跳到正文
只看论文
10月22日周四
  1. arXiv · 场景与世界生成

    DeXplicit: Making Physical State Explicit in Dexterous Manipulation World Models

    Before acting on an object, a robot needs to anticipate how its hand motion will change the physical world and what the resulting scene will look like. Existing approaches typically entangle action-conditioned dynamics, camera motion, and visual synthesis within a video generation model, or represent dynamics as point trajectories without explicit object structure. We introduce DeXplicit, a dexterous manipulation world model that instead predicts how the world evolves explicitly in metric 3D before rendering its visual consequences. Given an initial scene and a future hand trajectory, a lightweight 3D dynamics network predicts object-centric motion and deformation, providing structured and geometrically grounded future world states. These predictions then serve as a strong scaffold for video generation: appearance is transported through the predicted 3D motion and projected into future views, allowing the renderer to focus on completing uncertain and newly visible content rather than implicitly learning world dynamics from pixels. This explicit-3D-first decomposition also enables efficient causal generation, requiring only two denoising steps per block while conditioning on generated history. Across HOT3D, OakInk2, and PhysTwin, DeXplicit reduces moved-point error by 27-40% relative to PointWorld and improves object fidelity IoU+ by 47% on HOT3D under a shared renderer. Our design with two-step causal generation achieves a 57x speedup from our 14B bi-directional configuration in video synthesis.

  2. arXiv · 资产数据集与许可

    Evaluation of Front-End Feature Matching for Lunar Visual SLAM

    Lunar rovers will primarily rely on cameras for localization until navigation satellite services are deployed, and will continue to need them afterwards for relative navigation, hazard detection, and surface operations. Visual odometry (VO) and Simultaneous Localization and Mapping (SLAM) mostly estimate motion by matching features between frames. However, feature matchers are developed on urban, indoor and driving scenes, planetary studies compare them on orbital registration or within complete SLAM systems, and real lunar imagery with reference poses is scarce, so matching between rover frames on low-texture, harshly lit terrain, and its effect on VO, has not been compared. We benchmark eleven classical and learned detector-matcher combinations on nine synthetic, lunar-testbed, and terrestrial-analogue datasets, and a feature tracker, scoring pairwise pose accuracy against baseline and the trajectory error of the resulting VO. Under a common rotation-only metric, the SuperPoint pairings outperform SIFT with ratio-tested FLANN matching on both lunar testbeds and one analogue site, with smaller or unresolved differences on synthetic data. When rotation and translation direction are scored jointly, this SIFT pairing exceeds both SuperPoint pairings on four synthetic datasets, so synthetic results alone do not establish relative performance on real rover imagery. In stereo VO, the SuperPoint pairings are not consistently more accurate than SIFT, but only they keep tracking on all eight sequential datasets. Pairings with classical detectors depend on lighting while SuperPoint changes little: a low spotlight makes them more accurate than diffuse light on the same testbed views, consistent with more contrast and detected keypoints. SuperPoint with LightGlue is the fastest learned combination (16 ms per frame), and DINOv2 retrieval proposals still need geometric verification.

  3. arXiv · 场景与世界生成

    Anticipating Social Consequences with Distributional World Models

    Social world models predict how interactions develop after an action, helping agents negotiate, cooperate and mediate. In these situations, one action can lead to several plausible futures, hence predicting only one future can overlook consequential alternatives, while exploring them through repeated simulations increases inference cost. We propose DSWM, a Distributional Social World Model that predicts alternative responses, their probabilities and immediate and final outcome values in one output per candidate action. We train a 4B model on repeated simulations using supervised fine-tuning and GRPO. At inference time, the actor uses these possible consequences to choose an action. Across three seen and two unseen environments, with interaction backbones absent from training collection, DSWM achieves the best average performance on 13 of 15 metrics. Ablations show that retaining multiple futures and their predicted probabilities improves outcomes in all five environments.

  4. arXiv · 可控视频与镜头

    Auto-Director: Leveraging Multi-Agent Collaboration for Autonomous Video Editing

    Autonomous video editing from open-world resources remains a challenging problem in multimedia content creation. Retrieval-based methods are constrained by fixed source material and lack creative flexibility, while generative models struggle with long-horizon coherence and often diverge from user intent. Both paradigms fail to replicate the structured, multi-stage reasoning that underlies professional post-production. To address these limitations, we present Auto-Director, a tool-augmented multi-agent framework that mirrors professional editorial workflows: a Planning Agent decomposes abstract topics into fact-grounded scripts via retrieval-augmented planning; a Material Agent indexes open-world footage through shot segmentation and visual tagging; an Editor Agent assembles semantically aligned clips with MLLM-driven spatial adaptation; and a Critic Agent governs an iterative Reflect-and-Revise loop that checks scene-level alignment and pacing. To rigorously evaluate this new paradigm, we introduce VEdit-Bench, a benchmark of 100 expert-curated topics with human-annotated Golden Scripts and a five-dimensional metric suite covering Narrative Logic, Factual Accuracy, Semantic Alignment, Compositional Quality, and Stylistic Cohesion. Extensive experiments demonstrate that Auto-Director significantly outperforms baseline methods in both automated metrics and human evaluation, effectively bridging the gap between amateurs and professional-grade video production.

  5. arXiv · 场景与世界生成

    From Single-Actor Environments to Multi-Actor Event Worlds: Can LLMs Be Event World Models?

    World models help agents assess the consequences of their actions before acting. Many real systems, however, involve multiple actors who make their own decisions, can evolve even when the agent takes no action, and are only partially observed through event logs. We formalize these systems as event worlds and ask whether large language models can predict future observations from the log and the agent's proposed actions. We build complementary benchmarks from human Diplomacy games and real Polymarket trades. Our hierarchical framework compares prediction with and without other actors' actions, then tests multi-step rollouts, predictions for actions never taken, and usefulness to agents. Fine-tuning and more data improve prediction when all actors' recorded actions are given. With only the agent's action, models struggle when other actors strongly affect the outcome, and larger models and longer histories help little. This gap also limits multi-step prediction and agent decisions, with models overestimating plan success. Yet imagined experience and initialization from world models can still improve policy learning. Anticipating other actors' actions remains a central challenge for event world modeling.

