ICML 2026 Papers

Offline-to-Online Reinforcement Learning (O2O-RL) leverages an offline, pre-trained policy to minimize costly online interactions. Although data-efficient, O2O-RL is susceptible to shifts between offline and online distributions. Existing work aims to mitigate the harm of this shift by finetuning the policy on trajectory data sampled from a diffusion model. Inspired by this line of work, we propose DUAL: an efficient **D**iffusion **U**ncertainty-**A**ware framework for offline-to-online reinforcement **L**earning. DUAL utilizes the prior knowledge of the diffusion model to distill a fast-sampling diffusion actor policy and transition model in the offline phase. DUAL also employs a Laplace approximation and distance transition-state-shift detection, thereby using uncertainty quantification to improve exploration versus exploitation in the online phase. We formally show that our actor loss with the Laplace approximation provides a proxy for a principled estimate of epistemic uncertainty. Empirically, DUAL improves the online expected return over O2O-RL baselines across multiple settings and environments.

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.

Federated Multi-Objective Learning (FMOL) enables collaborative training of conflicting objectives but faces a compounded challenge: the recursive coupling between intra-task client drift and inter-task aggregation bias. We propose DREAM, a unified framework that jointly corrects these two coupled error sources through drift-aware control variates and momentum-smoothed local updates. On the server side, DREAM formulates multi-objective aggregation as a regularized quadratic program parameterized by a task correction matrix, which provides a generalized formulation that can flexibly adapt to scalarization, prioritization, and gradient manipulation strategies. Theoretically, we establish a linear speedup convergence rate of $\mathcal{O}(1/\sqrt{NT})$ for non-convex objectives. We further provide theoretical guarantees for the conflict-avoidant direction distance. In the strongly convex setting, DREAM achieves convergence in weighted sub-optimality and admits a unified Lyapunov analysis showing linear convergence to a regularization-dependent neighborhood. Numerical experiments on representative benchmarks validate the effectiveness of DREAM in multi-objective optimization.

Flaw reporting for deployed AI systems is fundamental to identifying system failures and improving AI safety. Yet the AI reporting ecosystem is fragmented: researchers who identify flaws often do not know what or where to report, and groups who receive reports rarely share them with other relevant stakeholders. As a result, good-faith reporters duplicate effort by submitting many different forms, and recipients lack standardized, triage-ready information. We survey 12 reporting systems published by AI developers, cybersecurity groups, and AI flaw aggregators, identifying five recurring design challenges spanning discoverability, scope, information collection, coordination, and guidance for strict-liability cases. Building on this analysis and feedback from 49 experts across 32 organizations representing developers, security researchers, and ecosystem coordinators, we introduce FLARE-AI, an open-source AI flaw reporting system designed for interoperability with existing systems. FLARE-AI streamlines flaw report creation by collecting triage-relevant information through conditional logic and early classification, then enables optional dissemination of standardized, machine-readable reports to multiple developers, coordinators, and incident registries from a single submission. By lowering barriers to reporting AI flaws and improving interoperability across stakeholders, FLARE-AI helps break down silos and accelerate remediation across the AI ecosystem.

We study first-order methods for solving monotone variational inequalities arising in min-max optimization. Classical approaches such as the extragradient method rely on two gradient queries per iteration, which limits their analysis and applicability in the online and stochastic settings. We propose a family of Generalized Optimistic Methods with Anchoring (GOMA), which combine two-time-scale optimistic updates with an anchoring term inspired by Halpern iteration. In the deterministic setting, GOMA achieves the optimal accelerated last-iterate rate $\mathcal{O}(1/k^2)$ on the squared gradient norm for monotone Lipschitz operators. In the stochastic setting with unbounded variance, a simplified single-call variant of GOMA achieves a last-iterate convergence rate of $\mathcal{O}(1/\sqrt{k})$ on the squared gradient norm. To the best of our knowledge, this is the first such guarantee for stochastic monotone Lipschitz VIs in the unconstrained setting without variance reduction or growing batches.

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.

Reinforcement Learning (RL) is pivotal for enhancing Large Language Model (LLM) reasoning, yet mainstream algorithms such as GRPO and DAPO remain constrained by a coarse-grained credit assignment paradigm, where all tokens within the same response receive the identical reward. In this paper, we propose **Dynamic Entropy Weighting**, systematically define entropy-based weight ratios $\frac{H_{i,t}}{\sum_{k=1}^{n} H_{k,t}}$ and similar variants to redistribute rewards and get fine-grained rewards through two new algorithms: **Group Token Policy Optimization (GTPO)**, which assigns an entropy-weighted reward to each token and synthesizes token-specific advantage function to drive the model toward optimal path, and the analogous algorithm **Sequence-Level GRPO (GRPO-S)**, which admits a completely similar design at the sequence level. Unlike methods using entropy as mere regularization, GTPO and GRPO-S establish a new state-of-the-art on AIME and MATH 500, outperforming prior entropy-guided baselines and validating our weighting mechanism.

