Recent years have witnessed remarkable progress in world models, which primarily aim to capture the spatiotemporal correlations between an agent's actions and the evolving environment. However, existing approaches often suffer from tight runtime coupling or depend on offline reward signals, resulting in substantial inference overhead or hindering end-to-end optimization. To overcome these limitations, we introduce WPT, a World-to-Policy Transfer training paradigm that enables online distillation under the guidance of an end-to-end world model. Specifically, we develop a trainable reward model that infuses world knowledge into a teacher policy by aligning candidate trajectories with the future dynamics predicted by the world model. Subsequently, we propose policy distillation and world reward distillation to transfer the teacher's reasoning ability into a lightweight student policy, enhancing planning performance while preserving real-time deployability. Extensive experiments on both open-loop and closed-loop benchmarks show that WPT achieves state-of-the-art performance with a simple policy architecture: it attains a 0.11 collision rate (open-loop) and achieves a 79.23 driving score (closed-loop), surpassing both world-model-based and imitation-learning methods in accuracy and safety. Moreover, the student sustains up to 4.9xfaster inference, while retaining most of the gains.
论文检索
输入标题、作者或关键词，从 100,903 篇学术成果中精准定位
Normalizing Flows (NFs) have been established as a principled framework for generative modeling. Standard NFs consist of a forward process and a reverse process: the forward process maps data to noise, while the reverse process generates samples by inverting it. Typical NF forward transformations are constrained by explicit invertibility, ensuring that the reverse process can serve as their exact analytic inverse. Recent developments in TARFlow and its variants have revitalized NF methods by combining Transformers and autoregressive flows, but have also exposed causal decoding as a major bottleneck. In this work, we introduce Bidirectional Normalizing Flow (BiFlow), a framework that removes the need for an exact analytic inverse. BiFlow learns a reverse model that approximates the underlying noise-to-data inverse mapping, enabling more flexible loss functions and architectures. Experiments on ImageNet demonstrate that BiFlow, compared to its causal decoding counterpart, improves generation quality while accelerating sampling by up to two orders of magnitude. BiFlow yields state-of-the-art results among NF-based methods and competitive performance among single-evaluation ("1-NFE") methods. Following recent encouraging progress on NFs, we hope our work will draw further attention to this classical paradigm.
Despite remarkable advancements in multimodal large language models (MLLMs), their fine-grained visual understanding is constrained by a primary reliance on sparse textual supervision. Existing efforts to introduce visual supervision typically do so during post-training, when visual representations have already been largely fixed, causing such signals to act mainly as auxiliary constraints rather than as a primary force for shaping perceptual features. In this paper, we aim to fundamentally reshape the model's perceptual backbone by incorporating vision supervision directly into the pre-training stage. We observe that pixel-level image patches and textual tokens naturally coexist in a shared, raw high-dimensional space characterized by an inherent input symmetry. Leveraging this insight, we propose UVU, a novel vision-language unified autoregressive framework that eschews vector quantization. It uniquely employs continuous visual encoding for lossless representation of visual inputs and proposes a large-scale iterative hierarchical clustering algorithm to construct a pixel-level visual codebook, thereby extending the vocabulary for unified supervision and enabling autoregressive generation of pixel-level image tokens alongside textual tokens. UVU effectively synergizes pixel-level visual perception with semantic-level visual understanding, internalizing visual reconstruction capabilities and unlocking the facilitative role of visual supervision in enhancing understanding in the pre-training stage. Extensive experiments across multiple tasks demonstrate that MLLMs are capable of achieving superior multimodal understanding performance under the supervised learning paradigm of UVU.