  6. arXiv · 可控视频与镜头

    AVAtt: A Free-Viewing Eye-Tracking Dataset and Benchmark for Visual Attention on Paintings

    Saliency and scanpath models are developed almost exclusively on natural photographs, yet paintings are composed by artists to guide the viewer's eye and differ from photographs in texture, colour and abstraction. We introduce \avatt, a free-viewing eye-tracking dataset on \nImages{} paintings spanning sixteen art movements and five centuries, with \nScanpaths{} scanpaths and \nFixations{} fixations from \nParticipants{} participants (\spPerImageMin--\spPerImageMax{} observers per painting). For every painting we release individual scanpaths with fixation timing and eye--screen distance, fixation maps, saliency maps smoothed at one degree of visual angle, temporal saliency videos and rich art metadata. We benchmark classic and deep saliency models against human agreement, revealing substantial headroom for art-aware models, and rank 44 scanpath-similarity measures, including a new language-free family of image-encoder metrics built on ten CNN, ViT and self-supervised encoders. With controls that remove viewing biases, we show that faces are fixated \faceEnrich$\times$ more than expected and within half a second, that local structure draws the eye, that composition rules fitted from image content predict where people look, and that art style shapes how much viewers agree. Dataset, code and a Python package are available at: https://huggingface.co/datasets/Amine-CV/AVAtt.

  7. arXiv · 创作 Agent 与装配

    Externalise the Memory: A Pre-Registered Study of Where an Embodied Agent's Experience Should Live

    Should an embodied agent's experience be baked into its policy weights, or kept external and retrieved at inference? A memory-augmented teacher (a VLM that retrieves lessons, with their episode images, from its own past attempts) produces trajectories on a vision-conditioned reach task; from the same trajectories we distil LoRA student policies and evaluate them under matched protocols, including one that re-attaches the teacher's frozen retrieval buffer to a fixed set of student weights. We pre-registered the hypotheses, arms, tests, and analysis rules before training and before any p-value. Distilling the memory-augmented behaviour into weights beat an action-only control by +1 pp (n.s.); the registered test proves underpowered for its own minimum effect, so we withdraw - rather than refute - the memory-in-weights claim. Attaching retrieval to a fixed set of student weights added +34 pp (z=5.15), reaching the pooled memory-augmented teacher reference (49.3%); the pre-registered protocol contrast agrees (+23 pp, z=3.36); under a pre-registered auto-reset sensitivity count the two are +36 and +29 pp. An exploratory behavioural probe with pre-registered per-arm predictions is consistent with behavioural distillation transferring the marginal correction but not the conditional, situation-specific correction that retrieval supplies. The recipe that follows is to externalise the memory: in follow-up controls registered before they ran, the same frozen buffer lifts the base model with no adapter to 48% (sensitivity 48%) and the other three adapters from 19-26% to 53-55% (sensitivity 17-21% to 50-55%), and the adapter adds nothing detectable once the buffer is attached. On a second, harder task, the retrieval advantage points the same direction but falls short of significance (+6 pp, 95% CI [-4.5, +16.4]); exploratory diagnostics tie the attenuation to memory coverage.

  8. arXiv · 渲染、神经场与导出

    SplatLab: Autonomous Scientific Discovery for 3DGS Research

    Reproducing a published 3DGS paper takes weeks of expert engineering before any extension can be proposed. Prior autonomous-research systems read papers, draft hypotheses, and write code, but are limited by short fixed templates, low executability, and weak traceability, and none has been demonstrated end-to-end in a live computer-vision subfield with intense activity. SplatLab is an autonomous research system for 3D Gaussian Splatting that addresses these limitations. Given a research prompt, a target-paper URL, or no input at all, it reproduces a published baseline, proposes an extension, trains and ablates across Mip-NeRF 360 and LLFF, writes the manuscript, and binds every reported number to a logged training run. Five design choices distinguish it: (i) candidates mutate plugins from a verified-reproduction backbone matched to published baselines; (ii) a cascade-weighted Elo tournament across eight islands scores candidates on measured GPU PSNR with verbal-gradient feedback; (iii) a standing falsifier subtracts Elo on failure, integrating stress-test survival into selection; (iv) every reported number is bound to a logged training run, and the run cannot terminate while any claim is unbacked; (v) a seven-category 3DGS taxonomy wraps EVOLVE-BLOCK markers in domain-aligned scaffolds transferable across 3DGS repositories. On PSNR-targeted runs, SplatLab's manuscripts beat their reproduced targets by +0.18 to +3.78 dB, including +3.78 dB on astrophotographic nebula rendering, for which no prior 3DGS method exists. The research cycle compresses from months of graduate-student time to days of mostly autonomous compute. Code, all manuscripts, experiment ledgers, and the cross-run paper store are available on the project page.

  9. arXiv · 渲染、神经场与导出

    Learning from Automatic 3D Gaussian Splatting Edits for Online Bayesian Scene Change Detection

    Online scene change detection requires integrating incoming observations to determine which regions have changed relative to a reference scene. We update change beliefs as new images arrive using generalised Bayesian evidence fusion. However, the prior probability of change and evidence models are unknown beforehand, and numerical Bayesian inference can be costly. We address these challenges by learning offline from automatic edits to a reference 3D Gaussian Splatting (3DGS) model. These edits generate labelled training views without manual annotations. Together with unchanged reference observations, they provide supervision for learning scene-specific evidence models and estimating the effective pixel prior and response statistics. Online prediction combines current evidence with historical support, while closed-form local Bayes factors, parameterized by the learned response statistics, update an accumulated 3D Gaussian field. Object-guided refinement produces the final change masks. Our method achieves state-of-the-art accuracy on PASLCD and CCS3D, with $67.64\%$ and $75.82\%$ mIoU, respectively. On PASLCD, the complete pipeline runs at $17.2$\,FPS over complete query sequences, the fastest among the evaluated online methods. Project page: https://changruizhu96.github.io/work/lego-scd/

  10. arXiv · 可控视频与镜头

    Latent Space Jam: Layout Guided Video Synthesis

    Adding control to video generation often requires training, which demands additional computational resources, data, or architectural changes. We propose a training-free approach that utilizes pretrained video diffusion models to control object layout and motion using coarse layouts. Our method operates in two passes: first, it steers spatiotemporal placement through prompt-guided cross-attention to produce a coarse visual guide; second, this guide conditions the same model to generate a high-quality, layoutconsistent video. The approach enables coordinated motion with strong temporal consistency, supporting diverse trajectories and object interactions without additional training. Extensive experiments show that our method consistently outperforms prior work in overlap and trajectory accuracy, as well as visual quality across static, dynamic, and multi-object scene settings.

