ICLR 2026 论文列表

Multimodal large language models (MLLMs) have advanced rapidly in recent years. However, existing approaches for vision tasks often rely on indirect representations, such as generating coordinates as text for detection, which limits performance and prevents dense prediction tasks like segmentation. To overcome these challenges, we introduce Patch-as-Decodable Token (PaDT), a unified paradigm that enables MLLMs to directly generate both textual and diverse visual outputs. Central to PaDT are Visual Reference Tokens (VRTs), derived from visual patch embeddings of query images and interleaved seamlessly with LLM's output textual tokens. A lightweight decoder then transforms LLM's outputs into detection, segmentation, and grounding predictions. Unlike prior methods, PaDT processes VRTs independently at each forward pass and dynamically expands the embedding table, thus improving localization and differentiation among similar objects. We further tailor a training strategy for PaDT by randomly selecting VRTs for supervised fine-tuning and introducing a robust per-token cross-entropy loss. Our empirical studies across four visual perception and understanding tasks suggest PaDT consistently achieving state-of-the-art performance, even compared with significantly larger MLLM models.

Diffusion models have demonstrated remarkable image generation capabilities, but their performance heavily relies on classifier-free guidance (CFG). While CFG significantly enhances image quality, evaluating both conditional and unconditional models at every denoising step leads to substantial computational overhead. Existing approaches mitigate this cost through distillation, training a student network to learn the guided predictions. In contrast, we take an orthogonal approach by refining the \textit{initial Gaussian noise}, a critical yet under-explored factor in the diffusion-based generation pipelines. Recent studies have explored noise optimization for specific tasks such as layout-conditioned generation and human preference alignment. However, whether refined noise alone can enable guidance-free high-quality image generation remains an open question. We introduce a noise refinement framework where a refining network is trained to minimize the difference between images generated by unguided sampling from the refined noise and those produced by guided sampling from the input Gaussian noise. Our method achieves CFG-like quality without modifying the diffusion model, preserving its prior knowledge and compatibility with techniques like DreamBooth LoRA. Additionally, the learned refining network generalizes across domains without retraining and seamlessly integrates with existing distilled models, further improving sample quality. Beyond its practical benefits, we provide an in-depth analysis of refined noise, offering insights into its role in the denoising process and its interaction with guidance. Our findings suggest that structured noise initialization is key to efficient and high-fidelity image synthesis. Code and weights will be publicly released.

We study the sample complexity of online reinforcement learning in the general setting of nonlinear dynamical systems with continuous state and action spaces. Our analysis accommodates a large class of dynamical systems ranging from a finite set of nonlinear candidate models to models with bounded and Lipschitz continuous dynamics, to systems that are parametrized by a compact and real-valued set of parameters. In the most general setting, our algorithm achieves a policy regret of $\mathcal{O}(N \epsilon^2 + \mathrm{ln}(m(\epsilon))/\epsilon^2)$, where $N$ is the time horizon, $\epsilon$ is a user-specified discretization width, and $m(\epsilon)$ measures the complexity of the function class under consideration via its packing number. In the special case where the dynamics are parametrized by a compact and real-valued set of parameters (such as neural networks, transformers, etc.), we prove a policy regret of $\mathcal{O}(\sqrt{N p})$, where $p$ denotes the number of parameters, recovering earlier sample-complexity results that were derived for linear time-invariant dynamical systems. While this article focuses on characterizing sample complexity, the proposed algorithms are likely to be useful in practice, due to their simplicity, their ability to incorporate prior knowledge, and their benign transient behaviors.

Reinforcement learning (RL) has become central to enhancing reasoning in large language models (LLMs). Yet on-policy algorithms such as Group Relative Policy Optimization (GRPO) often suffer in early training: noisy gradients from low-quality rollouts lead to unstable updates and inefficient exploration. We introduce Slow-Fast Policy Optimization (SFPO), a simple yet efficient mechanism to address the above limitations via decomposing each iteration into three stages: a short fast trajectory of inner steps on the same batch, a reposition step to control off-policy drift, and a final slow correction. This reposition-before-update design preserves the objective and rollout process unchanged, making SFPO plug-compatible with existing policy-gradient pipelines. Extensive experiments demonstrate that SFPO consistently improves stability, reduces number of rollouts, and accelerates convergence of reasoning RL training. Specifically, it outperforms GRPO by up to 2.80 points on math reasoning benchmarks. It also achieves up to 4.93$\times$ fewer rollouts and a 4.19$\times$ reduction in wall-clock time to match GRPO’s best accuracy.