VES-RFT: Rewarding Visual Evidence Sensitivity to Mitigate Hallucinations in Large Vision-Language Models
PDF ↗Vision-Language Models (VLMs) often over-rely on linguistic priors even when images are provided, leading to object hallucinations. We revisit object-wise hallucination from the perspective of how visual evidence shapes the model's uncertainty. For each input, we measure decision uncertainty with and without the image, and define a Visual Evidence Sensitivity (VES) signal as the image-attributable change in entropy. Building on this signal, we introduce Visual Evidence Sensitivity Reinforcement Fine-Tuning (VES-RFT), a training-time reinforcement fine-tuning method that explicitly rewards reliance on correct visual evidence. We pair this continuous, annotation-free signal with a verifiable reward that enforces factual object correctness by automatically checking generated object mentions against the image, yielding a computable objective without human annotations. We optimize the dual objective using critic-free GRPO with KL regularization, requiring only parallel image and no-image passes during training while preserving single-pass inference. Across multiple VLM families and benchmarks, VES-RFT consistently suppresses object hallucinations and improves robustness under ambiguity. Specifically, on LLaVA-7B, VES-RFT reduces 12.8 and 1.8 on CHAIR_S/CHAIR_I of MS-COCO, and increases POPE accuracy by 4.92%. Extensive experiments indicate that turning uncertainty into a learnable reward, paired with verifiable correctness signals, provides a scalable mechanism for training-time hallucination mitigation and stronger visual grounding.
UARE: A Unified Vision-Language Model for Image Quality Assessment, Restoration, and Enhancement
PDF ↗Image quality assessment (IQA) and image restoration are fundamental problems in low-level vision. Although IQA and restoration are closely connected conceptually, most existing work treats them in isolation. Recent advances in unified multimodal understanding-generation models demonstrate promising results and indicate that stronger understanding can improve generative performance. This motivates a single model that unifies IQA and restoration and explicitly studies how IQA can guide restoration, a setting that remains largely underexplored yet highly valuable. In this paper, we propose UARE, to our knowledge the first Unified vision-language model for image quality Assessment, Restoration, and Enhancement. Built on pretrained unified understanding and generation models, we introduce a two-stage training framework. First, a progressive, easy-to-hard schedule expands from single-type distortions to higher-order mixed degradations, enabling UARE to handle multiple degradations. Second, we perform unified fine-tuning of quality understanding and restoration with interleaved text-image data, aligning IQA signals with restoration objectives. Through multi-task co-training, UARE leverages IQA to boost restoration and enhancement performance. Extensive experiments across IQA, restoration, and enhancement tasks demonstrate the effectiveness of UARE. The code and models will be made available.
On Token's Dilemma: Dynamic MoE with Drift-Aware Token Assignment for Continual Learning of Large Vision Language Models
PDF ↗Multimodal Continual Instruction Tuning aims to continually enhance Large Vision Language Models (LVLMs) by learning from new data without forgetting previously acquired knowledge. Mixture of Experts (MoE) architectures naturally facilitate this by incrementally adding new experts and expanding routers while keeping the existing ones frozen. However, despite expert isolation, MoE-based continual learners still suffer from forgetting due to routing-drift: old-task tokens become mistakenly attracted to newly added experts, degrading performance on prior tasks. We analyze the failure mode at the token level and reveal the token's dilemma: ambiguous and old tokens in new-task data offer minimal learning benefit yet induce forgetting when routed to new experts, due to their ambiguous routing assignment during training. Motivated by this, we propose LLaVA-DyMoE, a dynamic MoE framework that incrementally expands the MoE with drift-aware token assignment. We characterize token types via their routing score distributions and apply targeted regularization. Specifically, a token-level assignment guidance steers ambiguous and old tokens away from new experts to preserve established routing patterns and alleviate routing-drift, while complementary routing score regularizations enforce expert-group separation and promote new-expert specialization. Extensive experiments demonstrate that our LLaVA-DyMoE effectively mitigates routing-drift-induced forgetting, achieving over a 7% gain in mean final accuracy and a 12% reduction in forgetting compared to baselines. The project page is zhaoc5.github.io/DyMoE.
