Pith. sign in

REVIEW 5 major objections 5 minor 34 references

XEWorld: Can Action-Conditioned World Models Generalize to Unseen Robot Embodiments?

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Trained on several robot arms, video world models transfer to an unseen body only insofar as it resembles the training fleet; held-out error tracks appearance distance, not kinematic workspace distance.

desk verdict A well-built cross-embodiment testbed with solid intervention findings, but the headline 'visual similarity governs generalization' claim rests on a five-robot correlation that is statistically fragile. read the letter →

arxiv 2608.05799 v1 pith:I6EHRYTQ submitted 2026-08-06 cs.RO cs.CV

Yixiang Chen , Jiabing Yang , Yuan Xu , Qisen Ma , Keji He , Peiyan Li , Kai Wang , Ziheng He
show 5 more authors
Xiangnan Wu Jing Liu Nianfeng Liu Yan Huang Liang Wang
This is my paper · ORCID
classification cs.ROcs.CV
keywords cross-embodimentgeneralizationaction-conditionedworldmodelsvideoroboticmanipulationvisualpatternmatchingkinematicsimilarityopticalflowactionrepresentationmodelbenchmark
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether action-conditioned world models—video generators trained to predict what a robot's actions will look like—can render a robot body they never saw during training, and it builds a controlled testbed, XEWorld, to force an answer. Five robot arms perform the same 25 tasks in scenes that are byte-identical across bodies, so the body itself is the only variable, and models are scored on held-out bodies across visual, morphological, kinematic, and object-dynamics metrics. The central finding is that these models transfer to an unseen body roughly to the degree that it looks like the training fleet, not to the degree that it moves like it: held-out error correlates with appearance distance (r=0.812) but weakly and unstably with forward-kinematics workspace distance (r=0.549, with a leave-one-out range that crosses zero). The paper concludes that current world models act as 2D visual pattern matchers that never bind abstract actions to physical bodies, and that true cross-embodiment simulation requires architectures that separate visual appearance from physical dynamics. This matters because learned simulators are proposed as engines for planning and policy evaluation; a simulator that only copies appearance will produce visually plausible but physically untrustworthy rollouts for any robot that looks new.

What carries the argument

The load-bearing instrument is XEWorld's paired-data protocol: for every task and seed, the scene layout, object poses, lighting, and camera are byte-identical across the five robot embodiments, so a performance drop on a held-out body is attributable to the embodiment alone. On that substrate the paper stacks a decoupled metric suite that scores visual quality, robot morphology (mask IoU and region LPIPS via SAM2 segmentation), robot kinematics (URDF-derived forward-kinematics keypoints and normalized DTW tracked by CoTracker3), and object dynamics; and two embodiment distances—the kinematic distance as a Chamfer distance over forward-kinematics reachable workspaces, and the appearance distance as a cosine distance over HSV-histogram plus Hu-moment descriptors of standardized renders. The reference world model, adapted from FlowWAM on a Wan2.2 video-diffusion backbone, separates three input streams (scene RGB, robot-masked optical flow as the action, and an embodiment-specification stream), which lets the paper vary the amount of target-embodiment information along an intervention ladder from an empty description to a per-frame forward-kinematics render, the central comparison that isolates spatial-temporal alignment as the binding constraint.

What would settle it

Render a held-out body whose head-camera appearance is nearly indistinguishable from a training robot (same colors, texture, silhouette) but whose joints and reachable workspace are substantially different: the visual-pattern-matcher thesis predicts near-training-quality rendering despite the kinematic gap, so a large error jump would falsify it. A cheaper calculation: recompute the five-point correlation with a learned perceptual distance between reference renders instead of the hand-crafted descriptor; the headline r=0.812 must survive that swap to be trustworthy.

Watch

Extended reading notes

Core claim

Current action-conditioned world models, evaluated strictly out of distribution, do not learn reusable physical dynamics. Asked to predict a rollout for a robot body outside the training fleet, they generate at a quality that tracks how visually different the new robot looks—measured by a hand-crafted appearance descriptor—rather than how differently it moves, measured by the Chamfer distance between forward-kinematics workspaces. The paper documents the behavioral fingerprints of this pattern-matching: numeric joint angles produce incoherent or dissolving bodies on unseen robots while pixel-space action signals (optical flow, masks, ray maps) largely rescue the rendering; static pictures or multi-view descriptions of the new robot barely help, whereas a perfectly time-aligned per-frame render of the target body nearly closes the gap; and fine-tuning on a handful of target demonstrations recovers appearance while driving catastrophic forgetting of the robots the model already knew. Because the same degradation appears across four different world-model architectures, the authors argue it is a shared architectural bottleneck rather than a defect of one model, and they conclude that cross-embodiment generalization requires innovations that decouple visual appearance from underlying physical dynamics.

