Searcharxiv⌕ Search

arXiv subjects

Wenjun Zeng

Publications and source records attributed to Wenjun Zeng.

At least 19 recordsLinked to original sources
arXiv2026

ReWorld: Multi-Dimensional Reward Modeling for Embodied Video Generation

Current embodied video generation models primarily emphasize visual generative quality while overlooking physical fidelity, dynamic consistency, and task logic, especially for contact-rich manipulation tasks, which limits their applicability to downstream tasks. In this paper, we introduce ReWorld, a framework aimed to employ reinforcement learning to align the video generation with physical realism, task completion capability, embodiment plausibility and visual quality. Specifically, we first construct a large-scale (~235K) video preference dataset and employ it to train a hierarchical reward model designed to capture multi-dimensional reward consistent with human preferences. We further propose a practical alignment algorithm that post-trains flow-based video generators using this reward through a computationally efficient PPO-style algorithm. Comprehensive experiments and theoretical analysis demonstrate that ReWorld significantly improves the physical fidelity, logical coherence, embodiment and visual quality of generated rollouts, outperforming previous methods.

cs.RO↗
arXiv2026

GeoNaviDiff: Transferring Disentangled Semantic Controls to Diffusion Features via Local Riemannian Navigation

Disentangled real-image manipulation requires factor-specific semantic control while preserving image fidelity and non-target content. StyleSpace $S$ provides interpretable controls but is limited by GAN inversion, whereas diffusion models offer stronger image realization without an explicit factorized control basis. We introduce GeoNaviDiff, an unsupervised cross-space framework that transfers disentangled StyleSpace interventions to local Riemannian trajectories in diffusion $h$-space. GeoNaviDiff constructs reconstruction-decoupled supervision by extracting a spherical Log-map tangent from a matched StyleGAN reconstruction-edit pair and parallel-transporting it to the real-image diffusion state. A shared, factor-conditioned Navigator predicts this tangent at inference, while norm-preserving geodesic realization and source-anchored rollout stabilize the reverse process without edited references or factor-specific models. Experiments across multiple datasets show that GeoNaviDiff outperforms the compared methods in fidelity, factor selectivity, edit locality, and trajectory smoothness.

cs.CV↗
arXiv2026

Euler: Towards Human-Centered Multi-Agent Mathematical Discovery

Mathematical research drives scientific discovery, and recent advances in AI for mathematics have shown that language-model agents can solve challenging problems and develop substantial mathematical proofs. Yet much of today's AI-for-mathematics work centers on obtaining a final answer. This emphasis leaves the exploration behind a result difficult for researchers to understand, while long-running investigations face fragmented context and the risk of unsupported claims propagating across reasoning steps. We present Euler, a human-centered mathematical research system built around a shared, structured representation of the research process. Euler organizes mathematical statements, assumptions, proof dependencies, alternative approaches, failed attempts, and verification evidence into connected research graphs. This representation allows researchers to trace how results are developed and supplies agents with persistent, task-relevant knowledge for planning, collaboration, and verification across sessions. Researchers can also select a graph node linked to a saved team state, revisit the agents' earlier reasoning, and explore alternative directions through continued interaction. We illustrate Euler's capabilities through three mathematical research case studies and describe the implemented system and its software validation.

cs.AI↗
arXiv2026

APM-Bench: Benchmarking Cross-session Persistent Memory for Egocentric Streaming Video Assistants

