arXiv subjects

Zuxuan Wu

Publications and source records attributed to Zuxuan Wu.

At least 19 recordsLinked to original sources

arXiv2026

Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Model

Recent Omni-Modal Generative Models (Omni-Models) have advanced content generation toward unified modeling of text, images, video, and audio. MiniMax-H3 exemplifies this transition by combining multimodal context understanding with joint audio-visual generation in a shared latent framework. Its unified architecture raises a fundamental question: Can multimodal alignment improve the model's world reasoning, and what new evaluation paradigms do omni-modal inputs enable? To investigate this question, this work introduces a comprehensive evaluation framework organized around four complementary dimensions of physical world reasoning. Unlike existing evaluation frameworks for video generation and world models, which are often constrained by limited input modalities and evaluation settings where prompts closely match the target video content, our evaluation is specifically designed to exploit the multimodal inputs of Omni-Model. We construct a diverse set of novel tasks that require models to integrate complementary information across modalities. Specifically, we consider four scenarios, including implicit prompts paired with multiple frames, audio-image, prefix-videos, and audio-video inputs. Every single modality provides only partial evidence about the underlying event, requiring the model to jointly reason over the complementary semantic cues to infer latent event states and future dynamics. Across 517 evaluation instances, MiniMax-H3 achieves an overall success rate of 41.97%. Video-based Decision Reasoning yields the highest success rate at 56.00%, while Audio-based Disambiguation Reasoning is the weakest, reaching only 27.40%. These results indicate that effective multimodal integration remains key to fully exploiting the benefits of diverse input modalities. The project is available at https://github.com/gulucaptain/MiniMax-H3-Reason.

cs.CV↗

arXiv2026

Bench2Dex: Benchmarking Visuo-Tactile Bimanual Dexterous Manipulation Across Dexterous Hands

Tactile sensing provides contact information that can be difficult to infer from vision alone, but tactile hardware for dexterous hands has not converged to a common design. Dexterous hands differ in finger structure, contact surfaces, and sensor layouts, while simulated tactile signals still differ from measurements produced by physical sensors. These factors make it difficult to study visuo-tactile manipulation across diverse dexterous hands within a consistent experimental setting. We present Bench2Dex, a simulation benchmark for visuo-tactile bimanual manipulation across 12 dexterous hands. We adapt existing robot models with a shared simulated tactile interface that converts local contact geometry into image-like tactile observations. The interface provides a consistent observation format across different hand morphologies without attempting to reproduce the output of a specific physical tactile sensor. Bench2Dex includes 26 bimanual manipulation tasks that involve tool use, articulated-object interaction, and multi-stage manipulation, together with about 1.3K human-teleoperated demonstrations. The benchmark provides synchronized visual, tactile, proprioceptive, action, and object-state observations, together with executable task metrics. For robustness, we group seven perturbation types into invariance axis, where the correct action does not change, and equivariance axis, where the correct action changes together with the perturbation. We evaluate ACT, Diffusion Policy, pi0.5, and GR00T N1.5 on Bench2Dex and report their performance and failure modes. Bench2Dex is meant as a platform for studying visuo-tactile learning across dexterous hands. It does not assume that simulated tactile observations can replace real tactile sensing; it offers a shared setting for algorithm development while tactile hardware and simulation models are still evolving.

cs.RO↗

arXiv2026

DiffusionOPD: A Unified Perspective of On-Policy Distillation in Diffusion Models

Reinforcement learning has emerged as a powerful tool for improving diffusion-based text-to-image models, but existing methods are largely limited to single-task optimization. Extending RL to multiple tasks is challenging: joint optimization suffers from cross-task interference and imbalance, while cascade RL is cumbersome and prone to catastrophic forgetting. We propose DiffusionOPD, a new multi-task training paradigm for diffusion models based on Online Policy Distillation (OPD). DiffusionOPD first trains task-specific teachers independently, then distills their capabilities into a unified student along the student own rollout trajectories. This decouples single-task exploration from multi-task integration and avoids the optimization burden of solving all tasks jointly from scratch. Theoretically, we lift the OPD framework from discrete tokens to continuous-state Markov processes, deriving a closed-form per-step KL objective that unifies both stochastic SDE and deterministic ODE refinement via mean-matching. We formally and empirically demonstrate that this analytic gradient provides lower variance and better generality compared to conventional PPO-style policy gradients. Extensive experiments show that DiffusionOPD consistently surpasses both multi-reward RL and cascade RL baselines in training efficiency and final performance, while achieving state-of-the-art results on all evaluated benchmarks.