Although LLMs have demonstrated improved performance by scaling parallel test-time compute, doing so relies on generating reasoning paths that are both diverse and accurate. For challenging problems, the forking tokens that trigger diverse yet correct reasoning modes are typically deep in the sampling tree. Consequently, common strategies to encourage diversity, such as temperature scaling, encounter a worsened trade-off between diversity and accuracy. Motivated by this challenge, we treat parallel reasoning as a set-of-next-token-prediction problem and incorporate a set-based global loss into Supervised Fine-Tuning (SFT) using bipartite matching between global forking tokens and unique reasoning traces. We observe that, whereas naive fine-tuning with multiple reasoning traces collapses these unique reasoning modes, our proposed method, Set Supervised Fine-Tuning (SSFT), preserves these modes and produces emergent global forking tokens. Experiments on multiple reasoning benchmarks show our SSFT method consistently outperforms SFT under both pass@1 and cons@k metrics.

Tokenization---the process of decomposing a given text into a sequence of subwords called tokens---is one of the key components in the development of language models. Particularly, auto-regressive language models generate texts token by token, i.e., by predicting the next-token distribution given the previous ones, and thus tokenization directly affects their efficiency in text generation. Since each language model has their own vocabulary as a set of possible tokens, they struggle to cooperate with each other at the level of next-token distributions such as model ensemble. In this paper, we establish a theoretical framework of lossless vocabulary reduction, which efficiently converts a given auto-regressive language model into the one with an arbitrarily small vocabulary without any loss in accuracy. As an application, we demonstrate that language models with different tokenization can cooperate with each other efficiently through their maximal common vocabulary.

Environments built for people are increasingly operated by a new class of economic actors: LLM-powered software agents making decisions on our behalf. These decisions range from our purchases to travel plans to medical treatment selection. Current evaluations of these agents largely focus on task competence, but we argue for a deeper assessment: how these agents choose when faced with realistic decisions. We introduce ABxLab, a framework for systematically probing agentic choice through controlled manipulations of option attributes and persuasive cues. We apply this to a realistic web-based shopping environment, where we vary prices, ratings, and psychological nudges, all of which are factors long known to shape human choice. We find that agent decisions shift predictably and substantially in response, revealing that agents are strongly biased choosers even without being subject to the cognitive constraints that shape human biases. This susceptibility reveals both risk and opportunity: risk, because agentic consumers may inherit and amplify human biases; opportunity, because consumer choice provides a powerful testbed for a behavioral science of AI agents, just as it has for the study of human behavior. We release our framework as an open benchmark for rigorous, scalable evaluation of agent decision-making.

Process reward models (PRMs) are a cornerstone of test-time scaling (TTS), designed to verify and select the best responses from large language models (LLMs). However, this promise is challenged by recent benchmarks where simple majority voting, which ignores PRM signals, occasionally outperforms standard PRM-based selection. This raises a critical question: How can we effectively utilize verification signals from PRMs for TTS? To address this, we start by developing a theoretical framework for optimally combining signals from both the LLM and the PRM. Our framework reveals that the optimal strategy is a weighted aggregation of responses, a strategy whose effectiveness hinges on estimating weights that capture the complex interplay between the models. Based on our theoretical results, we empirically show that these optimal weighting functions differ significantly across LLM-PRM pairs and, notably, often assign substantial negative weights. Motivated by these insights, we propose efficient pre-computation methods to calibrate these weighting functions. Extensive experiments across 5 LLMs and 7 PRMs demonstrate that our calibration method significantly boosts the TTS efficiency, surpassing the performance of vanilla weighted majority voting while using only $\sim 21.3\\%$ of the computation. Ultimately, our work demonstrates that investing in a more intelligent aggregation strategy can be a more convincing path to performance gains than simply scaling test-time computation.