In performative learning, the data distribution reacts to the deployed model—for example, because strategic users adapt their features to game it—which creates a more complex dynamic than in classical supervised learning. One should thus not only optimize the model for the current data but also take into account that the model might steer the distribution in a new direction, without knowing the exact nature of the potential shift. We explore how regularization can help cope with performative effects by studying its impact in high-dimensional ridge regression. We show that, while performative effects worsen the test risk in the population setting, when moving to the over-parameterized regime where the number of features exceeds the number of samples, the optimal regularization in the presence of performativity helps reduce the variance in the estimated parameters, thereby improving performance. We show that the optimal regularization scales with the overall strength of the performative effect, making it possible to set the regularization in anticipation of this effect. We illustrate this finding through empirical evaluations of the optimal regularization parameter on both synthetic and real-world datasets.

Evaluating representation similarity is fundamental to representation learning. However, existing metrics suffer from significant limitations: they lack interpretability due to shifting baselines, lack robustness to outliers, and are computationally intractable for large datasets, forcing reliance on heuristic approximations. To address this, we develop an ordinal-similarity framework, instantiated by the Triplet (TSI) and Quadruplet (QSI) Similarity Indices, which measure alignment by quantifying the consistency of ordinal relationships. We theoretically demonstrate this formulation is inherently interpretable, robust to outliers, and computationally efficient. Finally, we establish a formal equivalence between TSI and local neighborhood alignment, measured by Mutual Nearest Neighbors. Empirically, we validate these properties and show that ordinal similarity offers a scalable approach to measuring alignment, enabling practitioners to better understand and design representations.

While deep learning has revolutionized financial forecasting through sophisticated architectures, the design of the supervision signal itself is rarely scrutinized. We challenge the canonical assumption that training labels must strictly mirror inference targets, uncovering the Label Horizon Paradox: the optimal supervision signal often deviates from the prediction goal, shifting across intermediate horizons governed by market dynamics. We theoretically ground this phenomenon in a dynamic signal-noise trade-off, demonstrating that generalization hinges on the competition between marginal signal realization and noise accumulation. To operationalize this insight, we propose a bi-level optimization framework that autonomously identifies the optimal proxy label within a single training run. Extensive experiments on large-scale financial datasets demonstrate consistent improvements over conventional baselines, thereby opening new avenues for label-centric research in financial forecasting.

While Large Language Model (LLM) agents demonstrate proficiency in static benchmarks, their deployment in real-world scenarios is hindered by the dynamic nature of user queries, tool sets, and interaction dynamics.To address this generalization gap, we formalize **OpenAgent** (Tool-Use Agent in Open-World), a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions.To systematically diagnose its impact, we construct a controlled sandbox environment where we define fine-grained environmental shifts across a four-tier hierarchy, *Perception*, *Interaction*, *Reasoning*, and *Internalization*, and conduct a comprehensive series of experiments. Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning (SFT) and Reinforcement Learning suffer from varying degrees of performance degradation when confronting open environmental shifts.Building on these insights, we propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments. Our code will be released at: [https://github.com/LAMDA-NeSy/OpenAgent](https://github.com/LAMDA-NeSy/OpenAgent).

Cross-domain offline reinforcement learning leverages a source dataset to improve policy learning in a data-scarce target domain, but dynamics mismatch makes many source transitions kinematically infeasible and can cause negative transfer. Recent non-parametric geometric methods (e.g., standard optimal transport and k-nearest neighbors) avoid overfitting yet often yield only relative rankings under an implicit matching or retrieval budget, making performance sensitive to hand-tuned thresholds when the true cross-domain overlap is unknown. We formulate availability estimation as soft subset selection by learning a source reweighting that geometrically aligns with the target. We propose **R**obust **O**ffline unbalanced **O**ptimal **T**ransport (ROOT): (i) a robust ambiguity set for uncertainty under limited target samples, and (ii) an unbalanced transport objective that penalizes mass deviation, enabling a principled transport-or-discard mechanism. ROOT thus down-weights or discards high-cost source samples rather than forcing them onto the target support. Moreover, the induced weights decay exponentially with transport cost, guaranteeing outlier suppression. On D4RL dynamics-shift benchmarks, ROOT improves downstream offline RL and outperforms strong baselines on most tasks without task-specific threshold tuning.

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%.

