ICML 2026 Papers

Large language models (LLMs) are increasingly used as assistive interfaces for users who cannot reliably produce clean text due to accessibility constraints, yet existing evaluations assume iterative input repair and focus on task accuracy or generic noise robustness. We introduce Assistive Prompt Mediation (APM), a theory-grounded evaluation paradigm that reframes assistance as a constrained mediation problem: recovering latent user intent from accessibility-impaired input without clarification, while minimizing cognitive burden and hallucination risk. APM decomposes assistive quality along these axes and is instantiated across 8 languages, 4 accessibility-driven noise classes, and 10 frontier LLMs, with impairment severity yielding accessibility sensitivity curves. Results show that apparent robustness often masks trade-offs—high intent preservation frequently coincides with increased burden or hallucinated mediation, hallucination rates vary by more than $2\times$ across noise types, and assistive decisions exhibit bounded entropy ($<0.81$ normalized), indicating systematic rather than unstable behavior. These findings demonstrate that standard robustness metrics substantially overestimate assistive reliability and motivate evaluating LLMs as constrained mediators under accessibility-driven input degradation.

Accurate prediction in survival analysis with competing risks is challenged by rare event rates and limited effective sample sizes. Knowledge distillation offers a promising way to transfer information from an external teacher to improve a local student, but existing methods are overwhelmingly developed for uncensored outcomes and do not directly extend to survival analysis, where censored observations provide only partial information. Moreover, prior work often assumes that teacher and student share identical outcome definitions, whereas in competing risks settings, they may differ in outcome granularity and event definitions, further complicating knowledge transfer. To address these gaps, we propose DiSKD (Discrete Survival Knowledge Distillation), a deep learning framework for discrete-time competing risks that integrates teacher predictions via a cause-specific, time-dependent Kullback-Leibler divergence. DiSKD enables flexible and privacy-conscious transfer without requiring raw data sharing, remains robust to model misspecification or outcome-definition heterogeneity, and adaptively weights teacher guidance by emphasizing compatible teachers while down-weighting less relevant ones. Simulation studies and real-world applications demonstrate improved discrimination and calibration.

Outlier detection (OD) is widely used in practice; but its effective deployment on new tasks is hindered by lack of labeled outliers, which makes algorithm and hyperparameter selection notoriously hard. Foundation models (FMs) have transformed ML, and OD is no exception: Shen et al. (2025) introduced FOMO-0D, the first FM for OD, achieving remarkable performance against numerous baselines. This work introduces OUTFORMER, which advances FOMO-0D with (1) a mixture of synthetic priors and (2) self-evolving curriculum training. OUTFORMER is pretrained solely on synthetic labeled datasets and infers test labels of a new task by using its training data as in-context input. Inference is fast and zero-shot, requiring merely forward pass and no labeled outliers. Thanks to in-context learning, it requires zero additional work—no OD model training or bespoke model selection—enabling truly plug-and-play deployment. OUTFORMER achieves state-of-the-art performance on the prominent ADBench, as well as two new large-scale OD benchmarks that we introduce, comprising over 1,500 datasets, while maintaining speedy inference.

Personalized federated learning has emerged as a popular approach to training on devices holding statistically heterogeneous data, known as clients. However, most existing approaches require a client to have labeled data for training or finetuning in order to obtain their own personalized model. In this paper we address this by proposing FLowDUP, a novel method that is able to generate a personalized model using only a forward pass with unlabeled data. The generated model parameters reside in a low-dimensional subspace, enabling efficient communication and computation. FLowDUP's learning objective is theoretically motivated by our new transductive multi-task PAC-Bayesian generalization bound, that provides performance guarantees for unlabeled clients. The objective is structured in such a way that it allows both clients with labeled data and clients with only unlabeled data to contribute to the training process. To supplement our theoretical results we carry out a thorough experimental evaluation of FLowDUP, demonstrating strong empirical performance on a range of datasets with differing sorts of statistically heterogeneous clients. Through numerous ablation studies, we test the efficacy of the individual components of the method.

