pith. sign in

arxiv: 2606.28804 · v1 · pith:I2DMFDTQnew · submitted 2026-06-27 · 💻 cs.CV · cs.RO

ViPSim: Collaborating Visual and Parameter Spaces for Consistent Long-Horizon Embodied World Models

Longyu Chen , Heng Li , Wei Yang , Manqi Zhao , Dongsheng Jiang This is my paper

Pith reviewed 2026-06-30 09:28 UTC · model grok-4.3

classification 💻 cs.CV cs.RO
keywords embodied world modelsvisual parameter spacestrajectory consistencylong-horizon simulationrobot-object interactionsvision-language-actionembodied agents
0
0 comments X

The pith

Unifying visual priors with action parameters removes trajectory drift in long-horizon robot simulations.

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

Embodied world models suffer from a gap between low-dimensional actions and high-dimensional video outputs that produces accumulating drift and mismatched robot-object contacts over long rollouts. The paper addresses this by defining a Visual Space of pixel-aligned spatial priors (end-effector poses, camera views, depth geometry, morphological masks) and a Parameter Space of numerical drivers (raw action sequences and camera matrices). These two domains are unified so that every generated state is both geometrically anchored and numerically steered. If successful, the resulting simulators become reliable enough to serve as safe, repeatable benchmarks for vision-language-action systems.

Core claim

ViPSim achieves consistent long-horizon generation through the synergistic collaboration of Visual and Parameter Spaces. The Visual Space supplies dense structural grounding via pixel-aligned projections of end-effector pose, camera perspectives, depth-informed scene geometry, and robotic morphological masks. The Parameter Space supplies precise motion guidance via raw action sequences and camera matrices. By unifying the two, generated states remain simultaneously anchored by geometric boundaries and steered by numerical commands.

What carries the argument

Unification of Visual Space (explicit spatial priors) and Parameter Space (numerical drivers) so states are anchored by geometry and steered by actions.

If this is right

  • Trajectory consistency improves markedly across long-horizon rollouts.
  • Complex deformable interactions such as cloth folding emerge without explicit supervision.
  • Performance holds in out-of-distribution and cross-embodiment settings.
  • The framework remains agnostic to the choice of video-generation backbone.
  • The resulting models supply a high-fidelity base for automated evaluation and predictive control of embodied agents.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Better simulators could reduce the number of real-world trials needed to validate vision-language-action policies.
  • The same dual-space structure might extend to other embodied tasks that require both visual fidelity and precise control signals.
  • If the geometric correspondence holds, downstream tasks such as sim-to-real transfer for deformable manipulation become more tractable.

Load-bearing premise

Integrating pixel projections of poses, depths, masks, and camera views with raw action sequences and matrices supplies enough geometric correspondence to stop accumulated drift and inconsistent interactions.

What would settle it

A controlled long-horizon rollout experiment in which ViPSim still produces measurable trajectory drift or visibly inconsistent robot-object contacts despite the visual-parameter unification.

Figures

Figures reproduced from arXiv: 2606.28804 by Dongsheng Jiang, Heng Li, Longyu Chen, Manqi Zhao, Wei Yang.

Figure 1
Figure 1. Figure 1: Qualitative results on diverse tasks. The first sequence [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Chunk-based Autoregressive Generation Pipeline. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: ViPSim bridges the representation gap between numerical commands and scene dynamics through a dual-space [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Qualitative comparison of long-horizon temporal stability. We evaluate ViPSim against the baseline on tasks requiring [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison of deformable and rigid object manipulation between ViPSim and baseline methods. The first [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Analysis of dynamic compatibility, robustness, and generalization limits. (a) Unseen task generalization. (b) High [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Evaluation of action compliance through swapping. The [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗
read the original abstract

Embodied World Models (EWMs) have emerged as a scalable and risk-free paradigm for advancing embodied intelligence, enabling the safety-critical evaluation of Vision-Language-Action systems. However, their reliability as evaluation benchmarks and foundational simulators is often hindered by the representation gap between low-dimensional actions and high-dimensional video synthesis. This gap results in a lack of geometric correspondence, manifesting as accumulated trajectory drift and inconsistent robot-object interactions during long-horizon rollouts. To bridge this gap, we propose ViPSim, a framework that achieves consistent long-horizon generation through the synergistic collaboration of Visual and Parameter Spaces. We define the Visual Space as a domain of explicit spatial priors, integrating pixel-aligned projections of end-effector pose, camera perspectives, depth-informed scene geometry, and robotic morphological masks to provide dense structural grounding. Concurrently, the Parameter Space serves as a domain of numerical drivers, injecting raw action sequences and camera matrices to provide precise motion guidance. By unifying these two spaces, ViPSim ensures that the generated states are simultaneously anchored by geometric boundaries and steered by numerical commands. Extensive experiments demonstrate that ViPSim is backbone-agnostic and significantly enhances trajectory consistency. Notably, our approach exhibits emergent capabilities in generating complex interactions with deformable objects (e.g., cloth folding) and maintains robust performance in out-of-distribution and cross-embodiment scenarios, providing a high-fidelity foundation for the automated evaluation and predictive control of embodied agents.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes ViPSim, a framework for embodied world models that unifies a Visual Space (defined via pixel-aligned projections of end-effector pose, camera perspectives, depth-informed scene geometry, and robotic morphological masks) with a Parameter Space (raw action sequences and camera matrices) to enforce geometric correspondence. This is claimed to eliminate accumulated trajectory drift and inconsistent robot-object interactions in long-horizon rollouts, with the method being backbone-agnostic, yielding significant consistency gains, emergent capabilities (e.g., deformable object interactions like cloth folding), and robust out-of-distribution/cross-embodiment performance.