cs.LG↗

arXiv2026

VA-Judger: Reward Modeling from Human Preference Feedback for Joint Video-Audio Generation

Using reinforcement learning to post-train joint video-audio generation models requires a reward signal. Existing methods construct this reward by combining metrics for individual quality dimensions, including audio quality, visual fidelity, and synchronization. However, these metrics evaluate perceptual dimensions separately and fail to capture the overall semantic and temporal coherence among the text prompt, video, and audio that shapes human preferences. Optimizing models against these metrics encourages reward hacking, generating video-audio content that achieves high scores on these metrics yet appears incoherent or unfaithful to human viewers. To address this problem, we first construct a large-scale human-preference dataset VAPref-10K for joint video-audio generation, comprising 9K prompts and 10.3K fine-grained paired comparisons from open-source generation models. We also introduce the VA-Judger-Bench benchmark with both in-domain and out-of-domain model comparisons to evaluate whether reward models truly align with human preferences. We further propose VA-Judger, a chain-of-thought omni-reward model for joint video-audio generation. In particular, VA-Judger first learns from pairs with clear quality gaps to establish structured output and coarse preference discrimination, then distills reliable preference explanations for harder near-quality comparisons via rejection sampling verified against human annotations, and finally performs dimension-wise reinforcement learning that decomposes human feedback into individual quality dimensions for denser reward signals than a single binary preference label. Experiments show that VA-Judger outperforms metric baselines in predicting human preferences on both in-domain and out-of-domain evaluations. Using its human-aligned rewards for post-training audio-video generation model also yields significant improvements in generation quality.

cs.CV↗

arXiv2026

Large Discrete Policy: Advancing Explicit Behavior Modeling with Stochastic Iterative Scoring

Behavior policies are often formulated as continuous generative models, whose iterative denoising processes are expressive but difficult to interpret and prone to producing implausible actions. We propose the Large Discrete Policy (LDiP), a fully discrete behavior modeling framework that selects actions from a large vocabulary of physically plausible candidates. Rather than perturbing actions, LDiP improves expressivity through stochastic iterative scoring: it progressively re-scores and prunes candidates with score-space stochasticity, enabling fine-grained ranking and exploration among plausible actions while preserving an explicit decision process. Across end-to-end planning, closed-loop driving, robotic manipulation, and vision-language-action settings, LDiP consistently outperforms strong discrete and continuous baselines in autonomous driving, and exceeds or matches continuous generative policies in robotic manipulation. These results show that discrete policies, when equipped with effective scoring mechanisms, offer an expressive, plausible, and interpretable alternative for behavior modeling. Project website: https://zhenxinli.net/LargeDiscretePolicy/.

cs.RO↗

arXiv2026

Agentic Visual Generation: From Generative Models to Agentic Control

Visual generation is evolving from generative models used through a single invocation into agentic control processes that can plan, select tools, inspect intermediate synthesized outputs, revise failures, and reuse prior experience. In most existing systems, the controller is an LLM or VLM, while visual generation models serve as tools or executors. However, existing work lacks a consistent criterion for determining when a generation system becomes agentic. Planning depth, tool use, multi-role collaboration, and reinforcement learning are often treated as evidence of agenticity, even though none of them necessarily determines which generation decisions the controller can make. We organize the field according to what the controller can directly control in the generation process. At L1 Conditioning Control, the controller prepares the input to a predetermined generator but does not control which visual operation is executed. At L2 Execution Control, it selects and invokes actual generation, editing, rendering, or other content-modifying operations. At L3 Outcome-Adaptive Control, it observes an intermediate outcome and uses that observation to change a subsequent operation within the current task. At L4 Experience-Adaptive Control, it retains experience from completed tasks and uses that experience to change decisions on future tasks. L0 Fixed Support separately denotes generators, editors, evaluators, reward models, benchmarks, and fixed pipelines without a deployed controller that makes generation-level decisions. These levels describe a progressively broader decision-making scope rather than model size, system complexity, output quality, tool or role count, or training method. Applying this framework across image, video, editing, 3D, world, slide, and user-interface generation reveals how controller capabilities have evolved and how their mechanisms are distributed across levels.