Transformers are arguably the preferred architecture for language generation. In this paper, inspired by continued fractions, we introduce a new function class for generative modeling. The architecture family implementing this function class is named CoFrGeNets - Continued Fraction Generative Networks. We design novel architectural components based on this function class that can replace Multi-head Attention and Feed-Forward Networks in Transformer blocks while requiring much fewer parameters. We derive custom gradient formulations to optimize the proposed components more accurately and efficiently than using standard PyTorch-based gradients. Our components are a plug-in replacement requiring little change in training or inference procedures that have already been put in place for Transformer-based models thus making our approach easy to incorporate in large industrial workflows. We experiment on two very different transformer architectures GPT2-xl (1.5B) and Llama3 (3.2B), where the former we pre-train on OpenWebText and GneissWeb, while the latter we pre-train on the docling data mix which consists of nine different datasets. Results show that the performance on downstream classification, Q\& A, reasoning and text understanding tasks of our models is competitive and sometimes even superior to the original models with two thirds to half the parameters and shorter pre-training time. We believe that future implementations customized to hardware will further bring out the true potential of our architectures.

Learning from demonstration can be an effective way to teach robots task-oriented policies. However, in an interactive setting when demonstrations are limited by time or other budgetary constraints, it is challenging to find those that fix the learner's (remaining) errors. This is especially difficult for novice teachers: they may provide task-valid trajectories, often these fail to meaningfully improve the policy due to their lack of knowledge of learning mechanisms internal to the robot. This paper introduces CLASP (Collaborative Learning with Anchored State-space Partitions), which summarises the teaching process as a compact map of behavioural regions anchored in the teacher's own demonstrations. The map connects task failure to actionable changes to demonstrations by indicating what is going wrong in an intuitive way. It also enables difficulty-aware training that emphasises regions where learning is failing. Across diverse benchmarks, CLASP improves success by up to 20\% over offline and interactive baselines under the same demonstration budget, improves robustness under distribution shift by 14–20\%, and preserves behavioural diversity.

LLM-based deep research agents are largely built on the ReAct framework. This linear design makes it difficult to revisit earlier states, branch into alternative search directions, or maintain global awareness under long contexts, often leading to local optima, redundant exploration, and inefficient search. We propose Re-TRAC, an agentic framework that performs cross-trajectory exploration by generating a structured state representation after each trajectory to summarize evidence, uncertainties, failures, and future plans, and conditioning subsequent trajectories on this state representation. This enables iterative reflection and globally informed planning, reframing research as a progressive process. Empirical results show that Re-TRAC consistently outperforms ReAct by 15–20% on BrowseComp with frontier LLMs. For smaller models, we introduce Re-TRAC-aware supervised fine-tuning, achieving state-of-the-art performance at comparable scales. Notably, Re-TRAC shows a monotonic reduction in tool calls and token usage across rounds, indicating progressively targeted exploration driven by cross-trajectory reflection rather than redundant search.

Texture recognition is often framed as matching an image to a static training-set dictionary or codebook. In practice, this assumption is brittle: label-preserving transformations (illumination, scale, compression, blur) can shift test features away from the fixed training dictionary, producing a training-set codebook misalignment that limits accuracy. We propose STD-Former (Simple Texture Dictionary Transformer), a lightweight framework for image-conditioned texture dictionary encoding. Instead of comparing against a static codebook, STD-Former extracts a compact set of Intrinsic Textons (dictionary atoms / codewords) from the input image itself, yielding self-aligned representations at inference. Our design is intentionally simple and uses a decoupled two-stage recipe. In Stage 1, a Texture Dictionary Extractor (TDE) is pre-trained with a self-supervised Texton Coverage Loss that encourages the learned textons to collectively cover the image patch feature manifold. In Stage 2, a classifier is trained on the encoded dictionary representation; optionally, we add a Sparse Topological Loss derived from 0D persistent homology, which is equivalent to supervising only the (B-1) edges of a minimum spanning tree (MST) in each batch, providing efficient structure regularization. Across six standard texture benchmarks, STD-Former and STD-Former+ achieve new state-of-the-art results.