We study the neural program synthesis of $\textit{parameterized}$ function families through the lens of meta-learning with hypernetworks. Given a user intent $U$, a meta-learner $M_{\phi}$ produces a full weight set $\hat{\theta}=M_{\phi}(U)$ for a target neural network with fixed architecture $S$, and the instantiated network $m_{S,\hat{\theta}}(X)\to Y$ realizes the behavior intended for $U$. Classical hypernetworks typically $\textit{ignore the target network’s structure}$ and emit a flat list of weights; as a consequence, they fail to account for $\textit{neuron-permutation symmetry}$—many distinct parameterizations of $S$ implement the same function—so equivalent solutions are treated as different targets, fragmenting supervision and hurting out-of-distribution generalization. To address this, we propose $\textit{Meta-GNN}$, a hypernetwork that constructs a $\textit{neural graph}$ from the target architecture $S$ and applies $\textbf{structure-aware}$ message passing with parameter-tied encoders and decoders. This design reduces the search space during learning by collapsing equivalent classes of target networks, without loss of expressivity. Empirically, across modular arithmetic ($\textit{AddMod}$-$p$), array operations ($\textit{SumFirst}$-$n$), and inverse-rule tasks from 1D-ARC, $\textit{Meta-GNN}$ substantially improves learning and $\textbf{out-of-distribution generalization}$ compared to classic hypernetworks and direct $(U,X)\to Y$ baselines. Mechanistic analyses reveal $\textit{what is learned}$: on $\textit{AddMod}$-$p$ the synthesized Transformers recover the canonical clock representation and admit a compact closed-form map $U\mapsto\theta$. These results demonstrate that structure-aware Meta-GNNs enable reliable generalization to $\textit{unseen program parameterizations}$, providing a critical advance for the nascent field of neural program synthesis.

The exponential growth of video content highlights the importance of video summarization, a task that efficiently extracts key information from long videos. However, existing video summarization studies face inherent limitations in understanding complex, multimodal videos. This limitation stems from the fact that most existing architectures employ static or modality-agnostic fusion, which fails to account for the dynamic and frame-dependent variation in modality saliency that naturally occurs within a video. To overcome these limitations, we propose a novel architecture, TripleSumm, which adaptively weights and fuses the contributions of the three modalities at the frame level. Furthermore, a significant bottleneck for research into multimodal video summarization has been the lack of comprehensive benchmarks. Addressing this bottleneck, we introduce MoSu (Most Replayed Multimodal Video Summarization), the first large-scale benchmark that provides all three modalities. Our proposed TripleSumm demonstrates its superiority by achieving state-of-the-art performance by a large margin on four video summarization benchmarks, including MoSu.

Test-time entropy minimization helps adapt a model to novel environments and incentivize its reasoning capability, unleashing the model's potential during inference by allowing it to evolve and improve in real-time using its own predictions. However, pure test-time entropy minimization can favor non-generalizable shortcuts, such as inflating the logit norm and driving all predictions to a dominant class to reduce entropy, risking collapsed solutions (e.g., constant one-hot outputs) that trivially minimize the objective without meaningful learning. In this paper, we introduce ZeroSiam, an efficient asymmetric Siamese architecture tailored for test-time entropy minimization. ZeroSiam prevents collapse through asymmetry learning, which is efficiently achieved by a learnable predictor and a stop-gradient operator before the classifier. Beyond the collapse prevention effect, we provide empirical and theoretical evidence that ZeroSiam also absorbs and regularizes biases at testing, enhancing TTA effectiveness even when no collapse occurs. Despite its simplicity, extensive results show that ZeroSiam can avoid collapse and perform more stably over prior methods using negligible overhead, demonstrating efficacy on both vision and language tasks across challenging test scenarios and diverse models, e.g., tiny based models that are particularly collapse-prone.

The Homotopy paradigm, a general principle for solving challenging problems, appears across diverse domains such as robust optimization, global optimization, polynomial root-finding, and sampling. Practical solvers for these problems typically follow a predictor-corrector (PC) structure, but rely on hand-crafted heuristics for step sizes and iteration termination, which are often suboptimal and task-specific. To address this, we unify these problems under a single framework, which enables the design of a general neural solver. Building on this unified view, we propose Neural Predictor-Corrector (NPC), which replaces hand-crafted heuristics with automatically learned policies. NPC formulates policy selection as a sequential decision-making problem and leverages reinforcement learning to automatically discover efficient strategies. To further enhance generalization, we introduce an amortized training mechanism, enabling one-time offline training for a class of problems and efficient online inference on new instances. Experiments on four representative homotopy problems demonstrate that our method generalizes effectively to unseen instances. It consistently outperforms classical and specialized baselines in efficiency while demonstrating superior stability across tasks, highlighting the value of unifying homotopy methods into a single neural framework.