Significance. If the experimental results hold, the work addresses a practical bottleneck in embodied simulators for Vision-Language-Action evaluation by supplying explicit spatial priors alongside numerical drivers. The backbone-agnostic design and reported OOD robustness would be useful strengths for downstream predictive control and automated benchmarking.

major comments (2)
  1. [Abstract] Abstract: the central claim that unifying the two spaces 'ensures' elimination of trajectory drift rests on the sufficiency of the listed visual priors plus parameter inputs; no derivation or explicit integration mechanism (e.g., how masks and depth are fused with action sequences inside the generative model) is supplied in the abstract, which is load-bearing for the geometric-correspondence argument.
  2. [Abstract] The manuscript asserts 'extensive experiments' with 'significant enhancement' and 'robust' OOD performance, yet the abstract supplies neither quantitative metrics, baselines, error bars, nor ablation controls; without these, the strength of the consistency and emergent-capability claims cannot be assessed.
minor comments (2)
  1. [Abstract] The terms 'Visual Space' and 'Parameter Space' are introduced as novel domains but lack a concise formal definition or notation that would allow readers to distinguish them from standard conditioning inputs in video-generation models.
  2. [Abstract] The phrase 'backbone-agnostic' is used without specifying which backbones were tested or what invariance property is being claimed.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and the recommendation for minor revision. We address the comments on the abstract below and will incorporate targeted revisions to improve clarity and precision.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim that unifying the two spaces 'ensures' elimination of trajectory drift rests on the sufficiency of the listed visual priors plus parameter inputs; no derivation or explicit integration mechanism (e.g., how masks and depth are fused with action sequences inside the generative model) is supplied in the abstract, which is load-bearing for the geometric-correspondence argument.

    Authors: The abstract is a high-level summary; the explicit fusion mechanism (pixel-aligned visual priors combined with parameter inputs via the backbone-agnostic architecture) and geometric correspondence details are provided in Section 3, including the model diagram and equations for space unification. We agree the wording 'ensures' can be strengthened for precision in the abstract and will revise it to 'achieves through synergistic collaboration' while adding a brief clause on the integration of visual and parameter spaces. revision: yes

  2. Referee: [Abstract] The manuscript asserts 'extensive experiments' with 'significant enhancement' and 'robust' OOD performance, yet the abstract supplies neither quantitative metrics, baselines, error bars, nor ablation controls; without these, the strength of the consistency and emergent-capability claims cannot be assessed.

    Authors: We note that abstracts in this domain typically emphasize conceptual contributions over specific numbers due to space constraints, with all quantitative results (metrics, baselines, error bars, ablations) reported in Sections 4–5. However, to address the concern, we will revise the abstract to include representative quantitative highlights (e.g., consistency gains and OOD robustness figures) if length permits. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The manuscript presents ViPSim as an architectural framework proposal that unifies two defined spaces (Visual Space via explicit spatial priors and Parameter Space via numerical drivers) to address trajectory drift. No equations, derivations, fitted parameters, or first-principles predictions appear in the provided text; claims of consistency and emergent capabilities are asserted via experiments rather than reducing to self-definition or self-citation chains. The derivation chain is therefore self-contained as a descriptive hypothesis without internal reduction to inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 2 invented entities

The abstract introduces two new conceptual domains (Visual Space and Parameter Space) as the core contribution without independent evidence of their necessity or sufficiency beyond the stated experiments. No numerical free parameters are specified.

axioms (1)
  • domain assumption A representation gap exists between low-dimensional actions and high-dimensional video synthesis that causes accumulated trajectory drift and inconsistent interactions.
    Stated directly in the opening of the abstract as the motivating problem.
invented entities (2)
  • Visual Space no independent evidence
    purpose: Domain of explicit spatial priors providing dense structural grounding via pixel-aligned projections, camera perspectives, depth, and morphological masks.
    Defined in the abstract as one of the two collaborating spaces; no independent evidence supplied.
  • Parameter Space no independent evidence
    purpose: Domain of numerical drivers supplying raw action sequences and camera matrices for precise motion guidance.
    Defined in the abstract as the second collaborating space; no independent evidence supplied.