Achieving unified 3D perception and reasoning across tasks such as segmentation, retrieval, and relation understanding remains challenging, as existing methods are either object-centric or rely on costly training for inter-object reasoning. We present a novel framework that constructs a hierarchical language-distilled Gaussian scene and its 3D semantic scene graph without scene-specific training. A Gaussian pruning mechanism refines scene geometry, while a robust multi-view language alignment strategy aggregates noisy 2D features into accurate 3D object embeddings. On top of this hierarchy, we build an open-vocabulary 3D scene graph with Vision Language-derived annotations and Graph Neural Network-based relational reasoning. Our approach enables efficient and scalable open-vocabulary 3D reasoning by jointly modeling hierarchical semantics and inter/intra-object relationships, validated across tasks including open-vocabulary segmentation, scene graph generation, and relation-guided retrieval. Project page: https://dfki-av.github.io/ReLaGS/
Reinforcement learning (RL) has recently achieved remarkable success in eliciting visual reasoning within Multimodal Large Language Models (MLLMs). However, existing approaches typically train separate models for different tasks and treat image and video reasoning as disjoint domains. This results in limited scalability toward a multimodal reasoning generalist, which restricts practical versatility and hinders potential knowledge sharing across tasks and modalities.To this end, we propose OneThinker, an all-in-one reasoning model that unifies image and video understanding across diverse fundamental visual tasks, including question answering, captioning, spatial and temporal grounding, tracking, and segmentation. To achieve this, we construct the OneThinker-600k training corpus covering all these tasks and employ commercial models for CoT annotation, resulting in OneThinker-SFT-340k for SFT cold start.Moreover, we propose EMA-GRPO to handle reward heterogeneity in multi-task RL by tracking task-wise moving averages of reward standard deviations for balanced optimization. Extensive experiments on diverse visual benchmarks show that OneThinker delivers strong performance on 31 benchmarks, across 10 fundamental visual understanding tasks.Moreover, it exhibits effective knowledge transfer between certain tasks and preliminary zero-shot generalization ability, marking a step toward a unified multimodal reasoning generalist. All code, model, and data are released.
InfiniDepth: Arbitrary-Resolution and Fine-Grained Depth Estimation with Neural Implicit Fields
PDF ↗Existing depth estimation methods are fundamentally limited to predicting depth on discrete image grids. Such representations restrict their scalability to arbitrary output resolutions and hinder the geometric detail recovery. This paper introduces InfiniDepth, which represents depth as neural implicit fields. Through a simple yet effective local implicit decoder, we can query depth at continuous 2D coordinates, enabling arbitrary-resolution and fine-grained depth estimation. To better assess our method's capabilities, we curate a high-quality 4K synthetic benchmark from five different games, spanning diverse scenes with rich geometric and appearance details. Extensive experiments demonstrate that InfiniDepth achieves state-of-the-art performance on both synthetic and real-world benchmarks across relative and metric depth estimation tasks, particularly excelling in fine-detail regions. It also benefits the task of novel view synthesis under large viewpoint shifts, producing high-quality results with fewer holes and artifacts. Code and data will be made publicly available.
Large multimodal models (LMMs) are increasingly adopted as judges in multimodal evaluation systems due to their strong instruction following and consistency with human preferences. However, their ability to follow diverse, fine-grained evaluation criteria remains underexplored. We develop Multi-Crit, a benchmark for evaluating multimodal judges on their capacity to follow pluralistic criteria and produce reliable criterion-level judgments. Covering both open-ended generation and verifiable reasoning tasks, Multi-Crit is built through a rigorous data curation pipeline that gathers challenging response pairs with multi-criterion human annotations. It further introduces three novel metrics for systematically assessing pluralistic adherence, criterion-switching flexibility, and the ability to recognize criterion-level preference conflicts. Comprehensive analysis of 25 LMMs reveals that 1) proprietary models still struggle to maintain consistent adherence to pluralistic criteria--especially in open-ended evaluation; 2) open-source models lag further behind in flexibly following diverse criteria; and 3) critic fine-tuning with holistic judgment signals enhances visual grounding but fails to generalize to pluralistic criterion-level judgment. Additional analyses on reasoning fine-tuning, test-time scaling, and boundary consistency between open-source and proprietary models further probe the limits of current multimodal judges. As a pioneering study, Multi-Crit lays the foundation for building reliable and steerable multimodal AI evaluation.