Load-bearing premise

The headline finding rests on a five-point leave-one-out correlation: a hand-crafted appearance descriptor (a 16x16 HSV histogram plus seven Hu-moment silhouette features over nine renders) is assumed to capture the visual similarity the video model actually uses, and the paper itself reports the appearance correlation at p=0.075 with the kinematics correlation's leave-one-out range crossing zero.

Editorial extensions

If this is right

  • Numeric joint actions are a transfer bottleneck: replacing them with robot-masked optical flow cuts held-out LPIPS error (a perceptual image-similarity score) by 29% on Franka and 57% on Piper, so pixel-space action representations are a prerequisite for rendering unseen bodies.
  • Static descriptions of an unseen robot saturate almost immediately: one reference image helps modestly, nine views and an articulation clip add less than 2% in shape IoU, while a per-frame registered render cuts robot-region LPIPS by 50% on Franka and 45% on Piper—alignment, not information, is what the model lacks.
  • Few-shot adaptation is a localized patch: 25 demonstrations close most of the global LPIPS gap but only about half of the robot-region gap on Franka, and adapting one robot raises LPIPS error on a previously seen robot by 69%.
  • The bottleneck is architectural, not model-specific: all four world models evaluated (IRASim, Ctrl-World, EnerVerse-AC, and the reference model) degrade on held-out bodies, with Franka LPIPS error rising 38% to 221% over each model's seen-robot baseline.
  • The design rule the paper extracts is that future cross-embodiment simulators should keep actions in pixel space, provide time-aligned structural conditioning, and separate visual appearance from physical dynamics.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The five-point correlation is the load-bearing evidence, so extending the leave-one-out protocol to a larger fleet—ten or more bodies—is the cheapest decisive check, and the testbed's paired-data format ports directly to it; at p=0.075 the appearance correlation needs more points to firm up or dissolve.
  • A texture-swap experiment would turn correlation into causation: render a held-out body with the head-camera appearance of a training robot. Near-training-quality output would confirm the 2D-pattern-matcher diagnosis mechanistically; a large error would mean the descriptor is missing what the network actually uses.
  • The per-frame-render result sets an upper bound on the fix: since supplying the forward-kinematics render at every timestep nearly closes the gap, an architecture that internally predicts such a geometric render and then textures it with a separate appearance generator is a concrete, testable instantiation of the decoupling the paper calls for, scoreable with the same metrics.
  • A downstream risk the paper leaves implicit: a world model used as a learned simulator for a novel embodiment can produce visually plausible but physically wrong rollouts, and appearance-only benchmarks (LPIPS, SSIM, PSNR) would certify them while the kinematics dimension of XEWorld catches the failure; benchmarks for learned simulators should therefore include a kinematics-versus-appearance contr
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. The paper introduces XEWorld, a controlled cross-embodiment testbed for action-conditioned world models. It provides byte-aligned paired scenes across five bimanual robot embodiments, held-out evaluation protocols, and a decoupled metric suite covering visual quality, robot morphology, robot kinematics, and object dynamics. Using a FlowWAM-based reference model and three external world models (IRASim, Ctrl-World, EnerVerse-AC), the paper reports that pixel-space action inputs outperform numeric joint poses, static target-robot descriptions saturate quickly, per-frame aligned renders restore morphology, and few-shot adaptation improves target performance while causing forgetting on previously seen robots. A five-fold leave-one-embodiment-out correlation analysis is used to argue that generalization difficulty is governed by visual appearance similarity rather than kinematic workspace similarity.