  11. arXiv · 渲染、神经场与导出

    R4DGS: Referring Segmentation in 4D Gaussian Splatting

    Dynamic representations based on 4D Gaussian Splatting (4DGS) have shown strong capabilities in reconstruction and rendering, yet current methods remain limited in supporting complex natural-language referring in dynamic scenes. Existing 4D Gaussian methods are primarily designed for rendering-oriented dynamic modeling, while current semantic extensions often rely on scene-specific semantic optimization or are evaluated on relatively simple query settings. In this paper, we study Referring 4D Gaussian Splatting (R4DGS), a task for grounding temporally varying, multi-target or reasoning-intensive, and zero-target or distractor expressions in dynamic 4D scenes. To support this task, we introduce R4D-Bench-QA, a benchmark with structured query annotations. We further present ReferGaussian, a unified framework built upon 4DGS that converts low-level dynamic Gaussians into persistent, language-addressable entities for training-free referring inference. ReferGaussian organizes mask-supported Gaussian entities in an EntityBank, uses Multimodal Large Language Models (MLLMs) to decompose complex queries into trackable object phrases and structured referring constraints, and performs constraint-aware grounding without scene-specific semantic optimization. Experiments on R4D-Bench-QA and the 4D LangSplat protocol demonstrate improved temporal target selection and competitive spatiotemporal grounding without scene-specific semantic training. Project resources are available at https://chenbangpu.github.io/R4DGS/.

  12. arXiv · 可控视频与镜头

    Self-Distilling Future Video Representations for World Action Models

    World Action Models (WAMs) have emerged as a promising approach to learning generalist robot policies using rich spatiotemporal priors from pre-trained video diffusion models. A recent line of WAMs improves inference efficiency by conditioning a separate action decoder on intermediate future video representations extracted from a single denoising step, avoiding costly iterative denoising required for video generation. This design makes action prediction depend directly on the quality of these future representations. In this paper, we introduce Future REpresentation learning via self-Distillation with temporal context (FRED), a framework that trains WAMs to improve the quality of future video representations from noise in a single denoising step. Specifically, FRED uses self-distillation to predict future representations from noise, using representations extracted from ground-truth future observations as targets. We further condition the video model on historical observations, providing temporal context for predicting future representations better. Across diverse real-world manipulation scenarios, FRED consistently outperforms competing WAMs. It achieves particularly large gains on long-horizon and out-of-domain tasks under zero-shot environment generalization, and substantially improves human-to-robot skill transfer, reaching 50.0% task success on tasks absent from robot training data and demonstrated only in human videos, compared with 19.4% for the best competing WAM.

  13. arXiv · 渲染、神经场与导出

    TerraTransfer++: Demonstration-Free End-to-End Planner

    End-to-end autonomous driving has achieved state-of-the-art performance on benchmarks and real-world deployments. Its standard training recipe, however, is expensive across all existing paradigms: imitation learning requires fleet-scale driving logs whose behavior distribution only thinly covers the safety-critical tail; closed-loop reinforcement learning in a photorealistic simulator, while offering diverse behaviors, is bottlenecked by sample-efficiency, per-step cost of rendering, and a forward pass through a large vision backbone. Self-play in vectorized simulators changes the economics: at millions of rollout steps per second, one can obtain a state distribution with natural diversity in collisions, near-misses, and recoveries that no driving log contains, and a highly reactive policy emerges from a basic driving-quality reward. Our approach exploits this asymmetry. We pretrain a single policy by self-play, distill its rollouts into a flow-matching trajectory teacher that also plans, and train lightweight adapters on a frozen pretrained vision backbone to reproduce the latent space of either teacher through the action KL divergence, a batch-relational low-rank structural loss, and, for the planner, a trajectory-matching loss. Every target traces back to the self-play policy, so alignment never supervises against a logged trajectory: a paired dataset of (image, scene-state) frames suffices, with no need for the curated expert demonstrations that imitation pretraining is built on. From this one self-play teacher we obtain two end-to-end students: a closed-loop policy that outperforms every published imitation-based end-to-end method on the photorealistic 3D Gaussian splatting closed-loop benchmark HUGSim, and a planner that reaches 59.4 EPDMS on NAVSIM navhard, above every published method including DriveZero-Scale (57.1), and 95.5 PDMS on navtest, without human demonstrations.

  14. arXiv · 资产数据集与许可

    EgoHand: Monocular 3D Hand Reconstruction from Egocentric Video

    We introduce EgoHand, a monocular 3D hand reconstruction model that achieves state-of-the-art performance on egocentric video. Central to our approach is the curation of a high-quality training corpus of 14 million images, combining synthetic data, in-the-wild video, and multi-camera motion capture to balance scene diversity and supervision quality. The model adopts SOMA-Hand as its parametric representation and processes 2D hand crops with a transformer encoder, incorporating camera cues and an auxiliary 2D branch to improve geometric consistency and image alignment. Zero-shot evaluations on HOT3D, ARCTIC, and H2O demonstrate strong out-of-domain generalization, with lower joint-position, root-translation, and acceleration errors than competing baselines at an inference latency of approximately 9.69 ms. A blinded user preference study further shows that participants favor EgoHand's reconstructed hand poses over those from Apple EgoDex and Ropedia Xperience-10M. We release the EgoHand model and source code under the OpenMDW license, allowing free commercial use of the model.

  15. arXiv · 场景与世界生成

    Rethinking Video Autoencoders for Interactive World Models

    Video world models predict future frames from past frames and actions. They serve as simulators in which agents train, plan or are evaluated, and that humans can control interactively. They generate in the latent space of a video autoencoder, whose compression sets their cost per frame and how long each action waits to take effect. Yet world models often reuse the autoencoders of text-to-video generators, built for offline generation with no agent waiting and, without their training data and code, not retrainable at another compression. We therefore ask how, where and how much a world model should compress. We build CADRE, an open family of causal diffusion representation autoencoders spanning multiple spatio-temporal compressions. Its members differ only in a small bottleneck on frozen features and share a diffusion decoder self-distilled to one or a few steps. We train the same world model on each autoencoder in two domains, manipulation and navigation, with two environments each. CADRE forms most of the latency-quality frontier, especially with a moving camera. Our study also leads to four findings. At strong compression, a diffusion decoder is sharper than perceptual regression. In robot manipulation, temporal compression is more effective in the autoencoder than in the world model. Manipulation and navigation call for opposite compressions: temporal compression costs little when the camera is still, and spatial compression when it moves. Surprisingly, with a still camera, coarser spatial tokens hurt motion far more than image quality. Code, models and a live demo are at https://kyutai.org/blog/2026-10-21-cadre