The dynamics of gradient-based training in neural networks often exhibit nontrivial structures; hence, understanding them remains a central challenge in theoretical machine learning. In particular, the concept of feature unlearning, in which a neural network progressively loses previously learned features over long training, has gained attention. In this study, we consider the infinite-width limit of a two-layer neural network trained with a large-batch stochastic gradient, then derive differential equations with different time scales, revealing the mechanism and conditions for feature unlearning to occur. Specifically, we utilize the fast-slow dynamics: while an alignment of first-layer weights develops rapidly, the second-layer weights develop slowly. The direction of the flow on a critical manifold, determined by the slow dynamics, decides whether feature unlearning occurs. We give numerical validation of the result and derive theoretical grounding and scaling laws for the feature unlearning. Our results yield the following insights: (i) the strength of the primary nonlinear term in the data induces the feature unlearning, and (ii) an initial scale of the second-layer weights mitigates the feature unlearning. Our result should be understood as a population loss of alignment rather than finite-sample overfitting. Technically, our analysis utilizes Tensor Programs and singular perturbation theory.

Large language models (LLMs) have been globally adopted in various scenarios, making robust multilingual safety alignment a prerequisite for their reliable deployment across diverse languages. Despite recent advances, LLMs exhibit a substantial safety gap between high- and low-resource languages: models that can consistently refuse harmful requests in high-resource languages often fail to do so in low-resource languages. In this work, we reveal that such safety failures stem from insufficient representation-space separability between harmful and harmless prompts in low-resource languages. Through geometric analyses, we find that, compared to English, harmful prompts are significantly less separated from the manifold of harmless prompts, and that the resulting cross-lingual spatial margin gap is strongly correlated with attack success rates. Capitalizing on these insights, we propose Multilingual Spatial Margin Gap-based Optimization (SMO), a novel training strategy that exploits the well-aligned safety geometry of a dominant language (e.g., English) to enhance safety alignment in other languages. SMO explicitly leverages the spatial margin gap between English and target languages as an example-wise supervision signal, enabling effective cross-lingual transfer of safety capabilities while preserving the dominant language’s original performance. Experiments conducted on LLaMA-3.1-8B-Instruct and Qwen2.5-7B-Instruct demonstrate that SMO is capable of substantially reducing attack success rates in low-resource languages to near zero, often reaching zero, while maintaining strong general multilingual performance. Warning: This paper contains content that may be harmful.

We introduce vector diffusion wavelets (VDWs), a novel family of wavelets inspired by the vector diffusion maps algorithm that was introduced to analyze data lying in the tangent bundle of a Riemannian manifold. We show that these wavelets may be effectively incorporated into a family of geometric graph neural networks, which we refer to as VDW-GNNs. We demonstrate that such networks are effective on synthetic point cloud data, as well as on real-world data derived from wind field and neural activity measurements. Theoretically, we prove that these new wavelets have desirable frame theoretic properties, similar to traditional diffusion wavelets. Additionally, we prove that these wavelets have useful symmetries with respect to rotations and translations.

Modern AI systems are typically developed through multiple stages: pretraining, fine-tuning rounds, and subsequent adaptation or alignment, where each stage builds on the previous ones and updates the model in distinct ways. This raises a critical question of accountability: when a deployed model succeeds or fails, which stage is responsible, and to what extent? We pose the accountability attribution problem for tracing model behavior back to specific stages of the model development process. To address this challenge, we propose a general framework that answers counterfactual questions about stage effects: how would the model's behavior have changed if the updates from a particular stage had not occurred? Within this framework, we introduce estimators that efficiently quantify stage effects without retraining the model, accounting for both the data and key aspects of model optimization dynamics, including learning rate schedules, momentum, and weight decay. We demonstrate that our approach successfully quantifies the accountability of each stage to the model's behavior. Based on the attribution results, our method can identify stages associated with spurious correlations in image classification and text toxicity detection tasks and guide targeted follow-up interventions. Our approach provides a practical tool for model analysis and represents a significant step toward more accountable AI development.

