ICLR 2026 Papers

Chain-of-Thought (CoT) reasoning has proven effective in enhancing Large Language Models (LLMs) on complex tasks by decomposing problems into step-wise solutions. However, extending CoT to multi-modal settings remains challenging, as it requires modeling transitions of visual states alongside textual reasoning. Existing approaches often underperform due to limited capacity to model visual transitions or fragmented architectures. To overcome this limitation, we introduce Uni-CoT, a Unified Chain-of-Thought framework that captures structured visual transitions and seamlessly aligns them with textual logic, enabling coherent multimodal reasoning. To mitigate the computational and training challenges inherent to multi-modal reasoning, Uni-CoT introduces a two-level reasoning paradigm: a macro-level CoT for high-level planning and a micro-level CoT for localized subtask execution. This hierarchical design reduces computational overhead while maintaining coherence. Additionally, Uni-CoT incorporates a structured training paradigm with auxiliary tasks to stabilize optimization and improve generalization. Experiments on reasoning-driven image generation and understanding benchmarks demonstrate that Uni-CoT achieves state-of-the-art performance and remarkable generalization, underscoring its potential for complex multi-modal reasoning.

We study multi-user contextual bandits where users are related by a graph and their reward functions exhibit both non-linear behavior and graph homophily. We introduce a principled joint penalty for the collection of user reward functions $\\{f_u\\}$, combining a graph smoothness term based on RKHS distances with an individual roughness penalty. Our central contribution is proving that this penalty is equivalent to the squared norm within a single, unified _multi-user RKHS_. We explicitly derive its reproducing kernel, which elegantly fuses the graph Laplacian with the base arm kernel. This unification allows us to reframe the problem as learning a single "lifted" function, enabling the design of principled algorithms, LK-GP-UCB and LK-GP-TS, that leverage Gaussian Process posteriors over this new kernel for exploration. We provide high-probability regret bounds that scale with an _effective dimension_ of the multi-user kernel, replacing dependencies on user count or ambient dimension. Empirically, our methods outperform strong linear and non-graph-aware baselines in non-linear settings and remain competitive even when the true rewards are linear. Our work delivers a unified, theoretically grounded, and practical framework that bridges Laplacian regularization with kernelized bandits for structured exploration.

Time series foundation models (TSFMs) are a potential class of powerful, general-purpose tools for forecasting and related temporal tasks, but their behavior is strongly shaped by subtle inductive biases in their design. Rather than developing a new model and claiming that it is better than existing TSFMs, e.g., by winning on existing benchmarks, our objective is to understand how the various "knobs" of the training process affect model quality. Using a mix of theory and controlled empirical evaluation, we identify and show how various design choices (e.g., patch size, embedding choice, training objective, etc.) lead to implicit biases in fundamental model properties (e.g., temporal behavior, geometric structure, how aggressively or not the model regresses to the mean, etc.), and how these biases can be intuitive or counterintuitive, depending on properties of the model and data. We illustrate in a case study on outlier handling how multiple biases interact in complex ways.

Recent advances in deep reinforcement learning (RL) have achieved strong results on high-dimensional control tasks, but applying RL to reachability problems raises a fundamental mismatch: reachability seeks to maximize the set of states from which a system remains safe indefinitely, while RL optimizes expected returns over a user-specified distribution. This mismatch can result in policies that perform poorly on low-probability states that are still within the safe set. A natural alternative is to frame the problem as a robust optimization over a set of initial conditions that specify the initial state, dynamics and safe set, but whether this problem has a solution depends on the feasibility of the specified set, which is unknown a priori. We propose Feasibility-Guided Exploration (FGE), a method that simultaneously identifies a subset of feasible initial conditions under which a safe policy exists, and learns a policy to solve the reachability problem over this set of initial conditions. Empirical results demonstrate that FGE learns policies with over 50% more coverage than the best existing method for challenging initial conditions across tasks in the MuJoCo simulator and the Kinetix simulator with pixel observations.