To serve as real-world personal assistants, streaming video models need persistent memory that retains past experiences for later use. Yet existing streaming benchmarks and methods often focus on individual continuous videos or short clips, overlooking that real-world interactions are often intermittent and require memory to persist across interruptions. To fill this gap, we introduce APM-Bench, which reformulates real-world streaming interaction as multi-session life trajectories. It contains 549 sessions, 104 trajectories, and 2,719 candidates, spanning both objective and open-ended questions. Each session is a video with fine-grained annotations, and sessions within a trajectory revolve around related activities. Models then use persistent memory to answer questions about past sessions and provide proactive responses while maintaining real-time interaction. This raises challenges: persistent memory must be storable, selectively retain information, be injected at the right time, and remain efficient. Moreover, finite storage may leave required evidence unavailable, so assistants should recognize missing evidence. Therefore, we systematically evaluate general video models under different memory protocols and diverse specialized streaming memory systems, and test whether models acknowledge insufficient evidence. Our evaluation reveals a clear utility--latency--storage trade-off: existing methods still struggle to simultaneously achieve reliable long-term recall, low overhead, and effective proactive assistance across sessions. APM-Bench provides a comprehensive testbed for developing and comparing persistent memory systems under realistic streaming conditions. We hope it encourages future work that jointly considers utility, latency, and storage toward more practical persistent memory for real-world streaming assistants.

cs.CV↗
arXiv2026

GEM-Occ: From Visual Geometry Evidence to Embodied Semantic Occupancy Memory

Embodied agents exploring indoor environments require reliable semantic occupancy memory that persists across observations and revisits. Building such memory is challenging because each observation provides incomplete and uncertain geometric and semantic evidence. We introduce GEM-Occ, a Gaussian Evidence Memory framework that consolidates evidence accumulated over time into persistent semantic occupancy memory. Local predictions are converted into occupied semantic Gaussians and free-space ray evidence. Confidence- and visibility-aware causal updates integrate supporting observations, suppress occupancy contradicted by observed free space, and preserve previously observed structures through occlusion. A hierarchical memory organization supports continued mapping and efficient queries across connected indoor spaces. To evaluate this capability, we introduce HIOcc, a unified benchmark for embodied semantic occupancy memory. HIOcc establishes a shared semantic label space and evaluation framework spanning local prediction, room-level online mapping, and building-level mapping, while accommodating perspective and panoramic observations. Experiments on HIOcc demonstrate that GEM-Occ outperforms existing methods, enabling accurate semantic occupancy prediction and consistent online mapping across spatial scales with efficient memory usage and fast occupancy queries.

cs.RO↗
arXiv2026

EVOKE: Eliciting World Knowledge in Agents for Transferable Decision-Making

Large language models (LLMs) are increasingly deployed as agents for multi-step decision-making, yet transfer poorly to unseen environments. World-model methods address this by training agents to predict future observations, at the cost of additional training and errors that compound when predictions are used for planning. However, for LLM agents operating in digital environments, much of this world knowledge is already internalized during pretraining, which shifts the problem from acquiring it to eliciting it. We argue that typical post-training provides little pressure for such elicitation, since supervision under a single goal at each visited state inadvertently drives policies to rely on superficial contextual habits. We introduce EVOKE, a post-training method that supplies this pressure through goal diversity at fixed states. Motivated by theory showing that an agent competent across diverse goals must encode a world model recoverable from its action preferences, EVOKE holds the environment state and interaction history fixed and ranks the same candidate actions under alternative goals, forcing action preferences to change, so that a policy relying on contextual habits or single-goal correlations cannot order them correctly. This implicitly elicits the policy's pretrained world knowledge to inform decisions. We evaluate EVOKE across diverse tasks in three backbones, demonstrating improved task performance, unseen environment generalization, and data efficiency. We further conduct controlled analyses to better understand what drives these gains. These findings offer a new perspective on eliciting internalized world knowledge for transferable action through direct decision supervision.

cs.CL↗
arXiv2026

From Articulated Kinematics to Routed Visual Control for Action-Conditioned Surgical Video Generation