Significance. The experimental design is a real strength: strictly paired scenes isolate the embodiment variable, the metric suite separates appearance from kinematics, and the appendix tables report all ten metrics across all conditions. The intervention results (optical flow versus joint poses, saturation of static descriptions, benefit of per-frame renders, few-shot forgetting) are internally consistent and potentially reproducible. If the headline conclusion were established, the paper would be an important contribution because it would redirect architecture design toward decoupling visual appearance from physical dynamics. However, the central 'visual similarity governs generalization' claim currently rests on a five-robot correlational analysis whose statistical fragility and unvalidated appearance descriptor are not reflected in the abstract. The benchmark and its diagnostic findings are valuable; the main text needs to match the caution of its own Appendix E.

major comments (5)
  1. [Section 5.4, Table 14, Appendix E] The headline conclusion that cross-embodiment generalization is 'governed by visual similarity rather than physical kinematic similarity' is stronger than the reported statistics support. The appearance-distance correlation is r=0.812 with an exact two-sided permutation p=0.075 and n=5, which is above the conventional threshold; the kinematic correlation is r=0.549 with p=0.392 and leave-one-robot-out range [−0.155, 0.706]. The abstract and Section 5.4 present the contrast as 'strong/stable' versus 'weak/unstable' without these qualifications. I agree with the authors' own Appendix E that the appearance correlation is sign-stable under LOO, but a five-point correlation cannot support the causal 'governed by' wording. Please soften the headline to an association, report p-values and LOO intervals in the main text, and consider a Steiger-style test for the difference between the two dependent correlations or a model comparison on the five folds.
  2. [Section 3.3 and Appendix C.5] The appearance axis is a hand-crafted descriptor with no validation against the similarity structure actually used by a video diffusion model. The descriptor (16x16 HSV histogram plus seven Hu moments over nine views) may not reflect texture, context, or learned feature similarity, so the r=0.812 correlation could be measuring the wrong axis. Please validate the descriptor against model-based perceptual distances (e.g., LPIPS or DINO features) and show that the headline correlation is robust across descriptor choices. Additionally, the five robots vary in DoF, color, and scale; Table 14 shows DoF correlations above 0.9 for several morphology metrics, so the appearance-versus-kinematic contrast may be confounded with DoF and other covariates. A partial-correlation or matched-pair analysis is needed before the visual-governance claim can be considered established.
  3. [Tables 2-4 and Figure 2] All intervention values are episode means without standard errors or significance tests. The paired design (same task-seed-robot cells across conditions) permits paired tests, and several conclusions hinge on small differences. For example, Table 3 shows that moving from a reference image to nine views changes full-frame LPIPS by less than 1-2%, and the paper interprets this as saturation; without variance, it is impossible to tell whether the saturation is a real effect or noise. Please add error bars or paired statistical tests, or state explicitly how many seeds underlie each mean and whether the reported differences were significant.
  4. [Section 5.5 and Table 5] The claim that the generalization gap is a 'shared limitation' or 'universal' across current architectures is based on three external models, each evaluated with a single fine-tuned checkpoint and no variance estimates. The four columns in Table 5 show the same qualitative direction, which is useful, but the wording 'universal' and 'current world model architectures' exceeds the evidence. Please either add error bars and statistical comparisons across seeds, or qualify the claim to the specific evaluated models. The same caution applies to the abstract's statement that 'current models act primarily as 2D visual pattern matchers.'
  5. [Section 5.2 and abstract] The abstract says that successfully rendering an unseen embodiment zero-shot 'strictly requires' heavily grounded cues, but the experiments show large improvements from per-frame renders, not a proof of strict necessity. Optical flow alone already gives shape IoU of 0.667 on Franka and 0.773 on Piper. The wording should be softened to something like 'is substantially improved by' unless the authors define and demonstrate a success threshold that no less-grounded cue meets.
minor comments (5)
  1. [Figure 2] The label 'Raymap' is used in Figure 2 while the text and Table 2 use 'Ray map'; please unify the terminology.
  2. [Section 3.2 and Appendix C.3] PCK at alpha=0.1 and normalized DTW are defined only in the appendix; a brief main-text definition or a notation pointer would help readers of the main results.
  3. [Appendix C.2] The use of a ground-truth mask to initialize SAM2 in a predicted video could be seen as leaking oracle information into the morphology metrics; please justify this choice explicitly or provide a sensitivity check with random initialization.
  4. [General] The paper says the dataset is released but gives no URL or download instructions; if this is a benchmark contribution, a public release link and model checkpoint should be provided.
  5. [Table 5] The seen-fleet baseline is reported as a single pooled number; since the paper emphasizes isolating per-robot effects, a per-robot breakdown of the seen baselines would aid interpretation.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the benchmark conclusions are drawn from independent models and external distance metrics, not from the reference model's construction.