This paper introduces a novel watermarking method for diffusion models. It is based on guiding the diffusion process using the gradient computed from any off-the-shelf watermark decoder. The gradient is guided further using different image augmentations, increasing robustness to attacks against which the decoder was not originally robust, without retraining or fine-tuning. The methodology effectively allows to convert any post-hoc watermarking scheme into a scheme embedding the signal during the diffusion process. We show that this approach is complementary to watermarking techniques modifying the variational autoencoder at the end of the diffusion process. We validate the methods on different diffusion models and detectors. The watermarking guidance does not significantly alter the generated image for a given seed and prompt, preserving both the diversity and quality of generation.

Recent advances in diffusion and flow-based generative models have demonstrated remarkable success in image restoration tasks, achieving superior perceptual quality compared to traditional deep learning approaches. However, these methods either require numerous sampling steps to generate high-quality images, resulting in significant computational overhead, or rely on common model distillation, which usually imposes a fixed fidelity-realism trade-off and thus lacks flexibility. In this paper, we introduce OFTSR, a novel flow-based framework for one-step image super-resolution that can produce outputs with tunable levels of fidelity and realism. Our approach first trains a conditional flow-based super-resolution model to serve as a teacher model. We then distill this teacher model by applying a specialized constraint. Specifically, we force the predictions from our one-step student model for same input to lie on the same sampling ODE trajectory of the teacher model. This alignment ensures that the student model's single-step predictions from initial states match the teacher's predictions from a closer intermediate state. Through extensive experiments on datasets including FFHQ (256$\times$256), DIV2K, and ImageNet (256$\times$256), we demonstrate that OFTSR achieves state-of-the-art performance for one-step image super-resolution, while having the ability to flexibly tune the fidelity-realism trade-off. Code and pre-trained models will be publicly available.

We initiate a unified theoretical and algorithmic study of a key problem in weak-to-strong (W2S) generalization: when fine-tuning a strong pre-trained student with pseudolabels from a weaker teacher on a downstream task with spurious correlations, does W2S happen, and how to improve it upon failures? We consider two sources of spurious correlations caused by group imbalance: (i) a weak teacher fine-tuned on group-imbalanced labeled data with a minority group of fraction $\eta_\ell$, and (ii) a group-imbalanced unlabeled set pseudolabeled by the teacher with a minority group of fraction $\eta_u$. Theoretically, a precise characterization of W2S gain at the proportional asymptotic limit shows that W2S always happens with sufficient pseudolabels when $\eta_u = \eta_\ell$ but may fail when $\eta_u \ne \eta_\ell$, where W2S gain diminishes as $(\eta_u - \eta_\ell)^2$ increases. Our theory is corroborated by extensive experiments on various spurious correlation benchmarks and teacher-student pairs. To boost W2S performance upon failures, we further propose a simple, effective algorithmic remedy that retrains the strong student on its high-confidence data subset after W2S fine-tuning. Our algorithm is group-label-free and achieves consistent, substantial improvements over vanilla W2S fine-tuning.

While spiking neural networks (SNNs) provide a biologically inspired and energy-efficient computational framework, their robustness and the dynamic advantages inherent to biological neurons remain significantly underutilized owing to oversimplified neuron models. In particular, conventional leaky integrate-and-fire (LIF) neurons often omit the dynamic conductance mechanisms inherent in biological neurons, thereby limiting their capacity to cope with noise and temporal variability. In this work, we revisit dynamic conductance from a functional perspective and uncover its intrinsic role as a bio-inspired gating mechanism that modulates information flow. Building on this insight, we introduce the Dynamic Gated Neuron~(DGN), a novel spiking unit in which membrane conductance evolves in response to neuronal activity, enabling selective input filtering and adaptive noise suppression. We provide a theoretical analysis showing that DGN possess enhanced stochastic stability compared to standard LIF models, with dynamic conductance intriguingly acting as a disturbance rejection mechanism. DGN-based SNNs demonstrate superior performance across extensive evaluations on anti-noise tasks and temporal-related benchmarks such as TIDIGITS and SHD, consistently exhibiting excellent robustness. To the best of our knowledge, for the first time, our results establish bio-inspired dynamic gating as a key mechanism for robust spike-based computation, providing not only theoretical guarantees but also strong empirical validations. This work thus paves the way for more resilient, efficient, and biologically inspired spiking neural networks.