Tackling complex optimization problems often relies on expert-designed heuristics, typically crafted through extensive trial and error. Recent advances demonstrate that large language models (LLMs), when integrated into well-designed evolutionary search frameworks, can autonomously discover high-performing heuristics at a fraction of the traditional cost. However, existing approaches predominantly rely on verbal guidance, i.e., manipulating the prompt generation process, to steer the evolution of heuristics, without adapting the underlying LLM. We propose a hybrid framework that combines verbal and numerical guidance, the latter achieved by fine-tuning the LLM via reinforcement learning (RL) based on the quality of generated heuristics. This joint optimization allows the LLM to co-evolve with the search process. Our method outperforms state-of-the-art (SOTA) baselines across various optimization tasks, running locally on a single 24GB GPU using a 7B model with INT4 quantization. It surpasses methods that rely solely on verbal guidance, even when those use significantly more powerful API-based models.

Learning to simulate complex physical systems from data has emerged as a promising way to overcome the limitations of traditional numerical solvers, which often require prohibitive computational costs for high-fidelity solutions. Recent Graph Neural Simulators (GNSs) accelerate simulations by learning dynamics on graph-structured data, yet often struggle to capture long-range interactions and suffer from error accumulation under autoregressive rollouts. To address these challenges, we propose Information-preserving Graph Neural Simulators (IGNS), a graph-based neural simulator built on the principles of Hamiltonian dynamics. This structure guarantees preservation of information across the graph, while extending to port-Hamiltonian systems allows the model to capture a broader class of dynamics, including non-conservative effects. IGNS further incorporates a warmup phase to initialize global context, geometric encoding to handle irregular meshes, and a multi-step training objective that facilitates PDE matching, where the trajectory produced by integrating the port-Hamiltonian core aligns with the ground-truth trajectory, thereby reducing rollout error. To evaluate these properties systematically, we introduce new benchmarks that target long-range dependencies and challenging external forcing scenarios. Across all tasks, IGNS consistently outperforms state-of-the-art GNSs, achieving higher accuracy and stability under challenging and complex dynamical systems. Our project page: https://thobotics.github.io/neural_pde_matching.

While preference optimization is crucial for improving visual generative models, how to effectively scale this paradigm for visual generation remains largely unexplored. Current open-source preference datasets typically contain substantial conflicting preference patterns, where winners excel in some dimensions but underperform in others. Naively optimizing on such noisy datasets fails to learn meaningful preferences, fundamentally hindering effective scaling. To enhance the robustness of preference algorithms against noise, we propose Poly-DPO, which extends the DPO objective with an additional polynomial term that dynamically adjusts model confidence during training based on dataset characteristics, enabling effective learning across diverse data distributions from noisy to trivially simple patterns. Beyond biased patterns, existing datasets suffer from low resolution, limited prompt diversity, and imbalanced distributions. To facilitate large-scale visual preference optimization by tackling key data bottlenecks, we construct ViPO, a massive-scale preference dataset with 1M image pairs (1024px) across five categories and 300K video pairs (720p+) across three categories. Leveraging state-of-the-art generative models and diverse prompts ensures consistent, reliable preference signals with balanced distributions. Remarkably, when applying Poly-DPO to our high-quality dataset, the optimal configuration converges to standard DPO. This convergence validates both our dataset quality and Poly-DPO's adaptive nature: sophisticated optimization becomes unnecessary with sufficient data quality, yet remains valuable for imperfect datasets. We comprehensively validate our approach across various visual generation models. On noisy datasets like Pick-a-Pic V2, Poly-DPO achieves 6.87\% and 2.32\% gains over Diffusion-DPO on GenEval for SD1.5 and SDXL, respectively. For our high-quality VIPO dataset, models achieve performance far exceeding those trained on existing open-source preference datasets. These results confirm that addressing both algorithmic adaptability and data quality is essential for scaling visual preference optimization. All models and datasets will be released.