We introduce Mistake Attribution (MATT), a new task for fine-grained understanding of human mistakes in egocentric videos. While prior work detects whether a mistake occurs, MATT attributes the mistake to what part of the instruction is violated (semantic role), when in the video the deviation becomes irreversible (the Point-of-No-Return, PNR), and where the mistake appears in the PNR frame. We develop MisEngine, a data engine that automatically constructs mistake samples from existing datasets with attribution-rich annotations. Applied to large egocentric corpora, MisEngine yields EPIC-KITCHENS-M and Ego4D-M---two datasets up to two orders of magnitude larger than prior mistake datasets.We then present MisFormer, a unified attention-based model for mistake attribution across semantic, temporal, and spatial dimensions, trained with MisEngine supervision. A human study demonstrates the ecological validity of our MisEngine-constructed mistake samples, confirming that EPIC-KITCHENS-M and Ego4D-M can serve as reliable benchmarks for mistake understanding. Experiments on both our datasets and prior benchmarks show that MisFormer, as a single unified model, outperforms task-specific SOTA methods by at least 6.66%, 21.81%, 18.7%, and 3.00% in video-language understanding, temporal localization, hand-object interaction, and mistake detection, respectively. Project page: https://yayuanli.github.io/MATT/
Multimodal Large Language Models (MLLMs) often generate hallucinations, where the output deviates from the visual content. Given that these hallucinations can take diverse forms, detecting hallucinations at a fine-grained level is essential for comprehensive evaluation and analysis. To this end, we propose a novel task of multimodal fine-grained hallucination detection and editing for MLLMs. Moreover, we propose ZINA, a novel method that identifies hallucinated spans at a fine-grained level, classifies their error types into six categories, and suggests appropriate refinements. To train and evaluate models for this task, we constructed VisionHall, a dataset comprising 6.9k outputs from twelve MLLMs manually annotated by 211 annotators, and 20k synthetic samples generated using a graph-based method that captures dependencies among error types. We demonstrated that ZINA outperformed existing methods, including GPT-4o and Llama-3.2, in both detection and editing tasks.
Scenes in the real world are often composed of several static and dynamic objects. Capturing their 4-dimensional structures, composition and spatio-temporal configuration in-the-wild, though extremely interesting, is equally hard.Therefore, existing works often focus on one object at a time, while relying on some category-specific parametric shape model for dynamic objects. This can lead to inconsistent scene configurations, in addition to being limited to the modeled object categories. We propose COM4D (Compositional 4D), a method that consistently and jointly predicts the structure and spatio-temporal configuration of 4D/3D objects using only static multi-object or dynamic single object supervision. We achieve this by a carefully designed training of spatial and temporal attentions on 2D video input. The training is disentangled into learning from object compositions on the one hand, and single object dynamics throughout the video on the other, thus completely avoiding reliance on 4D compositional training data.At inference time, our proposed attention mixing mechanism combines these independently learned attentions, without requiring any 4D composition examples.By alternating between spatial and temporal reasoning, COM4D reconstructs complete and persistent 4D scenes with multiple interacting objects directly from monocular videos.COM4D provides state-of-the-art results in existing separate problems of 4D object and composed 3D reconstruction despite being purely data-driven.