cs.CV↗

arXiv2026

HandEdit: A Unified Benchmark for Egocentric Human-to-Robot Dexterous Hand Image Editing

Robotic manipulation with dexterous hands is a cornerstone of Embodied AI, yet its progress is stifled by the high cost of collecting embodiment-aware teleoperation data. While abundant egocentric videos of human hands offer a scalable alternative, the profound discrepancies in appearance, articulation, and camera viewpoints between human and robotic data raise significant challenges for co-training. Though existing general image-editing models demonstrate strong capabilities, they lack necessary embodiment-specific priors to fully bridge this gap. In this work, we present HandEdit, a unified large-scale embodiment-aware image-editing dataset and benchmark specifically designed to transform human hands and arms into various dexterous robotic embodiments within egocentric frames. HandEdit comprises over 200M editing instances derived from five diverse source datasets, covering 26 distinct URDFs, including 13 hand-only and 13 hand-arm configurations. Alongside the dataset, we establish a unified benchmark protocol with two tracks: Hand-only and Hand-Arm, supporting URDF-conditioned evaluation. We conduct extensive evaluations of 11 representative image-editing baselines using a multi-dimensional metric suite, including generic similarity metrics, VLM-based judgment, and embodiment-aware metrics. HandEdit serves as a critical resource at the intersection of image editing and robotics: it advances embodiment-aware editing models while enabling scalable dexterous robotic learning from abundant human video data, paving the way for more generalizable Embodied AI.

cs.RO↗

arXiv2026

VLZip: Unified Visual and Textual Compression for Interleaved Long-Context Modeling

Vision Language Models (VLMs) face significant challenges with ultra-long, interleaved image-text sequences due to the quadratic complexity of self-attention. Current solutions either resort to aggressive token pruning, risking irreversible information loss, or adopt efficient but less precise architectures, while largely ignoring the equally vital textual component. We introduce VLZip, a framework that unifies visual and textual compression for high-fidelity reasoning within a pure Transformer. At its core, VLZip hierarchically distills visual and textual segments into compact, layer-specific "soft prefixes" and injects them into each decoder layer's hidden states, drastically shortening the attention sequence while preserving fine-grained global context. To address deficient evaluations in the field, we also introduce LongVLBench, a new benchmark derived from video narratives that demands holistic, narrative-level reasoning. Extensive experiments show VLZip achieves leading performance on long-context multimodal reasoning, enabling training up to 120K tokens, a 6x increase over the baseline, and inference beyond 280K tokens with significantly reduced memory, while demonstrating the memory scalability to handle up to 2M tokens. By excelling at extreme context lengths where existing methods collapse, VLZip establishes an efficient and powerful new standard for long-context multimodal AI. Code is available at https://github.com/ShareLab-SII/VLZip.

cs.CV↗

arXiv2026

CopyCat: Improving Fine-Grained Subject Consistency in Subject-to-Image Models within Seconds

Recent subject-to-image models have achieved impressive progress in personalized image generation, yet they still struggle to preserve fine-grained subject-specific details. A major reason is the lack of high-quality fine-grained identity supervision: real paired data are expensive to collect, while synthesized training pairs often preserve only coarse subject appearance and fail to capture subtle subject-specific details. In this work, we propose CopyCat, a lightweight model-refinement framework that improves fine-grained subject consistency within only a few seconds. CopyCat performs a one-time refinement of a pretrained subject-to-image model by attaching a lightweight Fine-grained Consistency LoRA (FCLoRA) and optimizing it using a single proxy image, which is used as both the conditioning image and the reconstruction target. This exact self-reconstruction objective substantially simplifies the optimization task, enabling effective fine-grained refinement within only a few seconds. The refinement is performed only once; the resulting model can be directly applied to diverse unseen reference subjects and prompts without further subject-specific optimization. We further revisit subject-to-image LoRA training in double-stream diffusion transformers and find that adapting only the visual stream consistently improves subject consistency. Extensive experiments on DreamBench and XVerseBench demonstrate consistent improvements in fine-grained subject consistency across representative subject-to-image models under both single- and multi-subject settings.