Consider the problem of finding the best matching in a weighted graph where we only have access to predictions of the actual stochastic weights, based on an underlying context. If the predictor is the Bayes optimal one, then computing the best matching based on the predicted weights is optimal. However, in practice, this perfect information scenario is not realistic. Given an imperfect predictor, a suboptimal decision rule may compensate for the induced error and thus outperform the standard optimal rule. In this paper, we propose multicalibration as a way to address this problem. This fairness notion requires a predictor to be unbiased on each element of a family of protected sets of contexts. Given a class of matching algorithms $\mathcal C$ and any predictor $\gamma$ of the edge-weights, we show how to construct a specific multicalibrated predictor $\hat \gamma$, with the following property. Picking the best matching based on the output of $\hat \gamma$ is competitive with the best decision rule in $\mathcal C$ applied onto the original predictor $\gamma$. We complement this result by providing sample complexity bounds, and by performing numerical experiments.

This work proposes a framework LGKDE that learns kernel density estimation for graphs. The key challenge in graph density estimation lies in effectively capturing both structural patterns and semantic variations while maintaining theoretical guarantees. Combining graph kernels and kernel density estimation (KDE) is a standard approach to graph density estimation, but has unsatisfactory performance due to the handcrafted and fixed features of kernels. Our method LGKDE leverages graph neural networks to represent each graph as a discrete distribution and utilizes maximum mean discrepancy to learn the graph metric for multi-scale KDE, where all parameters are learned by maximizing the density of graphs relative to the density of their well-designed perturbed counterparts. The perturbations are conducted on both node features and graph spectra, which helps better characterize the boundary of normal density regions. Theoretically, we establish consistency and convergence guarantees for LGKDE, including bounds on the mean integrated squared error, robustness, and generalization. We validate LGKDE by demonstrating its effectiveness in recovering the underlying density of synthetic graph distributions and applying it to graph anomaly detection across diverse benchmark datasets. Extensive empirical evaluation shows that LGKDE demonstrates superior performance compared to state-of-the-art baselines on most benchmark datasets.

Pursuing training-free open-vocabulary semantic segmentation in an efficient and generalizable manner remains challenging due to the deep-seated spatial bias in CLIP. To overcome the limitations of existing solutions, this work moves beyond the CLIP-based paradigm and harnesses the recent spatially-aware dino.txt framework to facilitate more efficient and high-quality dense prediction. While dino.txt exhibits robust spatial awareness, we find that the semantic ambiguity of text queries gives rise to severe mismatch within its dense cross-modal interactions. To address this, we introduce VIsual-guided Prompt evolution (VIP) to rectify the semantic expressiveness of text queries in dino.txt, unleashing its potential for fine-grained object perception. Towards this end, VIP integrates alias expansion with a visual-guided distillation mechanism to mine valuable semantic cues, which are robustly aggregated in a saliency-aware manner to yield a high-fidelity prediction. Extensive evaluations demonstrate that VIP: I) surpasses the top-leading methods by 1.4%~8.4\% average mIoU, II) generalizes well to diverse challenging domains, and III) requires marginal inference time and memory overhead.

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.

Most reinforcement learning (RL) baselines maximize future cumulative rewards with a fixed single discount factor, which limits their performance in complex sequential decision-making tasks due to a failure to balance short-term objectives and long-term planning. To address this issue, this paper focuses on a multi-timescale critic framework, where each component corresponds to a Q-value with a distinct discount factor. Two key improvements are proposed: (1) A Neural Reward Decoder reconstructs the reward sequence from multi-scale Q-values, with value and reward reconstruction losses enhancing Q-value estimation consistency; (2) A cross-attention-based Q-weight predictor adaptively adjusts Q-value weights via current observations to generate the final Q-value for policy optimization. Extensive experiments on DMControl and CARLA benchmarks demonstrate that our method significantly outperforms state-of-the-art (SOTA) baselines. Furthermore, we validate the framework's generalizability by integrating it with both off-policy (SAC, DrQ-v2) and on-policy (PPO) algorithms, achieving consistent performance gains. The code is available in the supplementary material.

AI predictive systems increasingly support high-stakes decision making, yet robust decisions under uncertainty often rely on human capabilities beyond AI alone. This motivates collaborative approaches that combine human judgment with AI predictions. We study this problem through the lens of uncertainty quantification and introduce **Human-AI Collaborative Uncertainty Quantification**, a framework in which an AI system refines a human expert’s proposed prediction set subject to two principles: **counterfactual harm**, requiring that the AI not degrade correct human judgments, and **complementarity**, requiring recovery of correct outcomes the human missed. At the population level, we show that the optimal collaborative prediction set has a simple two-threshold structure over a single score function, governing pruning and augmentation relative to the human proposal. Building on this characterization, we develop offline and online calibration algorithms with **distribution-free** finite-sample guarantees. The online algorithm adapts to arbitrary distribution shifts, including settings where human behavior evolves through interaction with the AI. Empirically, we show that collaborative prediction sets outperform human-only and AI-only baselines, achieving improved coverage--efficiency tradeoffs across image classification, regression, and text-based medical decision making.