  16. arXiv · 场景与世界生成

    VoxProg: 3D Spatial Reasoning through Voxel Programming

    Language models' abilities to interpret tasks and generate programs offer a promising basis for 3D spatial reasoning. Yet following spatial instructions in 3D scenes requires programs to construct geometric references that language leaves implicit. In this work, we introduce VoxProg, which formulates 3D object placement through voxel programming: a language model translates task interpretations into executable programs over local voxel views of scene surfaces. Because these views index surface geometry on regular grids with explicit coordinates and neighborhood relations, an evidence program can isolate task-relevant regions and boundaries directly from scene meshes, and a downstream pose program can measure the resulting geometry to compute a target $SE(3)$ pose. Visual and numerical feedback from execution lets the model inspect intermediate selections and candidate placements to iteratively revise its code. To evaluate open-ended placement across functional use, geometric fit, and contextual arrangements, we introduce the 71-task VoxProg-Bench benchmark along with an evaluation suite pairing perceptual Intent Alignment and 3D Physical Validity. Across VoxProg-Bench and external benchmarks, VoxProg consistently outperforms relation-based, keypoint-based, and mesh-based baselines.

  17. arXiv · 场景与世界生成

    PhysiCalWorld: In-Context Physics Calibration for World Models

    Existing world models typically perceive a scene through current visual frame or a narrow history window. However, these observations may not reveal the physics properties that determine transition dynamics, such as an object's mass distribution, joint damping, or surface friction. Without calibrating these physics factors for world models, predictions tend to regress toward an empirical prior and induce severe drift in novel test scenes. To this end, we propose PhysiCalWorld, which calibrates action-conditioned world models to the instance-specific dynamics via in-context prompting, without fine-tuning or analytical physics modeling. Our method encodes an exploratory interaction clip collected within the test scene into a compact physics descriptor token, which grounds subsequent predictions in the scene's actual physical behavior. PhysiCalWorld is trained end-to-end with the latent prediction loss alone, without physics labels. At inference time, the descriptor token is extracted once and can be reused across rollouts with negligible compute overhead. We evaluate PhysiCalWorld across four environments covering diverse physics variations and show that it substantially improves rollout prediction accuracy and visual planning success rate. The learned descriptors also form a physically structured latent space and generalize beyond the physics ranges seen during training.

  18. arXiv · 场景与世界生成

    ALPS: Visual State Encoders for World Models

    World models turn visual representations into predictions and actions. We argue that their visual foundation should combine physical-state information, native language alignment, and pixel reconstruction, so that a single representation can support prediction, instruction, and inspection. However, existing encoders typically specialize in only one or two of these capabilities; we therefore introduce ALPS, a pretrained visual state encoder that unifies all three in a single set of patch tokens. Using only still images and text, we pretrain a ViT-B/16 from scratch with language, reconstruction, and spatial correspondence objectives, and then refine it via cross-view self-distillation. In a unified evaluation across fifteen public encoders, ALPS achieves the lowest aggregate LIBERO state-probe error among ViT-B models (matching the 3.5x larger V-JEPA 2-L) while maintaining strong language alignment and accurate pixel reconstruction. Downstream, frozen ALPS tokens reach the highest mean policy success on LIBERO (85.5%) when paired with our native text tower, match or exceed leading encoders on world-model planning across PushT, Wall, and Granular, and allow predicted future tokens to be decoded directly into pixels for visual inspection. We hope this work helps establish visual pretraining as a shared foundation for understanding, prediction, and action in world models.

  19. arXiv · 创作 Agent 与装配

    Multimodal Embodied Agents: A Unified Capability-centric Perspective from Computer-Use to Robot-Use

    Multimodal agents (MMAs) sustain interaction through reasoning, memory, tools, and feedback, most visibly as computer-use agents, while robotic systems (RSs) couple sensing and actuation under physical dynamics. We define multimodal embodied agents (MMEAs) as goal-directed systems that couple multimodal task reasoning with physical action and revise decisions from the resulting feedback. This raises a central question: what changes when a multimodal agent moves from computer-use to robot-use? We introduce PAPAV, a unified capability-centric framework of five recurring functions: Perceive the current state, Anticipate action effects, Plan a feasible course, Act through an interface or body, and Verify the outcome. Defined by function rather than architecture, these capabilities let PAPAV identify what all three share and then compare MMEAs with MMAs and with RSs. Five physical constraints, from partial observability to unverifiable outcomes, leave fewer choices fixed in advance and less room to reverse errors. Across 64 benchmarks, explicit evaluation centers on Act; only 5 assess Anticipate and 3 assess Verify, leaving both largely hidden behind task success. These constraints also frame the open challenges we identify. PAPAV therefore provides a common basis for designing reliable physical agents and measuring progress across all five capabilities rather than task success alone.

  20. arXiv · 角色动作与重定向

    SP-HLM: Sparse-Pose-Driven Motion Generation and Whole-Body Control for Humanoid Loco-Manipulation

    Humanoid loco-manipulation requires an effective interface between high-level task policies and whole-body control. Existing interfaces typically rely on a single action representation for both task intent and whole-body coordination. This may burden the high-level policy or leave one aspect underspecified. We present SP-HLM, a framework built on a sparse-pose-driven interface. The interface combines five task-critical body poses with a learned latent motion prior to represent spatial intent and whole-body coordination in complementary forms. SP-VLA, a vision-language-action model, first predicts task-relevant sparse poses and finger joint positions from multimodal observations and language instructions using asynchronous action-chunk inference. Given these sparse targets, SP-MGM, a motion generation model trained through two-stage distillation, reconstructs coordinated full-body motion in real time. A pretrained motion encoder maps this motion to a compact latent representation. SP-GMT, a general motion tracker, then integrates the explicit sparse targets with this implicit motion prior for whole-body control. Experiments in simulation and on a real humanoid robot demonstrate improved motion tracking, whole-body coordination, and loco-manipulation success over existing approaches. Additional results and videos are available on our \href{https://sp-hlm.github.io/}{\textcolor{blue}{project page}}.