Action-conditioned surgical video generation is a critical yet highly challenging problem for robotic surgery. The core difficulty is that low-dimensional control vectors must precisely govern complex image-space evolution. In this work, we propose a kinematic-to-visual lifting paradigm that converts articulated kinematics into a unified set of five image-aligned control modalities. Building on this representation, we introduce a hierarchically routed visual control framework that selectively activates the most relevant control modalities and motion scales. Instead of uniformly applying all control signals, our model performs hierarchical routing to dynamically allocate conditioning capacity. We further design kinematic-prior-guided routing loss functions to ensure physically meaningful, temporally stable, and efficient expert utilization. To improve efficiency, we propose a budgeted training and inference scheme that leverages routing-induced sparsity. By selectively discarding low-significance control pathways during training and execution, our approach enables adaptive computation that is complementary to standard distillation. We additionally construct a new benchmark with curated articulated annotations, obtained through human-in-the-loop semantic labeling and differentiable pose tracking, providing realistic supervision for action-conditioned surgical video generation. Extensive experiments demonstrate that our method consistently improves action faithfulness, visual fidelity, and cross-domain generalization over diverse baselines. Moreover, our efficient variant achieves substantial reductions in latency while maintaining strong control accuracy.

cs.CV↗
arXiv2026

FAN: Foresight Action Normalization for Continual Adaptation of Vision-Language-Action Models

Vision-Language-Action (VLA) models pre-trained on large-scale, closed datasets have demonstrated remarkable success across diverse robotic manipulation tasks. However, their long-term real-world deployment necessitates continuously acquiring new skills while retaining previously learned capabilities. While pioneering works have explored continual VLA adaptation using techniques such as experience replay and reinforcement fine-tuning, they overlook a foundational mechanism: action normalization, which determines the underlying coordinate system in which policies perceive and execute physical actions. To bridge this gap, we systematically evaluate five normalization strategies across four real-world task streams covering single-arm and bimanual manipulation. Our analysis reveals that existing protocols induce severe failure modes due to inter-task coordinate drift, limited motion coverage, or train-test coordinate mismatches. Motivated by these insights, we formulate three core design principles: consistency, coverage, and causality (3C), and introduce foresight action normalization (FAN). FAN estimates normalization statistics once from a small, task-independent calibration set prior to continual learning and freezes them throughout adaptation. Across all evaluated streams, FAN achieves the highest performance and demonstrates consistent robustness, providing insightful guidance for building stable action representations in achieving effective lifelong VLA adaptation.

cs.RO↗
arXiv2026

Visual Autoregressive Priors for RAW-to-sRGB Image Signal Processing

RAW-to-sRGB image signal processing (ISP) must recover perceptually faithful colors and fine details from sensor measurements, often under imperfect spatial alignment and missing camera metadata. This paper presents, to the best of our knowledge, the first application of visual autoregressive (VAR) next-scale prediction over a discrete image codebook to the RAW-to-sRGB ISP task. We adapt a frozen 1.10\,B-parameter VAR backbone for RAW-conditioned ISP with only 32.93\,M trainable parameters (2.99\%), and propose a frequency-decomposed color loss that separately supervises low-frequency tone via wavelet LL cosine similarity and chromatic edges via detail-band $\ell_1$. On the Zurich RAW-to-sRGB benchmark, the method improves PSNR-Y from 21.31 to 21.89\,dB and reduces LPIPS from 0.276 to 0.218 on the full 1,204-image test set. Diagnostic experiments show that the VAR prior preserves structure well, but continuous color transfer remains the dominant bottleneck: oracle affine correction recovers 3.8\,dB, while learned color heads yield marginal gains.

cs.CV↗
arXiv2026

Inter-X++: A Comprehensive Benchmark for Multimodal Human-Human Interaction Analysis