full rationale

The paper is an empirical benchmark study rather than a mathematical derivation. Its central claim, that cross-embodiment generalization is governed by visual similarity rather than physical kinematic similarity, is supported by a leave-one-embodiment-out correlation between an independently defined appearance descriptor (HSV histogram plus Hu moments, Appendix C.5) and measured LPIPS errors, and by the universal degradation of three independently sourced world models (IRASim, Ctrl-World, EnerVerse-AC). The reference model (FlowWAM) is the authors' own, but it is described in full and used as one of several architectures; no load-bearing conclusion rests on an unverified self-citation or a uniqueness theorem. The per-frame render and pixel-space action interventions are explicitly labeled as controlled conditions, not as free predictions, so they do not reduce to their own inputs. The fragility of the five-point correlation (p=0.075; kinematic LOO range crossing zero) is a statistical-support concern, not a circularity concern. I therefore find no circular step.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The paper introduces no new physical entities or fitted constants. Its load-bearing assumptions are about the validity of the simulation, the tracking and segmentation tools, and the embodiment-distance descriptors.

assumptions (3)
  • domain assumption Byte-identical scene rendering across embodiments preserves all physical dynamics except the robot body; the only varying factor is the embodiment.
    Section 3.1 claims byte-level alignment of scene layout, object poses, lighting, and camera across five robots. The inference that any performance drop is entirely due to the embodiment depends on the RoboTwin simulator's physics being independent of the robot body beyond the commanded actions.
  • domain assumption SAM2 and CoTracker3 provide sufficiently accurate masks and correspondences on generated videos for morphology and kinematics metrics.
    Section 3.2 and Appendix C.2-C.3 use SAM2 to segment predicted robot masks and CoTracker3 to track URDF-derived keypoints. Errors in these tools on synthetic, possibly distorted videos would propagate into the reported metrics.
  • ad hoc to paper The appearance distance descriptor (HSV histogram plus Hu moments) captures the visual similarity that determines video-generation difficulty.
    Appendix C.5 describes this hand-crafted descriptor. The paper does not validate that it correlates with the model's internal representation of visual appearance; it is chosen by the authors for this analysis and is load-bearing for the distance correlation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of XEWorld: Can Action-Conditioned World Models Generalize to Unseen Robot Embodiments?." pith.science (2026). https://pith.science/paper/I6EHRYTQ

@misc{pith2026260805799,
  author       = {Pith},
  title        = {Pith review of: XEWorld: Can Action-Conditioned World Models Generalize to Unseen Robot Embodiments?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/I6EHRYTQ}},
  note         = {Machine review of arXiv:2608.05799}
}
read the original abstract

Action-conditioned world models are promising learned simulators for robotic manipulation, yet evaluating them exclusively on training robots fails to reveal whether they capture physical dynamics or merely memorize visual patterns. To answer whether a model can faithfully render a robot it has never seen, we introduce XEWorld, a controlled cross-embodiment testbed for world models that isolates embodiments by evaluating held-out robots within physically identical scenes. Our systematic analysis uncovers a shared architectural bottleneck: current models act primarily as 2D visual pattern matchers whose generalization is governed by visual similarity rather than physical kinematic similarity. Driven by this limitation, they struggle to translate abstract numeric joint actions into coherent visual trajectories, and fail to predict dynamic visual changes from static initial observations. Consequently, successfully rendering an unseen embodiment zero-shot strictly requires heavily grounded cues, specifically pixel-space actions and explicit spatial-temporal alignment. Even when bypassing this zero-shot barrier via few-shot adaptation, the forced appearance recovery triggers catastrophic forgetting of seen embodiments. Together, these failures expose a critical inability to apply learned physical dynamics to novel visual appearances, highlighting that achieving true cross-embodiment generalization requires architectural innovations that decouple visual appearance from underlying physical dynamics.

Figures

Figures reproduced from arXiv: 2608.05799 by the authors.