We revisit Proximal Policy Optimization (PPO) from a function-space perspective. Our analysis decouples policy evaluation and improvement in a reproducing kernel Hilbert space (RKHS): (i) A kernelized temporal-difference (TD) critic performs efficient RKHS-gradient updates using only one-step state–action transition samples. (ii) a KL-regularized, natural-gradient policy step exponentiates the evaluated action-value, recovering a PPO/TRPO-style proximal update in continuous state-action spaces. We provide non-asymptotic, instance-adaptive guarantees whose rates depend on RKHS entropy, unifying tabular, linear, Sobolev, Gaussian, and Neural Tangent Kernel (NTK) regimes, and we derive a sampling rule for the proximal update that ensures the optimal $k^{-1/2}$ convergence rate for stochastic optimization. Empirically, the theory-aligned schedule improves stability and sample efficiency on common control tasks (e.g., CartPole, Acrobot), while our TD-based critic attains favorable throughput versus a GAE baseline. Altogether, our results place PPO on a firmer theoretical footing beyond finite-dimensional assumptions and clarify when RKHS-proximal updates with kernel-TD critics yield global policy improvement with practical efficiency.

The rapid progress of large language models (LLMs) has laid the foundation for multimodal models. Nevertheless, visual language models (VLMs) still face significant computational overhead when scaled from images to the video domain. When video data is too large (due to high frame rates and long durations), the inference cost of models increases sharply. This severely hinders their deployment and application in environments that require rapid responses and have limited computation resources. Token compression for input videos is one of the promising directions, as effective compression schemes can significantly reduce computational overhead. Most existing compression methods are based on training-free token merging strategies in either the spatial or temporal dimension. Although these methods reduce computational overhead, their training-free nature inevitably leads to information loss during token compression, resulting in a significant performance drop. To address these challenges, we propose a Memory-Augmented Reinforcement Learning-based Token Compression (MARC) method for efficient video understanding that integrates structured retrieval with RL-based distillation. Our proposed MARC is a retrieve-then-compress method, which employs a Visual Memory Retriever (VMR) tool and a Compression Group Relative Policy Optimization (C-GRPO) training strategy. The Visual Memory Retriever first segments videos into event-level fragments and selects query-relevant clips. The C-GRPO distills reasoning ability from a Teacher Network to a Student Network by encouraging the output of the student network to match the performance of the teacher network. Extensive experiments on six video benchmarks demonstrate that our compression method achieves nearly identical accuracy to the 64-frame Qwen2.5-VL-3B baseline while using only one frame’s worth of tokens as input, resulting in a 95% reduction in visual tokens. Moreover, our approach reduces GPU memory usage by 72% and generation latency by 23.9%. These results demonstrate the strong potential of our compression method as a robust solution for RL-based post-training compression of large-scale models, enabling practical deployment in latency-sensitive and resource-constrained applications such as real-time video question answering, surveillance, and autonomous driving.

While large and diverse datasets have driven recent advances in large models, identifying the optimal data mixture for pre-training and post-training remains a significant open problem. We address this challenge with FastMix, a novel framework that automates data mixture discovery while training only a single proxy model. Instead of relying on predefined heuristics or resource-intensive simulations, FastMix jointly optimizes mixture coefficients and model parameters, substantially improving efficiency and scalability over prior approaches. At the core of FastMix is a reformulation of mixture selection as a bilevel optimization problem. Under this reformulation, we show that optimizing mixture ratios is mathematically equivalent to assigning per-source loss weights under uniform source sampling. This embeds the mixture coefficients directly into the differentiable iterative optimization objective, enabling efficient, gradient-based optimization of both mixture and model. To solve the optimization problem, FastMix implements an approximate iterative optimization procedure, alternating between (i) updating model parameters on data sampled according to current mixture ratios (inner loop) and (ii) updating mixture ratios based on validation feedback (outer loop). Across pre- and post-training, FastMix outperforms baselines while drastically reducing search cost: in pre-training, it attains an average score of 48.2 with 1.3 GPU-hours ($\times 550$ vs. RegMix; $\times 55$ vs. CLIMB), and in post-training (SFT) it leads with 65.4 with a $+5.5$ gain over the next best, completing search in 2.2 GPU-hours compared to the 115 GPU-hours required by CLIMB/RegMix.