cs.CV↗

arXiv2026

EgoRecovery: Acquiring Failure Recovery Ability Through Human Recovery Demonstration

Robust embodied robots should be able to recover from failures and retry tasks in order to operate reliably in unstructured and noisy real-world environments. Achieving this capability requires training policies on data that captures recovery behaviors. However, collecting such data through robot teleoperation is difficult to scale, as it is time-consuming to induce diverse failure states, perform corrective actions, and reset the environment. This challenge is further exacerbated by the high diversity of failure modes, which demands substantially more recovery data than success demonstrations. In this work, we show that egocentric human data capturing failure recovery processes provides a scalable alternative. By efficiently arranging task-level failure configurations and recording short recovery segments, human operators can generate more than 10x as much valid recovery data per hour compared to robot teleoperation under our protocol. To address the embodiment gap between human and robot, we propose EgoRecovery, a co-training framework for learning recovery behavior, where human recovery demonstrations are aligned to a compact corrective-intent space shared with robot data, which captures the timing and magnitude of correction. Only a small number of robot recovery demonstrations are required to connect this intent to executable robot actions. At deployment, a learned recovery gate predicts when correction is needed from robot observations and activates the corrective intent only in recovery states. Experiments on real-world recovery tasks show that EgoRecovery improves success from failure starts over robot-only recovery, direct co-training with human recovery data, and direct intent-transfer baselines.

cs.RO↗

arXiv2026

Enhancing Vision Foundation Models via Multimodal Continual Pre-Training

Vision Foundation Models (VFMs) provide strong visual representations for a wide range of applications. In this work, we enhance prevailing VFMs through multimodal training, allowing them to effectively process visual inputs at varying resolutions while producing visual representations that are better aligned with language representations, regardless of their original pre-training objectives. To this end, we introduce M-CPT, a Multimodal Continual Pre-Training framework designed to improve the understanding capability of pre-trained VFMs while preserving their strong visual representation quality. M-CPT introduces a Continual Position Embedding (CPE) for handling flexible visual resolutions, along with a feature alignment objective that improves the consistency between visual and textual representations during multimodal training. Extensive experiments on leading VFMs, including DINOv2, SigLIP, and AIMv2, demonstrate that M-CPT consistently improves multimodal understanding performance while preserving strong performance on standard vision benchmarks such as classification and segmentation.

cs.CV↗

arXiv2026

WeEdit: A Dataset, Benchmark and Glyph-Guided Framework for Text-centric Image Editing

Instruction-based image editing aims to modify specific content within existing images according to user-provided instructions while preserving non-target regions. Beyond traditional object- and style-centric manipulation, text-centric image editing focuses on modifying, translating, or rearranging textual elements embedded within images. However, existing leading models often struggle to execute complex text editing precisely, frequently producing blurry or hallucinated characters. We attribute these failures primarily to the lack of specialized training paradigms tailored for text-centric editing, as well as the absence of large-scale datasets and standardized benchmarks necessary for a closed-loop training and evaluation system. To address these limitations, we present WeEdit, a systematic solution encompassing a scalable data construction pipeline, two benchmarks, and a tailored two-stage training strategy. Specifically, we propose a novel HTML-based automatic editing pipeline, which generates 330K training pairs covering diverse editing operations and 15 languages, accompanied by standardized bilingual and multilingual benchmarks for comprehensive evaluation. On the algorithmic side, we employ glyph-guided supervised fine-tuning to inject explicit spatial and content priors, followed by a multi-objective reinforcement learning stage to align generation with instruction adherence, text clarity, and background preservation. Extensive experiments demonstrate that WeEdit outperforms previous open-source models by a clear margin across diverse editing operations.