Reinforcement Learning with Verifiable Rewards (RLVR) is an effective paradigm for improving the reasoning capabilities of large language models. However, existing RLVR methods utilize rollouts in an indiscriminate and short-horizon manner: responses of heterogeneous quality within each prompt are treated uniformly, and historical rollouts are discarded after a single use. This leads to noisy supervision, poor sample efficiency, and suboptimal policy updates. We address these issues by formulating rollout scheduling in RLVR as a contextual bandit problem and proposing a unified neural scheduling framework that adaptively selects high-value rollouts throughout training. Each rollout is treated as an arm whose reward is defined by the induced performance gain between consecutive optimization steps. The resulting scheduler supports both noise-aware intra-group selection and adaptive global reuse of historical rollouts within a single principled framework. We provide theoretical justification by deriving sublinear regret bounds and showing that enlarging the rollout buffer improves the achievable performance upper bound. Experiments on six mathematical reasoning benchmarks demonstrate consistent gains in performance and training efficiency across multiple RLVR optimization methods.

Test-time compute scaling allocates inference computation uniformly, uses fixed sampling strategies, and applies verification only for reranking. In contrast, we propose a verifier-guided adaptive framework treating reasoning as iterative trajectory generation and selection. For each problem, the agent runs multiple inference iterations. In each iteration, it optionally produces a high-level plan, selects a set of reasoning tools and a compute strategy together with an exploration parameter, and then generates a candidate reasoning trajectory. A process reward model (PRM) serves as a unified control signal: within each iteration, step-level PRM scores are aggregated to guide pruning and expansion during generation, and across iterations, aggregated trajectory rewards are used to select the final response. Across datasets, our dynamic, PRM-guided approach consistently outperforms direct test-time scaling, yielding large gains on MATH-500 and several-fold improvements on harder benchmarks such as AIME24 and AMO-Bench. We characterize efficiency using theoretical FLOPs and a compute intensity metric penalizing wasted generation and tool overhead, demonstrating that verification-guided allocation concentrates computation on high-utility reasoning paths.

Hyperspectral image change detection (HSI-CD) suffers from severe annotation scarcity and complex change patterns, which fundamentally limit the effectiveness of directly fine-tuning pre-trained foundation models. Although semi-supervised learning provides a promising direction, existing approaches mainly rely on confidence-based pseudo-label selection, leading to limited data diversity or severe error propagation. In this paper, we propose Influence-Aware Semi-supervised Fine-tuning (IA-SFT), a novel framework that evaluates the influence of pseudo-labels on model decision behavior to identify truly valuable supervision signals. Instead of confidence-based selection, IA-SFT evaluates each low-confidence pseudo-label by measuring its impact on labeled data, enabling reliable filtering of high-value pseudo-labels with minimal noise. To further adapt foundation models to HSI-CD, we design an Adaptive Fusion Change Decoder (AFCD) that jointly models global semantic consistency and local change details. Extensive experiments on three benchmark datasets demonstrate that IA-SFT consistently improves pseudo-label quality and detection performance, achieving superior accuracy compared to state-of-the-art methods. Additional analyses validate the transferability of IA-SFT when integrated into different frameworks in a plug-and-play manner. Code will be released.