Recent diffusion models have achieved remarkable success in image relighting, and this success has quickly been reproduced in video relighting. Although these methods can relight videos under various conditions, their ability to explicitly control the illumination in the relighted video remains limited. Therefore, we present LightCtrl, the first controllable video relighting method that offers explicit control over the video illumination through a user-supplied light trajectory in a training-free manner. This is essentially achieved by leveraging a hybrid approach that combines pre-trained diffusion models: a pre-trained image relighting diffusion model is used to relight each frame individually, followed by a video diffusion prior that enhances the temporal consistency of the relighted sequence. In particular, to enable explicit control over dynamically varying lighting in the relighted video, we introduce two key components. First, the Light Map Injection module samples light trajectory-specific noise and injects it into the latent representation of the source video, significantly enhancing illumination coherence with respect to the conditional light trajectory. Second, the Geometry-Aware Relighting module dynamically combines RGB and normal map latents in the frequency domain to suppress the influence of the original lighting in the input video, thereby further improving the relighted video's adherence to the input light trajectory. Our experiments demonstrate that LightCtrl can generate high-quality video results with diverse illumination changes closely following the light trajectory condition, indicating improved controllability over baseline methods. The code will be released to facilitate future studies.

In many applications, treatment and outcome are confounded by unobservables, yet mediators remain unconfounded. The front‑door (FD) adjustment identifies causal effects through mediators even with unmeasured confounding. However, most estimators focus on *average treatment effects*, and work on *heterogeneous treatment effect* (HTE) estimation remain scarce. We address this gap with two *debiased* learners for heterogeneous FD effects: *FD‑DR‑Learner* and *FD‑R‑Learner*. Both attain fast, quasi-oracle rates (i.e., performance comparable to an oracle that knows the nuisances) even when nuisance functions converge as slowly as $n^{-1/4}$. Beyond theory, we demonstrate fast convergence and debiasedness in synthetic and real-world evaluations. Our results show that the proposed learners deliver robust and debiased HTE estimates under the FD scenario.

Large Language Models (LLMs) often hallucinate, generating content inconsistent with the input. Retrieval-Augmented Generation (RAG) and Reinforcement Learning with Human Feedback (RLHF) can mitigate hallucinations but require resource-intensive retrieval or large-scale fine-tuning. Decoding-based methods are lighter yet lack explicit hallucination control. To address this, we present \textbf{Token-Guard}, a token-level hallucination control method based on self-checking decoding. Token-Guard performs internal verification at each reasoning step to detect hallucinated tokens before they propagate. Candidate fragments are further evaluated in a latent space with explicit hallucination risk scoring, while iterative pruning and regeneration dynamically correct detected errors. Experiments on HALU datasets show Token-Guard substantially reduces hallucinations and improves generation accuracy, offering a scalable, lightweight solution for reliable LLM outputs. Our code is publicly available\footnote{\url{https://github.com/rhq945/Token-Guard}}.

Diffusion models are powerful generative models for continuous signals, such as images and videos. However, real-world digital data are quantized; hence, they take not continuous values but only a finite set of discrete values. For example, pixels in 8‑bit images can take only 256 discrete values. In existing diffusion models, quantization is either ignored by treating data as continuous, or handled by adding small noise to make the data continuous. Neither approach guarantees that samples from the model will converge to the finite set of quantized points. In this work, we propose a methodology to explicitly account for quantization within diffusion models. Specifically, by adopting a particular form of parameterization, we guarantee that samples from the reverse diffusion process converge to quantized points. In experiments, we demonstrate that our quantization-aware model can substantially improve the performance of diffusion models for density estimation, and achieve state‑of‑the‑art results on pixel‑level image generation in likelihood evaluation. In particular, for CIFAR‑10 image generation, the negative log‑likelihood improves substantially from 2.42 to 0.27, approaching the theoretical lower bound.