The primary obstacle for applying reinforcement learning (RL) to real-world robotics is the design of effective reward functions. While recently learning-based Process Reward Models (PRMs) are a promising direction, they are often hindered by two fundamental limitations: their reward models lack step-aware understanding and rely on single-view perception, leading to unreliable assessments of fine-grained manipulation progress; and their reward shaping procedures are theoretically unsound, often inducing a semantic trap that misguides policy optimization. To address these, we introduce Dopamine-Reward, a novel reward modeling method for learning a general-purpose, step-aware process reward model from multi-view inputs. At its core is our General Reward Model (GRM), trained on a vast 3,400+ hour dataset, which leverages Step-wise Reward Discretization for structural understanding and Multi-Perspective Reward Fusion to overcome perceptual limitations. Building upon Dopamine-Reward, we propose Dopamine-RL, a robust policy learning framework that employs a theoretically-sound Policy-Invariant Reward Shaping method, which enables the agent to leverage dense rewards for efficient self-improvement without altering the optimal policy, thereby fundamentally avoiding the semantic trap. Extensive experiments across diverse simulated and real-world tasks validate our approach. GRM achieves state-of-the-art accuracy in reward assessment, and Dopamine-RL built on GRM significantly improves policy learning efficiency. For instance, after GRM is adapted to a new task in a one-shot manner from a single expert trajectory, the resulting reward model enables Dopamine-RL to improve the policy from near-zero to 95% success with only 150 online rollouts (approximately 1 hour of real robot interaction), while retaining strong generalization across tasks.
We propose Coded-E2LF (coded event to light field), a computational imaging method for acquiring a 4-D light field using a coded aperture and a stationary event-only camera. In a previous work, an imaging system similar to ours was adopted, but both events and intensity images were captured and used for light field reconstruction. In contrast, our method is purely event-based, which relaxes restrictions for hardware implementation. We also introduce several advancements from the previous work that enable us to theoretically support and practically improve light field reconstruction from events alone. In particular, we clarify the key role of a black pattern in aperture coding patterns. We finally implemented our method on real imaging hardware to demonstrate its effectiveness in capturing real 3-D scenes. To the best of our knowledge, we are the first to demonstrate that a 4-D light field with pixel-level accuracy can be reconstructed from events alone. Our software and supplementary video are available from our project website: https://www.fujii.nuee.nagoya-u.ac.jp/Research/EventLF/.
Image memorability, i.e., how likely an image is to be remembered, has traditionally been studied in computer vision either as a passive prediction task, with models regressing a scalar score, or with generative methods altering the visual input to boost the image likelihood of being remembered. Yet, none of these paradigms supports users at capture time, when the crucial question is how to improve a photo memorability. We introduce the task of **Mem**orability **Feed**back (**MemFeed**), where an automated model should provide actionable, human-interpretable guidance to users with the goal to enhance an image future recall. We also present **MemCoach**, the first approach designed to provide concrete suggestions in natural language for memorability improvement (e.g., "emphasize facial expression," "bring the subject forward"). Our method, based on Multimodal Large Language Models (MLLMs), is training-free and employs a teacher-student steering strategy, aligning the model internal activations toward more memorable patterns learned from a teacher model progressing along least-to-most memorable samples. To enable systematic evaluation on this novel task, we further introduce **MemBench**, a new benchmark featuring sequence-aligned photoshoots with annotated memorability scores. Our experiments, considering multiple MLLMs, demonstrate the effectiveness of MemCoach, showing consistently improved performance over several zero-shot models. The results indicate that memorability can not only be predicted but also taught and instructed, shifting the focus from mere prediction to actionable feedback for human creators. Dataset and code will be publicly released upon publication.
Raw images taken in low-light conditions are very noisy due to low photon count and sensor noise. Learning-based denoisers have the potential to reconstruct high-quality images. For training, however, these denoisers require large paired datasets of clean and noisy images, which are difficult to collect. Noise synthesis is an alternative to large-scale data acquisition: given a clean image, we can synthesize a realistic noisy counterpart. In this work, we propose a general and practical noise synthesis method that requires only one single noisy image and one single dark frame per ISO setting. We represent signal-dependent noise with a Poisson distribution and introduce a Fourier-domain spectral sampling algorithm to accurately model signal-independent noise. The latter generates diverse noise realizations that maintain the spatial and statistical properties of real sensor noise. As opposed to concurrent approaches, our method neither relies on simplified parametric models nor on large sets of clean-noisy image pairs. It is accurate and practical. Moreover, our synthesis method leads to state-of-the-art performances on multiple low-light denoising benchmarks.