Figure 1
Figure 1. The XEWorld cross-embodiment testbed. We introduce XEWorld to investigate whether action-conditioned world models genuinely capture generalizable physical dynamics, or merely memorize embodiment-specific visual patterns. dynamic action inputs. Consequently, successfully rendering an unseen embodiment zero-shot requires explicit spatial￾temporal alignment, such as perfectly registered per-frame renderings. Alternativ… view at source ↗
Figure 2
Figure 2. Impact of targeted interventions on unseen-embodiment generation. Bars report relative improvements over the baseline for the two held-out robots across different intervention strategies. tion from its visual appearance. We adapt FlowWAM (Chen et al. 2026a), an action-conditioned embodied world model, and fine-tune it on our three training robots. We design the model to process three distinct input streams: (1) scen… view at source ↗
Figure 4
Figure 4. Adaptation benefit and forgetting. Fine-tuning on the held-out robot reduces its LPIPS error, but increases the LPIPS error on the previously seen UR5 robot. tion acts as a localized patch rather than a fundamental so￾lution. The network overwrites its internal representation of the old robots to memorize the visual patterns of the new one. This trade-off suggests that deploying these models in diverse environments … view at source ↗
Figures from the paper (5 more)
Figure 5
Figure 5. Figure 5: Correlation between embodiment distance and generalization error. understands physical kinematics. Therefore, a robot with a radically different mechanical structure is not harder for the model to generate unless it also lacks visual familiarity. 5.5 Does the Generaliz…
Figure 6
Figure 6. Figure 6: Action-interface transfer across additional episodes. Each task contributes paired Franka and Piper rows. Columns hold the task, seed, target robot, and rollout time fixed while varying only the action representation. Dense pixel-space controls more consistently retain…
Figure 7
Figure 7. Figure 7: Additional action-interface rollouts. Paired robot rows again isolate the action representation while holding scene state and rollout time fixed [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Robot conditioning across additional episodes. Each task again contributes paired Franka and Piper rows. Reference images, views, and articulation clips provide task-independent identity cues [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Additional robot-conditioning rollouts. The per-frame condition most directly restores target identity because its cue is aligned with the current rollout configuration [PITH_FULL_IMAGE:figures/full_fig_p020_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 9 canonical work pages

  1. [1]

    Proceedings of the computer vision and pattern recognition conference , pages=

    Robotwin: Dual-arm robot benchmark with generative digital twins , author=. Proceedings of the computer vision and pattern recognition conference , pages=

  2. [2]

    arXiv preprint arXiv:2506.18088 , year=

    Robotwin 2.0: A scalable data generator and benchmark with strong domain randomization for robust bimanual robotic manipulation , author=. arXiv preprint arXiv:2506.18088 , year=

    arXiv

  3. [3]

    arXiv preprint arXiv:2503.20314 , year=

    Wan: Open and advanced large-scale video generative models , author=. arXiv preprint arXiv:2503.20314 , year=

    arXiv

  4. [4]

    FlowWAM: Optical Flow as a Unified Action Representation for World Action Models

    FlowWAM: Optical Flow as a Unified Action Representation for World Action Models , author=. arXiv preprint arXiv:2607.13017 , year=

    work page Pith review arXiv

  5. [5]

    arXiv preprint arXiv:2406.14540 , year=

    Irasim: A fine-grained world model for robot manipulation , author=. arXiv preprint arXiv:2406.14540 , year=

    arXiv

  6. [6]

    arXiv preprint arXiv:2505.09723 , year=

    Enerverse-ac: Envisioning embodied environments with action condition , author=. arXiv preprint arXiv:2505.09723 , year=

    arXiv

  7. [7]

    The Fourteenth International Conference on Learning Representations , year=

    Ctrl-World: A Controllable Generative World Model for Robot Manipulation , author=. The Fourteenth International Conference on Learning Representations , year=

    work page

  8. [8]

    arXiv preprint arXiv:2511.19861 , year=

    Gigaworld-0: World models as data engine to empower embodied ai , author=. arXiv preprint arXiv:2511.19861 , year=

    arXiv

Show all 34 references
  1. [9]

    arXiv preprint arXiv:2501.03575 , year=

    Cosmos world foundation model platform for physical ai , author=. arXiv preprint arXiv:2501.03575 , year=

    arXiv

  2. [10]

    Conference on Robot Learning , pages=

    DreamGen: Unlocking Generalization in Robot Learning through Video World Models , author=. Conference on Robot Learning , pages=. 2025 , organization=

    2025

  3. [11]

    The Twelfth International Conference on Learning Representations , year=

    Learning Interactive Real-World Simulators , author=. The Twelfth International Conference on Learning Representations , year=