Dataset distillation (DD) aims to replace a full training set with a tiny synthetic one, yet current theories neither explain why heterogeneous matching objectives (gradient, distribution, trajectory) work nor provide a quantitative boundary for robustness under configuration changes (optimizer, architecture, augmentation). We propose **configuration-dynamics-error (CDE) analysis** for a broad class of matching-based DD methods, a unified generalization framework that treats a training configuration as an update operator inducing optimization dynamics and measures distillation robustness by the test-risk gap between models trained on distilled versus full data. Within this framework, gradient, distribution, and trajectory matching reduce the same dynamics-induced risk gap, explaining why these heterogeneous objectives can all support dataset distillation. CDE yields two predictive laws: within a fixed configuration, the gap decays as $\mathcal{O}(k^{-1/2})$ with the distilled set size $k$ until a configuration-dependent floor, explaining IPC saturation and indicating when reducing the floor is more valuable than enlarging $k$. Across configurations, an order-tight coverage law formalizes the utility boundary: the required $k$ grows linearly with the configuration diversity captured by covering-number complexity. Experiments with representative DD methods and configuration changes exhibit predictive behavior consistent with both laws.

Flow matching (FM) has shown promising results in data-driven planning. However, it inherently lacks formal guarantees for ensuring state and action constraints, whose satisfaction is a fundamental and crucial requirement for the safety and admissibility of planned trajectories on various systems. Moreover, existing FM planners do not ensure the dynamical consistency, which potentially renders trajectories inexecutable. We address these shortcomings by proposing SAD-Flower, a novel framework for generating $\textbf{S}$afe, $\textbf{A}$dmissible, and $\textbf{D}$ynamically consistent trajectories. Our approach relies on an augmentation of the flow with a virtual control input. Thereby, principled guidance can be derived using techniques from nonlinear control theory, providing formal guarantees for state constraints, action constraints, and dynamic consistency. Crucially, SAD-Flower operates without retraining, enabling test-time satisfaction of unseen constraints. Through extensive experiments across several tasks, we demonstrate that SAD-Flower outperforms various generative-model-based baselines in ensuring constraint satisfaction. Video and demos can be found at [sadflowerplanning.github.io](https://sadflowerplanning.github.io/project_web/).

Robotic manipulation benefits from foundation models that describe goals, but today's agents still lack a principled way to learn from their own mistakes. We ask whether natural language can serve as feedback, an error-reasoning signal that helps embodied agents diagnose what went wrong and correct course. We introduce LAGEA (Language Guided Embodied Agents), a framework that turns episodic, schema-constrained reflections from a vision language model (VLM) into temporally grounded guidance for reinforcement learning. LAGEA summarizes each attempt in concise language, localizes the decisive moments in the trajectory, aligns feedback with visual state in a shared representation, and converts goal progress and feedback agreement into bounded, step-wise shaping rewardswhose influence is modulated by an adaptive, failure-aware coefficient. This design yields dense signals early when exploration needs direction and gracefully recedes as competence grows. On the Meta-World MT10 and Robotic Fetch embodied manipulation benchmarks, LAGEA improves average success over the state-of-the-art (SOTA) methods by 9.0% on random goals, 5.3% on fixed goals, and 17% on fetch tasks, while converging faster. These results support our hypothesis: language, when structured and grounded in time, is an effective mechanism for teaching robots to self-reflect on mistakes and make better choices. Code will be released soon.

Due to limited resources and public safety concerns, deep reinforcement learning (RL) agents for many cyber-physical systems (e.g., autonomous vehicles) are first trained in simulators. However, when deployed in real world environments, they often suffer from performance degradation or safety violations because of the inevitable Sim2Real gap. Existing zero-shot approaches, such as robust safe RL and domain randomization, mitigate this issue but typically at the cost of degraded performance or residual safety risks when experiencing unmodeled system dynamics. To address these limitations, we propose a novel reinforcement learning framework that enables safe and efficient policy transfer via probabilistic latent embeddings and dynamic policy adaptation. We consider a family of Constrained Markov Decision Processes (CMDPs) under different environment contexts. By leveraging latent context variable in meta-RL, the proposed framework infers the latent representation of the environment from simulated experiences. Furthermore, it incorporates a distributional RL formulation, which allows risk levels of the deployed policy to be adjusted dynamically, based on the estimation accuracy of the latent context variable. This strategy promotes safety at the early deployment stage and improves efficiency through fast policy adaptation under the Sim2Real gap.