pith-pipeline@v0.9.1-grok · 5802 in / 1315 out tokens · 53659 ms · 2026-06-30T09:28:50.011854+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 5 canonical work pages · 4 internal anchors

  1. [1]

    Physical Intelligence, Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, et al.π 0.5: a vision-language-action model with open-world generalization, 2025

    2025

  2. [2]

    Depthvla: Enhancing vision-language-action models with depth-aware spatial reasoning, 2025

    Tianyuan Yuan, Yicheng Liu, Chenhao Lu, Zhuoguang Chen, Tao Jiang, and Hang Zhao. Depthvla: Enhancing vision-language-action models with depth-aware spatial reasoning, 2025

    2025

  3. [3]

    Evo-0: Vision-language-action model with implicit spatial understanding, 2025

    Tao Lin, Gen Li, Yilei Zhong, Yanwen Zou, Yuxin Du, Jiting Liu, Encheng Gu, and Bo Zhao. Evo-0: Vision-language-action model with implicit spatial understanding, 2025

    2025

  4. [4]

    Gemini Robotics: Bringing AI into the Physical World

    Gemini Robotics Team, S Abeyruwan, J Ainslie, JB Alayrac, MG Are- nas, T Armstrong, A Balakrishna, R Baruch, M Bauza, M Blokzijl, et al. Gemini robotics: Bringing ai into the physical world, 2025.URL https://arxiv. org/abs/2503.20020, 1:6, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Univla: Learning to act anywhere with task-centric latent actions, 2025

    Qingwen Bu, Yanting Yang, Jisong Cai, Shenyuan Gao, Guanghui Ren, Maoqing Yao, Ping Luo, and Hongyang Li. Univla: Learning to act anywhere with task-centric latent actions, 2025

    2025

  6. [6]

    Gigabrain-0: A world model-powered vision-language-action model.arXiv preprint arXiv:2510.19430, 2025

    GigaBrain Team, Angen Ye, Boyuan Wang, Chaojun Ni, Guan Huang, Guosheng Zhao, Haoyun Li, Jie Li, Jiagang Zhu, Lv Feng, et al. Gigabrain-0: A world model-powered vision-language-action model. arXiv preprint arXiv:2510.19430, 2025

    work page arXiv 2025

  7. [7]

    Vid2world: Crafting video diffusion models to interactive world models, 2025

    Siqiao Huang, Jialong Wu, Qixing Zhou, Shangchen Miao, and Ming- sheng Long. Vid2world: Crafting video diffusion models to interactive world models, 2025

    2025

  8. [8]

    Worldgym: World model as an environ- ment for policy evaluation, 2025

    Julian Quevedo, Ansh Kumar Sharma, Yixiang Sun, Varad Suryavanshi, Percy Liang, and Sherry Yang. Worldgym: World model as an environ- ment for policy evaluation, 2025

    2025

  9. [9]

    Worldeval: World model as real-world robot policies evaluator, 2025

    Yaxuan Li, Yichen Zhu, Junjie Wen, Chaomin Shen, and Yi Xu. Worldeval: World model as real-world robot policies evaluator, 2025

    2025

  10. [10]

    Towards high-consistency embodied world model with multi- view trajectory videos, 2025

    Taiyi Su, Jian Zhu, Yaxuan Li, Chong Ma, Zitai Huang, Hanli Wang, and Yi Xu. Towards high-consistency embodied world model with multi- view trajectory videos, 2025

    2025

  11. [11]

    Enact: Evaluating embodied cognition with world modeling of egocentric interaction, 2025

    Qineng Wang, Wenlong Huang, Yu Zhou, Hang Yin, Tianwei Bao, Jianwen Lyu, Weiyu Liu, Ruohan Zhang, Jiajun Wu, Li Fei-Fei, and Manling Li. Enact: Evaluating embodied cognition with world modeling of egocentric interaction, 2025

    2025

  12. [12]

    Gen2act: Human video generation in novel scenarios enables generalizable robot manipulation, 2024

    Homanga Bharadhwaj, Debidatta Dwibedi, Abhinav Gupta, Shubham Tulsiani, Carl Doersch, Ted Xiao, Dhruv Shah, Fei Xia, Dorsa Sadigh, and Sean Kirmani. Gen2act: Human video generation in novel scenarios enables generalizable robot manipulation, 2024

    2024

  13. [13]

    Enerverse: Envisioning embodied future space for robotics manipulation, 2025

    Siyuan Huang, Liliang Chen, Pengfei Zhou, Shengcong Chen, Zhengkai Jiang, Yue Hu, Yue Liao, Peng Gao, Hongsheng Li, Maoqing Yao, and Guanghui Ren. Enerverse: Envisioning embodied future space for robotics manipulation, 2025