  21. arXiv · 角色动作与重定向

    RLMuscle: Muscle-Driven Character Animation via Reinforcement Learning

    Body deformation caused by muscle contraction plays a vital role in the visual realism of character animation, yet in most 3D animation pipelines artists have to manually specify muscle activations, which is time-consuming and gives no guarantee of physical realism. To address this, we propose a new method in which a reinforcement learning (RL) controller discovers per-muscle activations that track reference motions without manual tuning. Existing RL muscle controllers stop at 1D muscles; we go one step further, keeping the 1D model inside the RL control loop as a fast surrogate, and using a 3D volumetric simulation to produce the final deformation. Because both stages share the very same constitutive law, the muscle deformation is produced by the same activations that generated the motion. We further add a measured electromyography (EMG) prior that steers the redundant activation solution toward real human muscle recruitment. To achieve stable, efficient, and realistic simulation, we propose a novel geometric multigrid for Extended Position-Based Dynamics (XPBD) that is built on a coarse tetrahedral cage with barycentric-coordinate prolongation and closes a V-cycle on tetrahedral meshes. Within this GPU-parallelized solver, we inject the same Hill-type force-length law as an energy-based fiber constraint that turns per-fiber activation into realistic muscle contraction and bulging. We demonstrate volumetric muscle simulation on a 194-muscle human body with a diverse set of motions (walk, kick, swim, overhead-squat, etc.), simulating the full-body model at 18 ms per frame on a single RTX 4090. Our work takes a step toward character animation that is not only visually convincing but also biomechanically grounded. The website with code and videos is at https://chunleili.github.io/project-page-rlmuscle/.

  22. arXiv · 可控视频与镜头

    RealMe: Immersive Volumetric Video from Minimal Captures

    This paper addresses the problem of creating immersive volumetric videos from sparse and low-resolution capture resources. Conventional dense multi-view reconstruction methods can produce high-quality volumetric videos, but their reliance on costly capture setups makes them impractical in real-world scenarios. Recent approaches use generative models to relax capture requirements, but still suffer from limited resolution and 4D inconsistency. To overcome these constraints, we propose RealMe, a unified framework built around a high-resolution video diffusion model that enhances limited captures into high-resolution (2K), 4D-consistent videos for volumetric video reconstruction. Specifically, RealMe constructs auxiliary data from capture videos and encodes them into latents along with layout-aware positional embeddings. The video diffusion model adapts factorized spatio-temporal attention to the multi-view capture layout, denoises these latents across viewpoints and time, and decodes them into enhanced videos to refine the 4D representation. This adaptation helps the model capture temporal dynamics and multi-view relationships, leading to more 4D-consistent generation. To reduce the data gap between generated videos and real captures, we apply a latent frequency loss that provides frequency-domain supervision in latent space. Extensive experiments on the public DNA-Rendering benchmark and our self-captured RealVV dataset demonstrate that RealMe is able to create immersive volumetric videos and outperforms state-of-the-art methods in visual fidelity and 4D consistency.

  23. arXiv · 渲染、神经场与导出

    ClunoGS: Compact Language 3D Gaussian Splatting via Unified Optimization

    Language 3D Gaussian Splatting enables open-vocabulary scene understanding by associating Gaussian primitives with language features. These semantic attributes increase model storage, while most existing 3DGS compression methods are designed mainly to preserve RGB rendering quality. One straightforward approach is to apply sparsification and quantization sequentially. However, pruning based only on rendering quality can remove Gaussians that have small RGB contributions but are important for semantic prediction. Once removed, their semantic information cannot be recovered by later quantization. We propose ClunoGS, a compact framework for Language 3DGS that combines semantics-aware Gaussian sparsification and vector quantization in a unified constrained objective. Semantic supervision is used during sparsification, and the objective jointly considers rendering quality, semantic accuracy, and compactness. We optimize the unified objective with alternating primal, sparsification, quantization, and dual updates. We evaluate ClunoGS on open-vocabulary localization and segmentation using the LERF and ScanNet-10 benchmarks. Under aggressive compression, our unified, semantics-aware formulation substantially alleviates the degradation of semantic representations caused by sequential compression and semantics-agnostic Gaussian sparsification. Ablations further show that jointly coordinating sparsification and quantization better preserves semantic accuracy than applying them independently.

  24. arXiv · 资产数据集与许可

    Every Alarm Has a Cause: An Open Generator of Telecom Alarm Storms with Exact Root Causes, Bound to NORIA-O, RFC 8632 and TMF642

    Alarm root-cause analysis is studied on data that cannot be shared. Operator logs are confidential, so what the field publishes is a causal graph between alarm types rather than per-incident causes, a fixed scenario set, gated, or never released. Two recent benchmarks carry per-incident root causes at scale; neither states a download location or a licence. We give a generator rather than a dataset. It runs on real backbone topologies from the Internet Topology Zoo and records the exact root cause and the propagation path of every alarm -- a by-product of generation, not a labelling pass. Every emitted field binds to a term in NORIA-O, RFC 8632, RFC 8345 or TMF642, and an unknown term fails the run. We ship fixed configurations, splits, a scorer and nine baselines. A one-line rule -- rank the device that raised the earliest critical alarm -- takes top-1 at 0.650 against 0.575 for our own temporal method, and wins 13 of 15 runs over three backbones and five seeds. An LLM given the same view reaches 0.600, the same model allowed to query the graph 0.625, and the rule wins on real operator alarms too, at 0.773: a property of alarm streams, not of our generator. One fault class separates the table: when a site loses power its PoP reports nothing, so every method that ranks alarmed devices scores 0.000 or near it, while the two that walk the topology to a silent candidate take 1.000 and 0.833.

  25. arXiv · 场景与世界生成

    Looking Ahead: Modeling Future Evolution in Autoregressive Video Generation

    Autoregressive video diffusion models generate videos chunk by chunk, yet producing plausible continuations does not guarantee that they capture how scenes evolve. Under teacher forcing, strong temporal redundancy may allow local visual continuity to serve as a shortcut for current-chunk denoising, reducing the pressure to model visual evolution. We introduce Future Chunk Prediction (FCP), an auxiliary objective jointly optimized with teacher-forced denoising that predicts temporally separated chunks without access to the intervening visual context. Drawing supervision directly from training videos, FCP requires no additional pretrained visual encoder and leaves autoregressive inference unchanged. Our method achieves superior physical plausibility over standard teacher-forced autoregressive training with a lower overall training compute budget, highlighting the effectiveness and efficiency of future-chunk supervision for modeling visual evolution. Intriguingly, our visualizations suggest that FCP fosters a more structured representation space, which may be valuable for building autoregressive world models.