cs.CV↗

arXiv2026

SPEED: One-Step Pixel Diffusion for High-quality Video Frame Interpolation

Despite the success of diffusion models in Video Frame Interpolation (VFI), existing methods still suffer from two critical limitations. First, latent diffusion inevitably loses fine-grained details when reconstructing images from latent representations back to the pixel space. Second, multi-step sampling incurs prohibitive memory consumption and inference latency. To address these issues, we propose SPEED, a one-step pixel diffusion framework for high-quality VFI. Specifically, SPEED employs a progressive multi-stage architecture with dynamic patch scaling to effectively learn multi-scale motion, structural, and appearance representations. Furthermore, we propose a novel Noise-Update-Only Attention mechanism to prevent semantic degradation of the clean condition frames while reducing the computational overhead by nearly 50%. Besides, we introduce a Drift-aware Timestep Sampling strategy coupled with a tailored training objective to directly predict images in the pixel space, enabling one-step inference without compromising the quality of the generated frames. Extensive experiments show that SPEED achieves state-of-the-art performance. On SNU-FILM, SPEED reduces LPIPS by 8.8% while delivering 63.3% faster inference and 10.6% lower memory usage. On challenging 4K benchmarks, it further surpasses prior methods by up to 51.5% in LPIPS.

cs.MM↗

arXiv2026

SegDiff: Segmented Trajectory Diffusion for Consistent and Adaptive Robot Manipulation

Imitation learning enables robots to acquire manipulation skills from demonstrations by mapping observations to actions. Existing approaches predict either short-horizon continuous action sequences or discrete keyposes. However, continuous prediction methods suffer from compounding errors due to short prediction horizons and struggle with multi-modal action distributions, whereas keypose-based methods necessitate an external planner, constraining real-time applicability. To address these challenges, we introduce SegDiff, a closed-loop visuomotor policy that integrates the strengths of both paradigms. SegDiff decomposes demonstrations into motion segments between keyposes and learns to predict the continuous trajectory from the current state to the next keypose, enabling long-horizon prediction with real-time refinement. Furthermore, we leverage the capability of diffusion models and DDIM inversion to propose a Dynamic Temporal Ensembling mechanism, which allows the policy to efficiently respond to dynamic environments and mitigate discontinuities caused by inconsistent multi-modal sampling. SegDiff demonstrates significant performance gains over existing approaches across various simulated and real-world scenarios, indicating its strong ability to reason over extended temporal dependencies while maintaining real-time adaptability and control stability.

cs.RO↗

arXiv2026

Zero-Human Demonstration End-to-end Autonomous Driving with Trajectory Scorer

Human demonstrations are widely considered the cornerstone of end-to-end (E2E) autonomous driving despite human demonstration's scarcity for long-tail and safety-critical scenarios. Nonetheless, current E2E autonomous driving (AD) training paradigms continue to rely on human demonstrations. Imitation learning (IL) requires human demonstrations for training, whereas reinforcement learning (RL) has emerged as a promising alternative to reduce this dependency. However, most existing RL methods for E2E AD still rely implicitly on human demonstrations. A pure rewards-based RL method can overcome the need for human demonstrations, but general RL policy gradient methods suffer from the cold-start problem. In this paper, we propose ZTRS (Zero-human demonstration end-to-end autonomous driving with TRajectory Scorer) - a complete RL-based E2E planning paradigm trained solely on real-world images and rule-based rewards, entirely without human demonstration. Through our proposed Exhaustive Policy Optimization (EPO), a policy gradient variant tailored for enumerable trajectory actions and dense supervision, ZTRS enables the model to generalize better to long-tail driving scenarios. We demonstrate this generalization through our SOTA performance against IL approaches on both long-tail Navhard and closed-loop HUGSIM datasets. Project page: https://zhenxinli.net/ZTRS/.

cs.RO↗

arXiv2026

HiMoE-VLA: Hierarchical Mixture-of-Experts for Generalist Vision-Language-Action Policies