  4. [12]

    Advances in Neural Information Processing Systems , volume=

    ivideogpt: Interactive videogpts are scalable world models , author=. Advances in Neural Information Processing Systems , volume=

  5. [13]

    Forty-first International Conference on Machine Learning , year=

    Genie: Generative interactive environments , author=. Forty-first International Conference on Machine Learning , year=

  6. [14]

    arXiv preprint arXiv:2606.17030 , year=

    Qwen-robotworld technical report: Unifying embodied world modeling through language-conditioned video generation , author=. arXiv preprint arXiv:2606.17030 , year=

    arXiv

  7. [15]

    Proceedings of the IEEE/CVF international conference on computer vision , pages=

    Scalable diffusion models with transformers , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  8. [16]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Vbench: Comprehensive benchmark suite for video generative models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  9. [17]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    Worldscore: A unified evaluation benchmark for world generation , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  10. [18]

    2024 IEEE International Conference on Robotics and Automation (ICRA) , pages=

    Open x-embodiment: Robotic learning datasets and rt-x models , author=. 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages=. 2024 , organization=

    2024

  11. [19]

    arXiv preprint arXiv:2505.14986 , year=

    Anybody: A benchmark suite for cross-embodiment manipulation , author=. arXiv preprint arXiv:2505.14986 , year=

    arXiv

  12. [20]

    arXiv preprint arXiv:2601.12993 , year=

    Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization , author=. arXiv preprint arXiv:2601.12993 , year=

  13. [21]

    arXiv preprint arXiv:2512.13100 , year=

    OXE-AugE: A Large-Scale Robot Augmentation of OXE for Scaling Cross-Embodiment Policy Learning , author=. arXiv preprint arXiv:2512.13100 , year=

  14. [22]

    Conference on Robot Learning , pages=

    Towards Embodiment Scaling Laws in Robot Locomotion , author=. Conference on Robot Learning , pages=. 2025 , organization=

    2025

  15. [23]

    International Conference on Learning Representations , volume=

    Sam 2: Segment anything in images and videos , author=. International Conference on Learning Representations , volume=

  16. [24]

    arXiv preprint arXiv:2410.11831 , year=

    Cotracker3: Simpler and better point tracking by pseudo-labelling real videos , author=. arXiv preprint arXiv:2410.11831 , year=

    arXiv

  17. [25]

    Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

    The unreasonable effectiveness of deep features as a perceptual metric , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  18. [26]

    European conference on computer vision , pages=

    Raft: Recurrent all-pairs field transforms for optical flow , author=. European conference on computer vision , pages=. 2020 , organization=

    2020

  19. [27]

    arXiv preprint arXiv:2602.03793 , year=

    BridgeV2W: Bridging Video Generation Models to Embodied World Models via Embodiment Masks , author=. arXiv preprint arXiv:2602.03793 , year=

  20. [28]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

    Kinema4d: Kinematic 4d world modeling for spatiotemporal embodied simulation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  21. [29]

    arXiv preprint arXiv:2607.19343 , year=

    Masked Visual Actions for Unified World Modeling , author=. arXiv preprint arXiv:2607.19343 , year=

    arXiv

  22. [30]

    arXiv preprint arXiv:2602.08971 , year=

    Worldarena: A unified benchmark for evaluating perception and functional utility of embodied world models , author=. arXiv preprint arXiv:2602.08971 , year=

  23. [31]

    arXiv preprint arXiv:2606.02800 , year=

    Cosmos 3: Omnimodal world models for physical ai , author=. arXiv preprint arXiv:2606.02800 , year=

    arXiv

  24. [32]

    arXiv preprint arXiv:2509.22578 , year=

    Egodemogen: Novel egocentric demonstration generation enables viewpoint-robust manipulation , author=. arXiv preprint arXiv:2509.22578 , year=

  25. [33]

    arXiv preprint arXiv:2605.17912 , year=

    WorldArena 2.0: Extending Embodied World Model Benchmarking on Modality, Functionality and Platform , author=. arXiv preprint arXiv:2605.17912 , year=

    arXiv

  26. [34]

    arXiv preprint arXiv:2410.05363 , year=

    Towards world simulator: Crafting physical commonsense-based benchmark for video generation , author=. arXiv preprint arXiv:2410.05363 , year=

    arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.