The nature of the relationship between Bayesian sampling and stochastic gradient descent in neural networks has been a long-standing open question in the theory of deep learning. We shed light on this question by modeling the long runtime behaviour of SGD as diffusion on porous media. Using singular learning theory, we show that the late stage dynamics are strongly impacted by the degeneracies of the loss surface. From this we are able to show that under reasonable choices of hyperparameters for vanilla SGD, the local steady state distribution of SGD (if it exists) is effectively a tempered version of the Bayesian posterior over the weights which accounts for local accessibility constraints.

Existing dominant methods for audio generation include Generative Adversarial Networks (GANs) and diffusion-based methods like Flow Matching. GANs suffer from slow convergence and potential mode collapse during training, while diffusion methods require multi-step inference that introduces considerable computational overhead. In this work, we introduce Flow2GAN, a two-stage framework that combines Flow Matching training for learning generative capabilities with GAN fine-tuning for efficient few-step inference. Specifically, given audio's unique properties, we first improve Flow Matching for audio modeling through: 1) reformulating the objective as endpoint estimation, avoiding velocity estimation difficulties when involving empty regions; 2) applying spectral energy-based loss scaling to emphasize perceptually salient quieter regions. Building on these Flow Matching adaptations, we demonstrate that a further stage of lightweight GAN fine-tuning enables us to obtain one-step generator that produces high-quality audio. In addition, we develop a multi-branch network architecture that processes Fourier coefficients at different time-frequency resolutions, which improves the modeling capabilities compared to prior single-resolution designs. Experimental results indicate that our Flow2GAN delivers high-fidelity audio generation from Mel-spectrograms or discrete audio tokens, achieving better quality-efficiency trade-offs than existing state-of-the-art GAN-based and Flow Matching-based methods. Online demo samples are available at \url{https://flow2gan.github.io}.

Causal representation learning has attracted significant research interest during the past few years, as a means for improving model generalization and robustness. Causal representations of interventional image pairs (also called ``actionable counterfactuals'' in the literature), have the property that only variables corresponding to scene elements affected by the intervention / action are changed between the start state and the end state. While most work in this area has focused on identifying and representing the variables of the scene under a causal model, fewer efforts have focused on representations of the interventions themselves. In this work, we show that an effective strategy for improving out of distribution (OOD) robustness is to focus on the representation of actionable counterfactuals in the latent space. Specifically, we propose that an intervention can be represented by a Causal Delta Embedding that is invariant to the visual scene and sparse in terms of the causal variables it affects. Leveraging this insight, we propose a method for learning causal representations from image pairs, without any additional supervision. Experiments in the Causal Triplet challenge demonstrate that Causal Delta Embeddings are highly effective in OOD settings, significantly exceeding baseline performance in both synthetic and real-world benchmarks.

Based on the prevalent intermediate fusion (IF) and late fusion (LF) frameworks, multimodal representation learning (MML) demonstrates its superiority over unimodal representation learning. To investigate the intrinsic factors underlying the empirical success of MML, research grounded in theoretical justifications from the perspective of generalization error has emerged. However, these provable MML studies derive the theoretical findings based on LF, while theoretical exploration based on IF remains scarce. This naturally gives rise to a question: **Can we design a comprehensive MML approach supported by the sufficient theoretical analysis across fusion types?** To this end, we revisit the IF and LF paradigms from a fine-grained dimensional perspective. The derived theoretical evidence sufficiently establishes the superiority of IF over LF under a specific constraint. Based on a general $K$-Lipschitz continuity assumption, we derive the generalization error upper bound of the IF-based methods, indicating that eliminating the distribution incoherence can improve the generalizability of IF-based MML methods. Building upon these theoretical insights, we establish a novel IF-based MML method, which introduces the informatic constraint and performs distribution cohering. Extensive experimental results on multiple widely adopted datasets verify the effectiveness of the proposed method.