  26. arXiv · 角色动作与重定向

    DewHand: Direct Metric World-Space Hand Motion Reconstruction from Egocentric Videos

    Recovering metric world-space hand motion from egocentric video supports understanding human interactions and transferring them to embodied agents. Three challenges complicate this task: multi-stage pipelines propagate errors from independently estimated hand and camera motion; monocular observations leave physical scale ambiguous; and occlusion and out-of-view hands admit multiple plausible motion trajectories. We present DewHand, a unified model for direct joint reconstruction of world-space hand and camera motion. To avoid compounding errors from multi-stage estimation, DewHand jointly predicts hand and camera motion in a shared world coordinate system within a single-stage pipeline. To resolve metric-scale ambiguity, it combines scene-level geometric features from VGGT-$Ω$ with metric-space supervision, while hand-specific features preserve fine-grained articulation. To handle occlusion and out-of-view hands, we formulate reconstruction as conditional motion generation with flow matching. Under the same architecture, this generative formulation outperforms direct regression in both accuracy and temporal smoothness with only two sampling steps. Experiments on HOT3D and ARCTIC demonstrate improved world-space accuracy and temporal consistency over state-of-the-art methods. DewHand reduces rigid-aligned world-space error relative to StableHand from 32.9 to 28.4 mm on HOT3D and from 76.2 to 36.2 mm on ARCTIC. It also recovers camera scale within 10% on two thirds of HOT3D clips, versus about a quarter for SLAM. Moreover, in zero-shot evaluations on HoloAssist, H2O, and TACO, captured with different devices, DewHand achieves the lowest rigid-aligned world-space error (WA-MPJPE) on each dataset.

10月21日周三
  1. arXiv · 场景与世界生成

    GeoMem-VLA: Internalizing 3D Scene Understanding into Temporal Memory for Vision-Language-Action Models

    Geometric structure is essential for robotic manipulation and understanding the 3D world. Most VLA models rely on 2D representations of current observations, while geometry-aware approaches often require additional 3D information. However, geometry inferred independently from individual observations is often discarded rather than accumulated into a persistent task representation. Our key insight is that action generation needs a compact, task-relevant geometric abstraction that can be continuously updated, rather than a complete, renderable 3D reconstruction at every step. We introduce GeoMem-VLA, a VLA framework that internalizes explicit 3D supervision into dedicated geometric tokens and maintains them as temporal memory for action generation, without directly aligning the VLM's native visual embeddings with the teacher feature space. During training, a frozen geometry foundation model supervises learnable GeoPlan tokens through a geometric flow-matching objective over the current scene point map and future end-effector tracks. These tokens are integrated with the memory of observation representations, where historical information is retrieved, adaptively fused with current observations, and consolidated to retain event frames as the task progresses. At inference, the policy retrieves historical representations and fuses them with the current working representation to predict actions, without invoking the 3D teacher or explicitly reconstructing the scene. After benchmark-specific downstream adaptation, GeoMem-VLA achieves 98.9% average success on LIBERO and 91.10% / 91.24% on RoboTwin's clean/randomized settings. Across five real-world tasks, GeoMem-VLA achieves 78.0% average success. These results suggest that internalizing and remembering 3D scene understanding is a promising approach to robot control.

  2. arXiv · 场景与世界生成

    Spatially Gated Demographic Steering for Debiasing Multi-Person Text-to-Image Generation

    Text-to-Image models reproduce societal stereotypes, and existing mitigation applies one demographic correction to the whole scene. In multi-person scene generation this creates a trade-off: global steering improves aggregate balance only by homogenizing the people within an image. Steering-based methods like Embedding Arithmetic drive the intra-image diversity below the unsteered baseline. Text- and latent-projection approaches shift the aggregate skew with only marginal gains in intra-image diversity. We present Spatially Gated Steering (SGS), which steers the demographics of each person independently by extracting per-person masks from the Diffusion Transformer's (DiT) internal attention maps at a critical denoising timestep and applying a distinct demographic vector within each mask for the remaining steps. SGS needs no fine-tuning, inversion or layout specifications, and uses a lightweight segmentor over frozen DiT attention features to localize subjects at early timesteps. On skewed occupational prompts, SGS reaches high intra-image racial and gender diversity while keeping aggregate balance competitive and image quality near baseline, a result we validate on two architecturally distinct diffusion transformers, FLUX.1-Dev and SD3.5-Large. Codebase and model weights are available at https://github.com/cvims/spatially-gated-steering

  3. arXiv · 可控视频与镜头

    TernScope: Model-Hardware Co-Design for Efficient Ternary Endoscopic Segmentation

    Edge medical segmentation must reconcile small model size, high accuracy, and fast inference. TernScope addresses this tradeoff through model--hardware co-design, learning a compact segmentation network whose low-bit operators also simplify hardware execution. Split-channel large-kernel blocks preserve spatial context with limited computation, while branch fusion removes redundant inference paths. Ternary Progressive Quantization-aware Training (TPQ) then adapts the fused network from deep to shallow layers, through the decoder, and finally to the input and prediction layers. All 25 convolutions use ternary weights and 8-bit activations, enabling packed storage and multiplier-free computation. The resulting model uses 0.2483 M parameters and 0.0949 GMACs while retaining competitive segmentation accuracy on in-domain and external datasets. Under the same 45 nm technology, TernScope reduces compute-array area by 88.46% compared with the original Gemmini, a systolic-array accelerator with 8-bit $\times$ 8-bit multipliers. These results demonstrate a path from accurate lightweight medical segmentation to compact low-bit hardware.

  4. arXiv · 渲染、神经场与导出

    AutoPhyX: Automatic Text-Condition Physics Property Generation

    High-fidelity 4D physical simulation requires accurate physical parameter assignments that align with diverse material properties. Previous methods struggle with critical bottlenecks: they either rely on computationally intensive per-scene optimization or predict parameters from purely visual inputs. The latter suffers from inherent ambiguity, e.g., the inability to distinguish the stiffness of rubber from its appearance alone. To bridge this gap, we propose AutoPhyX, a text-conditioned framework for predicting spatially-varying physical properties. Since training robust feed-forward models requires fine-grained supervision, we first introduce a novel part-controllable data generation pipeline that decomposes complex 3D assets into semantically distinct components and pairs them with physically plausible, diverse parameters and corresponding text descriptions. Leveraging this dataset, AutoPhyX employs a cross-modal modulation mechanism where text features dynamically modulate visual features for precise physical grounding. By formulating property prediction within a voxel field, AutoPhyX ensures compatibility with diverse 3D representations, including meshes, point clouds, Gaussian Splatting, and NeRF. Experiments demonstrate that our method enables physically plausible, text-driven parameter assignment in a single forward pass. It achieves high accuracy and robust generalization to in-the-wild objects, paving the way for future physics-intensive embodied and robotic applications.