MimiCAT: Mimic with Correspondence-Aware Cascade-Transformer for Category-Free 3D Pose Transfer
PDF ↗3D pose transfer aims to transfer the pose-style of a source mesh to a target character while preserving both the target's geometry and the source's pose characteristic. Existing methods are largely restricted to characters with similar structures and fail to generalize to category-free settings (e.g., transferring a humanoid's pose to a quadruped). The key challenge lies in the structural and transformation diversity inherent in distinct character types, which often leads to mismatched regions and poor transfer quality. To address these issues, we first construct a million-scale pose dataset across hundreds of distinct characters. We further propose MimiCAT, a cascade-transformer model designed for category-free 3D pose transfer. Instead of relying on strict one-to-one correspondence mappings, MimiCAT leverages semantic keypoint labels to learn a novel soft correspondence that enables flexible many-to-many matching across characters. The pose transfer is then formulated as a conditional generation process, in which the source transformations are first projected onto the target through soft correspondence matching and subsequently refined using shape-conditioned representations. Extensive qualitative and quantitative experiments demonstrate that MimiCAT generalizes plausible poses across diverse character morphologies, surpassing prior approaches restricted to narrow-category transfer (e.g., humanoid-to-humanoid).
Visual Speaker Authentication (VSA) verifies identity by analyzing lip dynamics during prompted speech, offering enhanced privacy compared to full-face methods while maintaining discriminability for high-security applications. However, recent advances in talking face generation (TFG) have enabled realistic forgeries that closely mimic lip dynamics in sync with speech, posing severe threats to VSA systems. Prevailing defenses rely heavily on supervised classifiers trained on known forgeries via empirical risk minimization, resulting in poor generalization to unseen attacks, dependency on continuously updated fake data, and complete failure in the absence of effective forgery priors. In this paper, we revisit the design of forgery detectors and argue that over-reliance on fake priors hinders the exploitation of rich authenticity signals inherently present in real videos. We propose a novel detector trained exclusively on authentic data, learning forgery-aware representations through three key components: (1) lightweight modules that capture forgery-indicative statistics from real videos; (2) an asymmetric contrastive objective that compacts real samples while repelling potential forgeries in representation space; and (3) a theoretically grounded regularizer that shapes real representations into a tractable, isotropic Gaussian. To support rigorous evaluation, we introduce a benchmark suite spanning diverse TFG forgeries. Across eight modern forgery attacks and ten state-of-the-art (SOTA) detectors, our method achieves over a 10% reduction in error rates while preserving identity-verification capability with minimal overhead, and demonstrates consistent gains on datasets that better emulate real-world scenarios.
Vision-Language-Action (VLA) models are promising for embodied intelligence, yet they often overlook the predictive and temporal-causal structure underlying visual dynamics. World-model VLAs address this by predicting future frames, but waste capacity reconstructing redundant backgrounds. To overcome these limitations, we introduce CoWVLA, a new "Chain of World" paradigm that unifies world-model temporal reasoning with a disentangled latent motion representation. First, a pretrained video VAE serves as a latent motion extractor, explicitly factorizing video segments into structure and motion latents. Then, during pre-training, the VLA learns from an instruction and an initial frame to infer a continuous latent motion chain and predict the segment's terminal frame. Finally, during co-fine-tuning, this latent dynamic is aligned with discrete action prediction by jointly modeling sparse keyframes and action sequences in a unified autoregressive decoder. This design preserves the world-model benefits of temporal reasoning and world knowledge while retaining the compactness and interpretability of latent actions, enabling efficient visuomotor learning. Extensive experiments on robotic simulation benchmarks show that CoWVLA outperforms existing world-model and latent-action approaches and achieves moderate computational efficiency, highlighting its potential as a more effective VLA pretraining paradigm. Project website: https://fx-hit.github.io/cowvla-io.