While Large Vision-Language Models (LVLMs) achieve strong performance in multimodal tasks, hallucinations continue to affect their reliability. Among the three categories of hallucinations, which include object, attribute, and relation, relation hallucinations account for the largest proportion but have received the least attention. To address this challenge, we propose \textbf{ChainMPQ} (\textbf{M}ulti-\textbf{P}erspective \textbf{Q}uestions guided Interleaved Text-image Reasoning \textbf{Chain}), a training-free method that improves relational inference in LVLMs by utilizing accumulated textual and visual memories. ChainMPQ first extracts subject and object keywords from the question to enhance the corresponding image regions. It then constructs multi-perspective questions that focus on the three core components of a relationship: the subject, the object, and the relation that links them. These questions are sequentially input to the model, with textual and visual memories from earlier steps providing supporting context for subsequent ones, thereby forming an interleaved chain of image and text that guides progressive relational reasoning. Experiments on multiple LVLMs and benchmarks show that ChainMPQ substantially reduces relation hallucinations, while ablation studies further validate the effectiveness of its three core modules.

Scaling laws for language model training traditionally characterize how performance scales with model size and dataset volume. Prior work has explored architecture variants and data treatments such as dataset filtering and noise injection in language model pretraining; however, these studies have not formalized data quality within a principled scaling law. We introduce a dimensionless data-quality parameter Q, and propose a quality-aware scaling law extending the Chinchilla framework to predict loss as a joint function of model size, data volume, and data quality. The law is motivated by an effective-sample-size and information-theoretic view of noisy or redundant corpora, and it admits two practical estimators for Q: (i) a corruption rate proxy and (ii) a deficiency measure. Through synthetic experiments in neural machine translation and autoregressive modeling--where we systematically control data quality via multiple levels of noise injection--we show that loss scales predictably with data quality and that higher-quality data can substantially reduce model size and hence compute requirements. Our results demonstrate a sublinear decay of effective data with quality and robustness to moderate data corruption; out-of-sample evaluations further validate the predictive form of the law. Unlike prior empirical analyses, our work establishes an explicit, generalizable law for data quality, offering concrete guidance for balancing data curation effort and model scale in large-scale pretraining.

RAG (Retrieval-Augmented Generation) systems and web agents are increasingly evaluated on multi-hop deep search tasks, yet current practice suffers from two major limitations. First, most benchmarks leak the reasoning path in the question text, allowing models to follow surface cues rather than discover reasoning chains autonomously. Second, evaluation is typically reduced to a single pass rate, which collapses diverse behaviors into one score and obscures whether failures stem from inadequate search, poor knowledge use, or inappropriate refusal. To address these issues, we present WebDetective, a benchmark of hint-free multi-hop questions paired with a controlled Wikipedia sandbox that ensures full traceability of model actions, and a holistic evaluation framework that separates search sufficiency, knowledge utilization, and refusal behavior. Our evaluation of 25 state-of-the-art models reveals systematic weaknesses across all architectures: models struggle with knowledge utilization despite having sufficient evidence and demonstrate near-absent appropriate refusal when evidence is lacking. These patterns expose a fundamental gap—today's systems excel at executing given reasoning paths but fail when required to discover them. We develop an agentic workflow EvidenceLoop that explicitly targets the challenges our benchmark identifies, incorporating verification loops and systematic evidence tracking that improve both search and synthesis capabilities. This baseline demonstrates that WebDetective's diagnostic framework can guide concrete architectural improvements, establishing our benchmark as a critical tool for developing genuinely autonomous reasoning systems rather than pattern-following agents.

Understanding human behavior traits is central to applications in human-computer interaction, computational social science, and personalized AI systems. Such understanding often requires integrating multiple modalities to capture nuanced patterns and relationships. However, existing resources rarely provide datasets that combine behavioral descriptors with complementary modalities such as facial attributes and biographical information. To address this gap, we present PersonaX, a curated collection of multimodal datasets designed to enable comprehensive analysis of public traits across modalities. PersonaX consists of (1) CelebPersona, featuring 9444 public figures from diverse occupations, and (2) AthlePersona, covering 4181 professional athletes across 7 major sports leagues. Each dataset includes behavioral trait assessments inferred by three high-performing large language models, alongside facial imagery and structured biographical features. We analyze PersonaX at two complementary levels. First, we abstract high-level trait scores from text descriptions and apply five statistical independence tests to examine their relationships with other modalities. Second, we introduce a novel causal representation learning (CRL) framework tailored to multimodal and multi-measurement data, providing theoretical identifiability guarantees. Experiments on both synthetic and real-world data demonstrate the effectiveness of our approach. By unifying structured and unstructured analysis, PersonaX establishes a foundation for studying LLM-inferred behavioral traits in conjunction with visual and biographical attributes, advancing multimodal trait analysis and causal reasoning. The code is available at https://github.com/lokali/PersonaX.