The capability to perceive and synthesize human-human interactions is fundamental to developing intelligent digital human systems. However, existing datasets and modeling approaches are fundamentally constrained by low-fidelity kinematics, the omission of dexterous hand gestures and a severe lack of rich multimodal annotations. Furthermore, fragmented interaction representations and inconsistent evaluation protocols also impede fair and rigorous benchmarking. To systematically address these bottlenecks, we present Inter-X++, a comprehensive and large-scale benchmark designed to empower versatile HHI analysis. Captured via a novel hybrid motion capture system, Inter-X++ provides 11,388 high-fidelity interaction sequences and over 8.1M frames, featuring precise whole-body movements and detailed finger articulations. Meanwhile, we enrich the data foundation with multifaceted annotations, including hierarchical fine-grained textual descriptions, interaction categories, causal interaction orders, the relationship and personality of the subjects, as well as vertex-level contact maps and physically regularized constraints. Leveraging these elaborate annotations, we formulate a unified testing ground comprising four categories of downstream tasks that symmetrically span both generative and perceptive paradigms. To eliminate benchmarking ambiguities, we systematically standardize the interaction representations and evaluation protocols. Finally, we go beyond dataset construction to propose OpenHHI, a single and unified HHI representation and modeling framework that jointly optimizes interaction reconstruction and semantic understanding. Extensive experiments reveal that OpenHHI achieves state-of-the-art performance on both generation and perception tasks. This definitively proves that our unified representation successfully bridges interaction understanding and generation simultaneously.

cs.CV↗
arXiv2026

Dual Distribution Estimation for Zero-shot Noisy Test-Time Adaptation with VLMs

While test-time adaptation (TTA) empowers vision-language models to adapt without costly retraining, it remains highly vulnerable to out-of-distribution (OOD) outliers prevalent in real-world applications. This discrepancy motivates Noisy TTA (NTTA), an online task to filter noisy OOD samples on the fly while maximizing in-distribution (ID) classification accuracy. Existing zero-shot NTTA approaches typically rely on test-time discriminative training, leading to overconfident misclassifications and significantly degraded inference efficiency. To address these limitations, we propose a novel framework named Dual Distribution Estimation (DDE), shifting the zero-shot NTTA paradigm from instance-level learning to training-free Gaussian distribution modeling. DDE incorporates two novel modules: Positive Feature Distribution Estimation (PFDE) and Negative Label Distribution Estimation (NLDE). PFDE explicitly models class-wise inclusion and exclusion Gaussian distributions to formulate a calibrated contrastive score, robustly enhancing ID accuracy. In parallel, NLDE improves OOD identification by explicitly modeling the negative label distribution to mine highly discriminative labels, effectively mitigating spurious correlations. Extensive experiments show that on the large-scale ImageNet benchmark, DDE achieves an improvement of 3.70\% in harmonic mean accuracy and reduces the FPR95 for OOD detection by 6.20\%, while ensuring highly scalable and efficient online inference. Furthermore, DDE is zero-shot and training-free, demonstrating remarkable robustness in data-scarce scenarios. Codes are available at https://github.com/ZhuWenjie98/DDE.

cs.CV↗
arXiv2026

Can Vision-Language-Action Models Learn from Real-World Data Continually without Forgetting?

Vision-Language-Action (VLA) models provide a promising foundation for general-purpose robotics, yet their real-world deployment demands the ability to continually acquire new skills without forgetting prior ones. While recent studies have explored continual learning for VLA models in simulated settings, the challenge remains largely unexamined under realistic physical conditions. To bridge this gap, we construct a real-world continual learning benchmark comprising ten diverse sequential manipulation tasks across both single-arm and bimanual configurations. Through extensive experiments on this benchmark, we find that naive sequential fine-tuning leads to severe catastrophic forgetting, whereas a well-configured experience replay (ER) approach can effectively mitigate forgetting and outperform joint multi-task training under equivalent computational budgets. Notably, by synthesizing our empirical findings, we successfully achieve stable continual learning across the full 10-task heterogeneous stream, retaining previously acquired capabilities while adapting to diverse new skills in real-world deployment. This work presents an empirical study grounded in real-world continual VLA learning and offers actionable insights for deploying robust, long-lived robotic policies.

cs.RO↗
arXiv2026

MMOOC: A Comprehensive Benchmark for Out-of-Context Evaluation in Multimodal Large Language Models