  5. arXiv · 可控视频与镜头

    Sparsity Should Be ALLOTted by Output: Training-Free Allocation of Block-Sparse Attention for Video Diffusion Transformers

    Full 3D attention is costly in video Diffusion Transformers. Block-sparse attention reduces this cost by computing only selected attention blocks. However, existing methods either overlook the potential to optimize sparsity allocation or rely on heuristic rules or local metrics to guide it. In this work, we observe that sparsifying attention in different layers and at different denoising steps affects final-video fidelity to substantially different degrees. Yet the identities of the most sensitive layer-step pairs remain nearly unchanged across inputs. Building on these findings, we propose Allocating Layer-step Latency by Output Tests (ALLOT), a training-free sparsity allocation method. Treating each layer-step pair as a cell, ALLOT estimates each cell's contribution to final-video error through (i) cell probing and (ii) trajectory calibration, then jointly optimizes cell densities under a measured-latency budget. It preserves existing selection strategies and kernels, replacing only the allocation component. At equal or shorter wall-clock time, ALLOT outperforms every compared allocation strategy with XAttention, with an 11.8 dB PSNR gain on HunyuanVideo. Extensive experiments across four additional published pipelines on Wan2.1, Wan2.2, and HunyuanVideo show that ALLOT outperforms or ties the default in all nine settings.

  6. arXiv · 场景与世界生成

    Events, Not Pixels: Where World Models Get the Return Wrong

    World models are increasingly used for planning, where a planner ranks candidate actions by returns imagined by the model. However, these models are judged mainly by how well they render frames, even though a visually faithful rollout can still assign the wrong return. To identify where this mismatch arises, we introduce a transition-level diagnostic that replaces imagined transitions with real ones and measures the reduction in return error against a budget-matched random control. Our analysis yields two findings. (1) Event transitions, the few steps where a reward is earned or an episode ends, are rendered as well as other steps (median pixel-error ratio 0.85), yet restoring them removes a median of 70% of the return error. In contrast, restoring transitions with the largest MSE, SSIM or LPIPS error removes less than a random choice. Across the 44 Atari-100k model--collection pairs that pass our reliability checks, including three released families, event patches outperform random patches in 39 (sign test p=1.4 x 10^{-7}), with a median gain of +44 percentage points. (2) The remaining error is structured: a complementary patch separates errors at true events from those elsewhere, while an eight-step timing tolerance recovers few events missed by our deterministic models. This structure also guides repair: a rule recorded before four repair outcomes were known predicts when a diffusion model repairs these failures. Event patches improve plan ranking, and the same concentration appears on control tasks with rare reward events.

  7. arXiv · 场景与世界生成

    LiveAct: Towards Real-Time Multi-Agent World Models with Heterogeneous Action Interfaces

    Interactive video world models predict how an environment evolves under user actions, but most existing systems assume one controlled agent and one task-specific action interface. We study multi-agent video world modeling, where one model generates a joint visual future for multiple controlled agents while interpreting heterogeneous controls ranging from binary game buttons to continuous robot commands. This task presents two complementary challenges. First, every agent-indexed action stream must control its intended visual entity independently. Second, the model must adapt the meaning of each control variable across different tasks, where the same tensor dimension may denote an attack, steering command, or robot command. We introduce LiveAct, centered on a Heterogeneous Action Interface (HAI) whose two components mirror this decomposition. To support independent control, Agent-Aware Spatial Modulation uses language-driven spatial gates and agent-specific attention to bind each action stream to its intended visual agent. To accommodate heterogeneous controls, Schema-Guided Action Adaptation describes each control variable in language and modulates its semantic token with the synchronized numerical value. We further transfer the complete interface to a blockwise causal generator through teacher forcing and rollout-aware distribution matching, enabling efficient streaming generation while retaining agent-specific control and action semantics. We instantiate the formulation on six multi-agent sources spanning games and multi-robot manipulation, and evaluate visual quality, action following, agent attribution, and latency. Across games and robot manipulation, LiveAct reaches 0.898 and 0.942 agent-attribution accuracy without agent locations or identity crops, and the causal generator attains 17.70 FPS at $512\times 448$.

  8. arXiv · 可控视频与镜头

    Region-Aware Multi-View Inspection Using Deep Learning and Domain-Informed Decision Fusion

    Automated optical inspection of precision components with tiny surface defects remains challenging when different regions of the same component exhibit different defect characteristics and, therefore, different acceptance criteria. This paper presents a region-aware, dual-model inspection framework that combines instance segmentation of localized seat-region defects with image-level classification of global shaft-region anomalies. Rather than treating these two model detections as sufficient evidence for component rejection, a deterministic decision fusion layer that formalizes industrial Standard Operating Procedures is applied to the model outputs to make the final component level decision. A PLC-synchronized imaging system acquires 18 seat and 18 shaft images over a full 360-degree rotation to provide multi-view circumferential coverage. The framework was evaluated on 380 automotive fuel injector needles, covering representative operational defect types, using identical model predictions in all configurations. Direct model-output aggregation via simple union logic (where any single-view detection flags a rejection) identified all 104 non-conforming components but produced a component level false rejection rate of 57.6%. Inclusion of a complete decision layer maintained the observed non-conforming (NOK) recall of 100% while reducing the false rejection rate to 10.9%, corresponding to an 81.1% reduction in false-positive component decisions. The results demonstrate that formalized manufacturing acceptance criteria can complement sensitive visual models by translating frame-level detections into practical and traceable component-level decisions.

  9. arXiv · 渲染、神经场与导出

    Progressive Permutation Alignment with Temperature-Controlled Spectral Sharpening

    In implicit neural representations (INRs), 3D signals are encoded directly in network weights, making stable weight-space operations essential. However, permutation symmetry causes independently trained yet functionally equivalent models to appear distant in weight space, often inducing large loss barriers under linear interpolation. Existing neuron alignment methods either commit early to discrete matches and become trapped in inconsistent local optima or rely on computationally expensive gradient optimization over continuous relaxations, both of which are unreliable in challenging cases such as NeRF. We adopt progressive alignment, a layer-wise strategy that follows inter-layer dependencies from input to output. At each layer, a correlation matrix is estimated and aligned with a discrete anchor obtained by the Hungarian algorithm, after which temperature-controlled spectral annealing gradually sharpens the correlation structure toward a permutation matrix. This update is deterministic and lightweight, avoiding the inner-loop gradient optimization required by continuous relaxation methods. A final refinement stage further optimizes the network layer-by-layer via weight matching, starting from the aligned solution. Experiments on MLPs, CNNs, and Neural Radiance Fields (NeRFs) show that the proposed method reliably recovers neuron correspondences and achieves stable alignment across architectures. In practical experiments aligning independently trained NeRF models, our method reduces the loss barriers under weight interpolation compared with existing approaches.

