, substantial computation is often required to generate multiple candidate sequences from the same prompt. This enables exploration of different reasoning paths toward the correct solution, however, allocates the same compute budget for each prompt. Grounded on the assumption that different prompts carry different degrees of complexity, and thus different computation needs, we propose EAGer, a training-free generation method that leverages model uncertainty through token-wise entropy distribution to reduce redundant computation and concurrently improve overall performance. EAGer allows branching to multiple reasoning paths only in the presence of high-entropy tokens, and reallocates the saved compute budget to instances where exploration of alternative paths is most needed. We validate EAGer across multiple open-source models on complex reasoning benchmarks, with gains specifically demonstrated on AIME 2025. When target labels are accessible -- as in RLVR training pipelines -- EAGer achieves up to +37% in Pass@k and 59% fewer tokens; in test-time settings it still yields +12% in Pass@k and 64% fewer tokens compared to Full Parallel Sampling.

Looping, reusing a block of layers across depth, and depth growing, training shallow-to-deep models by duplicating middle layers, have both been linked to stronger reasoning, but their relationship remains unclear. We provide a mechanistic unification: looped and depth-grown models exhibit convergent depth-wise signatures, including increased reliance on late layers and recurring patterns aligned with the looped or grown block. These shared signatures support the view that their gains stem from a common form of iterative computation. Building on this connection, we show that the two techniques are adaptable and composable: applying inference-time looping to the middle blocks of a depth-grown model improves accuracy on some reasoning primitives by up to $2\times$, despite the model never being trained to loop. Both approaches also adapt better than the baseline when given more in-context examples or additional supervised fine-tuning data. Additionally, depth-grown models achieve the largest reasoning gains when using higher-quality, math-heavy cooldown mixtures, which can be further boosted by adapting a middle block to loop. Overall, our results position depth growth and looping as complementary, practical methods for inducing and scaling iterative computation to improve reasoning.

Diffusion-based image synthesis has made AI-generated images (AIGI) increasingly photorealistic, raising urgent concerns about authenticity in applications such as misinformation detection, digital forensics, and content moderation. Despite the substantial advances in AIGI detection, how to correct detected AI-generated images with visible artifacts and restore realistic appearance remains largely underexplored. Moreover, few existing work has established the connection between AIGI detection and artifact correction. To fill this gap, we propose GenShield, a unified autoregressive framework that jointly performs explainable AIGI detection and controllable artifact correction in a closed loop from diagnosis to restoration, revealing a mutually reinforcing relationship between these two tasks. We further introduce a Visual Chain-of-Thought based curriculum learning strategy that enables self-explained, multi-step "diagnose-then-repair" correction with an explicit stopping criterion. A high-quality dataset with large-scale "artifact-restored" pairs is also constructed alongside a unified evaluation pipeline. Extensive experiments on our correction benchmark and mainstream AIGI detection benchmarks demonstrate state-of-the-art performance and strong generalization of our method.

Inference-time scaling via chain-of-thought (CoT) reasoning is a major driver of state-of-the-art LLM performance, but it comes with substantial latency and compute costs. We address a fundamental theoretical question: *how many* reasoning tokens are required to solve a problem as input size grows? By extending the bounded attention prefix oracle (BAPO) model--an abstraction of LLMs that quantifies the information flow required to solve a task--we prove lower bounds on the CoT tokens required for three canonical BAPO-hard tasks: binary majority, triplet matching, and graph reachability. We show that each requires $\Omega(n)$ reasoning tokens when the input size is $n$. We complement these results with matching or near-matching upper bounds via explicit constructions. Finally, our experiments with frontier reasoning models show approximately linear reasoning token scaling on these tasks and failures when constrained to smaller reasoning budgets, consistent with our theoretical lower bounds. Together, our results identify fundamental bottlenecks in inference-time compute through CoT and offer a principled tool for analyzing optimal reasoning length.

Large Language Models (LLMs) have shown promising results in automating formal verification. However, existing approaches treat proof generation as a static, end-to-end prediction over source code, relying on limited verifier feedback and lacking access to concrete program behaviors. We present ExVerus, a counterexample-guided framework that enables LLMs to reason about proofs using behavioral feedback via counterexamples. When a proof fails, ExVerus automatically generates and validates counterexamples, and then guides the LLM to generalize them into inductive invariants to block these failures. Our evaluation shows that ExVerus significantly improves proof accuracy, robustness, and token efficiency over the state-of-the-art prompting-based Verus proof generator.