Multimodal Large Language Models (MLLMs) have achieved strong performance on a wide range of vision-language tasks, but often fail under imperfect or shifted contexts. A reliable MLLM should refuse truly out-of-context (OOC) questions with subject-level context shifts while still answering shifted in-context (Shifted IC) questions with non-subject context shifts. Existing benchmarks mainly target OOC or visually unanswerable questions, but overlook answerable Shifted IC cases and cover limited OOC shifts. To fill this gap, we present MMOOC, a large-scale benchmark for evaluating refusal and robust answering abilities of MLLMs. MMOOC contains over 41K image-question pairs, including answerable Shifted IC cases and unanswerable OOC cases, spanning three question formats, eight shift types and six visual scenarios, with data quality ensured through MLLM-based filtering and human verification. We evaluate model responses using Accuracy and Refusal Rate, and further introduce an LLM-as-a-Judge metric to assess the correctness of model reasoning. Experiments on diverse MLLMs show that current models still struggle to balance answer-ability and refusal under shifted contexts. We further analyze key failure patterns and show that post-training can improve robustness. MMOOC will be made publicly available.

cs.CV↗
arXiv2026

RPG-VST: Robust Poisson-Gaussian Variance Stabilization for Blind RAW Denoising

Variance stabilization with the generalized Anscombe transform (GAT) enables frozen Gaussian denoisers to process Poisson--Gaussian (PG) RAW noise, but its reliability depends on fitted shot/read-noise parameters. In blind single-image deployment, these parameters are estimated from low-texture RAW statistics that are often corrupted by residual texture, clipping, defective pixels, and read-noise floors. Such contamination yields heavy-tailed log-variance residuals, making ordinary least-squares PG calibration brittle and causing severe tail failures despite favorable average PSNR. We propose RPG-VST, a robust no-reference variance-stabilization framework for blind RAW denoising. RPG-VST estimates PG parameters separately for each color filter array (CFA) plane using a Student-$t$ log-variance objective with robust tile statistics and physical constraints. It then estimates the stabilized-domain noise level $σ_z$ from tile variance ratios and uses it as a reliability signal. For each image, RPG-VST selects the robust fit or the conventional OLS fit according to which produces $σ_z$ closer to unit variance, requiring no clean reference or learned threshold. On SID Sony SID$50$, SIDD, and ELD with frozen SwinIR and Restormer denoisers, RPG-VST improves mean PSNR in all six dataset--backbone settings. It reduces severe tails, defined as cases whose PSNR gain over Direct is below $-1$ dB, in four settings and leaves them unchanged in the other two. On SIDD, it yields $+1.83/+1.92$ dB and reduces severe tails from $44/36$ to $7/4$. Ablations show that the $σ_z$ gate prevents regressions of ungated robust fitting on read-noise-dominated ELD captures.

eess.SP↗
arXiv2026

Bridging 3D Gaussians and Semantic Occupancy for Comprehensive Open-Vocabulary Scene Understanding from Unposed Images

Comprehensive 3D scene understanding from sparse, unposed images requires a model to recover renderable geometry, open-vocabulary semantics, and free/occupied 3D space without relying on external camera calibration. Recent feed-forward Gaussian methods improve pose-free reconstruction and semantic rendering, but their Gaussian primitives are mainly optimized through image-space objectives and remain weakly constrained in unobserved regions. We propose \textit{COVScene}, a pose-free semantic Gaussian framework that couples renderable Gaussian primitives with a dense semantic occupancy field through differentiable volumetric lifting. Instead of converting Gaussians to voxels only at evaluation time, COVScene lifts the predicted semantic Gaussians inside the training computation graph, so volumetric regularization provides gradients to Gaussian opacity, geometry, and semantic features. The framework combines a semantic-aware Geometry Transformer, multi-task Gaussian decoding, geometric foundation distillation, and occupancy entropy regularization to support novel view synthesis, open-vocabulary semantic querying, and semantic occupancy prediction within a single representation. Experiments on ScanNet and ScanNet++ show that COVScene maintains competitive rendering quality, improves open-vocabulary segmentation, and achieves stronger semantic occupancy prediction than the self-supervised baseline without direct voxel-level supervision.