  10. arXiv · 场景与世界生成

    AdvFidelity: Unifying Preference Alignment and Few-Step Image Generation within Reinforcement Learning

    Practical image generation requires both human preference alignment and few-step efficiency, yet combining reinforcement learning (RL) with distillation is often multi-stage and may erode RL gains We observe that few-step RL progressively improves structural coherence, even with specialized rewards, but leaves low-level artifacts that these rewards may fail to penalize. To address this gap, we propose AdvFidelity, which integrates artifact-sensitive fidelity supervision into RL through an online adversarial reward. This enables joint reward alignment and high-fidelity few-step generation within a single RL stage, without distribution matching distillation (DMD) or other explicit distillation objectives. We train a discriminator to rank high-quality references (positives) above content-matched reconstructions (negatives) produced by the current generator. Reconstructing moderately noised references largely preserves style and scene layout while exposing low-level degradations. This encourages the discriminator to assess artifacts rather than content differences. With negatives reconstructed online, alternating updates adapt fidelity supervision to the generator's evolving failure modes. References can come from existing target-domain data or full-step model samples. Built on LeapAlign, AdvFidelity backpropagates gradients from both rewards through the few-step sampling trajectories used at inference. Across two model backbones and three task settings, AdvFidelity matches full-step LeapAlign in alignment, task performance, and overall generation quality and remains highly competitive with other evaluated full-step RL methods, using only 4 or 8 steps without CFG.

  11. arXiv · 渲染、神经场与导出

    AOV-GS: Active Open-Vocabulary Semantic Mapping with 3D Gaussian Splatting

    Robots deployed in human-centered environments need maps that support both geometric navigation and open-vocabulary semantic queries beyond a fixed label set. A key challenge is enriching such maps with semantic information without relying on manually annotated segmentation masks or closed-set semantic categories. Existing geometry-driven active mapping maximizes coverage but treats semantics as a passive by-product of the collected trajectory, whereas active semantic mapping typically depends on supervised semantic segmentation. We present AOV-GS, an active 3D Gaussian mapping system that guides exploration using frozen foundation models. During coarse exploration, AOV-GS re-ranks geometrically promising candidate views according to CLIP mask-embedding novelty computed from SAM instance segmentation, biasing exploration toward semantically informative regions. Frozen SAM and CLIP provide instance masks and language-aligned embeddings throughout mapping, while a lightweight scene-specific language field is fitted offline directly to the collected CLIP embeddings rather than manually annotated masks, enabling open-vocabulary text-query segmentation on novel views. In a controlled comparison using the same SAM-CLIP perception stack, AOV-GS improves average language-field mIoU over geometry-only active planning by 1.3 percentage points on Replica (34.6% on eight scenes) and by 1.8 points on Matterport3D (23.1% on five scenes). With CorrCLIP, AOV-GS reaches 35.6% on Replica, 2.0 points above the matched Geometry-based + CorrCLIP baseline. Novel-view PSNR remains comparable to the geometry-only baseline on both datasets while depth L1 improves on Replica. Code is available at https://github.com/OptimusFaber/AOV-GS.

  12. arXiv · 可控视频与镜头

    NeuralSRender: G-Buffer Guided Video Diffusion for Sim-to-Real Rendering

    The scarcity of high-quality annotated visual data remains a critical bottleneck for autonomous-driving perception. Simulators provide training data at scale, but the limited realism of synthetic videos compromises perception performance in real-world applications. To address this limitation, we propose NeuralSRender, a sim-to-real video translation framework that enhances the photorealism of synthetic videos while maintaining geometric consistency with the source scenes. Built on a pretrained video diffusion transformer (DiT), NeuralSRender uses a lightweight projector to inject geometric and material priors from G-buffers, and LoRA enables parameter-efficient adaptation. The model is trained in two stages: it first learns G-buffer conditioning from synthetic videos with accurate G-buffers, then adapts to real-world appearance using real videos with estimated G-buffers. During the second stage, an additional LoRA module is trained while the projector and first-stage LoRA remain frozen to reduce interference from estimation noise. NeuralSRender introduces only 1.45\% parameter overhead and can be trained on consumer-grade GPUs with 24 GB of memory. To support training and evaluation, we introduce GenReal, a large-scale driving video dataset comprising synthetic and real-world videos with corresponding G-buffers. Experiments demonstrate improved photorealism with competitive geometric consistency. Training perception models on our translated videos instead of raw simulator renders improves detection $\mathrm{AP}_{50}$ on nuImages from 35.1 to 40.6 and segmentation mIoU on Cityscapes from 32.4 to 40.8. Our code, data, and model weights are available at: https://github.com/WHU-USI3DV/NeuralSRender.

  13. arXiv · 创作 Agent 与装配

    ROCOCO: Robust Outline Completion of Occluded Curvilinear Objects

    We propose a learning-based generative model for completing curvilinear features under occlusion by environmental obstacles. To this end, we adopt a flow matching-based approach that supports the inpainting of occluded regions with the appropriate curvilinear features. To eliminate the reliance on privileged information about the occlusion a priori, we also train a UNet-based mask prediction model, which can identify regions for inpainting given only the occluded observation. The components of our approach are validated via experiments using procedurally-generated test cases, and the complete pipeline is validated using large public datasets for curvilinear objects. Against off-the-shelf inpainting baselines, ROCOCO is the strongest method by a wide margin when no occlusion mask is given, and is competitive with mask-supervised inpainting when one is, achieving best on full-image error and second only to LaMa on the occluded region, while using a completer of $7.7$M parameters against LaMa's $51$M and Stable Diffusion's $860$M.

  14. arXiv · 可控视频与镜头

    How Representational Priors Shape Generalization in World Action Models

    World action models (WAMs) are shown to substantially diverge under generalization performance based on their generative backbones. We investigate this gap through the representational geometry that WAMs inherit from their backbones and how policy training reorganizes it. Our central hypothesis is that a useful representational prior would structure geometry that separates nuisance information from the in-distribution manifold in advance of training as a policy. We quantify this with nuisance orthogonality; orthogonality of the feature under visual variation outside the empirical in-distribution manifold. Across various image and video backbones, we show that a higher orthogonality from the representational prior is associated with stronger robustness to visual variations after WAM training, enabling policy generalization measurable prospectively for world models. In addition, analyzing the resulting robust WAMs reveals nuisance suppression; relative suppression of feature displacement caused by such variations and robust action generation of the policy. Together, these findings provide a geometric perspective on how the representational prior of the world model shapes the generalization performance of WAMs.