Generalist vision--language--action (VLA) policies are typically trained on heterogeneous mixtures of robot demonstrations spanning diverse embodiments, action spaces, and observation configurations. Modeling such heterogeneity with a shared dense action module can induce negative transfer, particularly when action spaces or visual observations differ across data sources. We address this issue with HiMoE-VLA, a VLA framework built around a Hierarchical Mixture-of-Experts (HiMoE) action module. HiMoE uses Action-Space MoE layers at the input/output boundaries to specialize computation for distinct action spaces, Heterogeneity-Balancing MoE layers in neighboring layers to provide balanced capacity for residual variation in observations, scenes, and embodiments, and dense Transformer blocks in the middle to integrate shared representations. Two auxiliary objectives further guide this hierarchy: a contrastive Action-Space Regularization objective for boundary specialization and a load-balancing objective for stable expert utilization. HiMoE-VLA reaches 3.98 on CALVIN, 98.0\% on LIBERO, and 75.0\% and 63.7\% average success on real xArm7 and ALOHA tasks; under controlled heterogeneous co-training, it turns the negative transfer observed in strong baselines into positive transfer. The code and models are publicly available at https://github.com/ZhiyingDu/HiMoE-VLA.

cs.RO↗

arXiv2026

Ask-to-Clarify: Resolving Instruction Ambiguity through Multi-turn Dialogue

Embodied agents are intelligent systems designed to perceive, reason, and act within the physical world. While the robotics community has long strived to build such versatile agents, a fundamental limitation persists: most current VLA-based models operate under a rigid ``Listen-and-Act'' paradigm. These systems assume instructions are unambiguous and execute them in a passive fashion, preventing them from resolving uncertainty through dialogue. To address this, we propose Ask-to-Clarify, a unified end-to-end framework that seamlessly integrates multi-turn disambiguation dialogue with low-level visuomotor control, eliminating the reliance on high-level action primitives or external planners. Specifically, Ask-to-Clarify synergizes a VLM-based Cognitive Planner with a Diffusion-based Motor Executor. To bridge the disparity between high-level disambiguation and low-level execution, we introduce a Semantic-Visual Alignment Adapter, which functions as a cross-modal interface to synthesize semantic intent with visual perceptual streams. Furthermore, we observe severe catastrophic forgetting: visuomotor fine-tuning completely erases dialogue capabilities. To overcome this, we propose a two-stage knowledge-insulation training strategy, effectively decoupling dialogue logic from physical manipulation. Extensive evaluations across 11 real-world tasks demonstrate that \framework{} significantly outperforms existing methods, offering a promising path toward building truly collaborative embodied agents.

cs.RO↗

arXiv2026

Seeing Touch from Motion: A Unified Modality-Aware Visuo-Tactile Policy with Tactile Motion Correlation

Visuo-Tactile policies leveraging optical tactile sensors have shown great promise in contact-rich manipulation. These sensors achieve high spatial resolution and multi-dimensional force sensing by utilizing an internal camera to monitor the deformation of their elastic gel surface, thereby indirectly inferring tactile cues. Despite their advantages, extracting fine-grained contact states necessary for contact-rich manipulation remains an open challenge. Existing methods typically use either raw images or cumulative motion fields to represent tactile cues. However, both are prone to perception ambiguity. Raw tactile images mainly capture appearance changes, while cumulative motion fields only reflect the aggregate gel deformation. Consequently, distinct fine-grained contact states can exhibit highly similar patterns, making it difficult to explicitly distinguish subtle contact variations. To address this issue, we explore the dynamic priors of tactile motion and discover that the correlation between transient and cumulative motion can explicitly distinguish fine-grained contact states. Based on this insight, we propose a motion-aware tactile representation to facilitate contact-rich manipulation. Beyond tactile representation, effective fusion of tactile and visual modalities is also critical. Most existing fusion methods either directly concatenate features from each modality or train modality-specific networks separately and fuse their outputs. However, these strategies struggle to simultaneously model cross-modal interactions and preserve modality-specific characteristics. In this work, we take advantage of the Mixture-of-Transformers architecture and propose a unified modality-aware visuo-tactile policy that captures cross-modal complementarity while maintaining modality-specific properties.

cs.RO↗