cs.CV↗
arXiv2026

OmniNWM: Omniscient Driving Navigation World Models

Autonomous driving world models are expected to work effectively across three core dimensions: state, action, and reward. However, existing methods are typically restricted to fragmented modality modeling, short-horizon drift, and imprecise action control, while lacking intrinsic mechanisms for policy evaluation. In this paper, we introduce OmniNWM, an Omniscient panoramic Navigation World Model that addresses all three dimensions within a consistent probabilistic framework. For State, OmniNWM generates panoramic videos of RGB, semantics, metric depth, and 3D occupancy, ensuring pixel-level alignment across modalities with joint distribution modeling. To mitigate autoregressive exposure bias, we propose a structured panoramic forcing strategy to stabilize long-horizon generation via stochastic manifold thickening. For Action, we introduce canonical geometric action encoding with normalized panoramic Plücker ray-maps. This representation decouples motion dynamics from sensor intrinsics, enabling precise, zero-shot trajectory control across heterogeneous datasets and camera configurations. For Reward, we derive intrinsic occupancy-grounded dense rewards directly from generated 3D volumes, establishing a reliable closed-loop simulation cycle for evaluating diverse planning agents. Extensive experiments demonstrate that OmniNWM achieves SOTA performance in generation fidelity and control precision, with remarkable zero-shot robustness to novel scenes on NuPlan and in-house datasets with distinct camera rigs. Project page is available at https://arlo0o.github.io/OmniNWM/.

cs.CV↗
arXiv2026

DeformGen: Dynamics-Based Topology Augmentation for Deformable Manipulation Policy Learning

Demonstration augmentation is proposed for cost-efficient data acquisition, but existing methods are fundamentally limited in deformable manipulation due to two challenges: (1) the state space is high-dimensional with physics-induced constraints, making valid configurations impossible to reach via low-dimensional pose perturbations; and (2) trajectory transfer is non-equivariant, as material points no longer move rigidly together under deformation. We present DeformGen, a dynamics-based augmentation framework that achieves topological diversity for deformable objects. For the state challenge, DeformGen expands the valid state distribution by applying localized physical disturbances and forward-simulating the dynamics to obtain topology-coherent, physically plausible deformable states. For the trajectory challenge, DeformGen transfers source manipulation trajectories via deformation-field warping, which lifts per-particle displacements into a continuous spatial function to adapt the end-effector trajectory consistently with the deformed geometry. In this way, our method jointly augments the state distribution and its associated manipulation behavior. Experiments on high-fidelity deformable manipulation benchmarks show that DeformGen generally improves policy learning compared with training on the original demonstrations alone and with rigid-style augmentation baselines.

cs.RO↗
arXiv2026

ImageWAM: Do World Action Models Really Need Video Generation, or Just Image Editing?

World Action Models (WAMs) commonly rely on video generation to bridge visual world modeling and robot control. However, video-based WAMs face three coupled limitations: dense multi-frame future tokens make inference costly, full video prediction spends capacity on action-irrelevant temporal and appearance details, and long-horizon future imagination may introduce errors that mislead action prediction. These issues raise a simple question: Does world action model really need video generation? We propose ImageWAM, a simple WAM framework that repurposes pretrained image editing models for robot action prediction. In contrast to video generation, image editing provides a better-matched prior: it only needs to model a target-frame transformation, focuses on action-relevant current-to-target visual differences, and grounds task instructions to localized visual changes through edit pretraining. In practice, ImageWAM does not decode the target frame at inference time; instead, it conditions a flow-matching action expert on the KV caches produced by image-editing denoising, using them as a compact world-action context. ImageWAM outperforms standard VLA baselines and matching competitive WAMs without additional policy pretraining across different simulator and real-world experiments. It also reduces FLOPs to 1/6 and latency to 1/4 of video-based WAMs. Attention analysis further shows that editing caches focus on task-relevant change regions, supporting image editing as an effective alternative to video-based world-action modeling.

cs.CV↗