    2025

  14. [14]

    Genie envi- sioner: A unified world foundation platform for robotic manipulation, 2025

    Yue Liao, Pengfei Zhou, Siyuan Huang, Donglin Yang, Shengcong Chen, Yuxin Jiang, Yue Hu, Jingbin Cai, Si Liu, Jianlan Luo, Liliang Chen, Shuicheng Yan, Maoqing Yao, and Guanghui Ren. Genie envi- sioner: A unified world foundation platform for robotic manipulation, 2025

    2025

  15. [15]

    Cosmos World Foundation Model Platform for Physical AI

    Niket Agarwal, Arslan Ali, Maciej Bala, Yogesh Balaji, Erik Barker, Tiffany Cai, Prithvijit Chattopadhyay, Yongxin Chen, Yin Cui, Yifan Ding, et al. Cosmos world foundation model platform for physical ai. arXiv preprint arXiv:2501.03575, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  16. [16]

    Ctrl- world: A controllable generative world model for robot manipulation, 2025

    Yanjiang Guo, Lucy Xiaoyang Shi, Jianyu Chen, and Chelsea Finn. Ctrl- world: A controllable generative world model for robot manipulation, 2025

    2025

  17. [17]

    Learning real- world action-video dynamics with heterogeneous masked autoregression, 2025

    Lirui Wang, Kevin Zhao, Chaoqi Liu, and Xinlei Chen. Learning real- world action-video dynamics with heterogeneous masked autoregression, 2025

    2025

  18. [18]

    Unified world models: Coupling video and action diffusion for pretraining on large robotic datasets, 2025

    Chuning Zhu, Raymond Yu, Siyuan Feng, Benjamin Burchfiel, Paarth Shah, and Abhishek Gupta. Unified world models: Coupling video and action diffusion for pretraining on large robotic datasets, 2025

    2025

  19. [19]

    Unifolm-wma-0: A world-model-action (wma) framework under unifolm family, 2025

    Unitree. Unifolm-wma-0: A world-model-action (wma) framework under unifolm family, 2025

    2025

  20. [20]

    Wristworld: Generating wrist-views via 4d world models for robotic manipulation, 2025

    Zezhong Qian, Xiaowei Chi, Yuming Li, Shizun Wang, Zhiyuan Qin, Xiaozhu Ju, Sirui Han, and Shanghang Zhang. Wristworld: Generating wrist-views via 4d world models for robotic manipulation, 2025

    2025

  21. [21]

    Magicworld: Interactive geometry- driven video world exploration, 2025

    Guangyuan Li, Siming Zheng, Shuolin Xu, Jinwei Chen, Bo Li, Xiaobin Hu, Lei Zhao, and Peng-Tao Jiang. Magicworld: Interactive geometry- driven video world exploration, 2025

    2025

  22. [22]

    Roboscape: Physics-informed embodied world model, 2025

    Yu Shang, Xin Zhang, Yinzhou Tang, Lei Jin, Chen Gao, Wei Wu, and Yong Li. Roboscape: Physics-informed embodied world model, 2025

    2025

  23. [23]

    Learning primitive embodied world models: Towards scalable robotic learning, 2025

    Qiao Sun, Liujia Yang, Wei Tang, Wei Huang, Kaixin Xu, Yongchao Chen, Mingyu Liu, Jiange Yang, Haoyi Zhu, Yating Wang, Tong He, Yilun Chen, Xili Dai, Nanyang Ye, and Qinying Gu. Learning primitive embodied world models: Towards scalable robotic learning, 2025

    2025

  24. [24]

    Tesseract: Learning 4d embodied world models, 2025

    Haoyu Zhen, Qiao Sun, Hongxin Zhang, Junyan Li, Siyuan Zhou, Yilun Du, and Chuang Gan. Tesseract: Learning 4d embodied world models, 2025

    2025

  25. [25]

    Orv: 4d occupancy-centric robot video generation, 2025

    Xiuyu Yang, Bohan Li, Shaocong Xu, Nan Wang, Chongjie Ye, Zhaoxi Chen, Minghan Qin, Yikang Ding, Zheng Zhu, Xin Jin, Hang Zhao, and Hao Zhao. Orv: 4d occupancy-centric robot video generation, 2025

    2025

  26. [26]

    Tracegen: World modeling in 3d trace space enables learning from cross-embodiment videos, 2025

    Seungjae Lee, Yoonkyo Jung, Inkook Chun, Yao-Chih Lee, Zikui Cai, Hongjia Huang, Aayush Talreja, Tan Dat Dao, Yongyuan Liang, Jia-Bin Huang, and Furong Huang. Tracegen: World modeling in 3d trace space enables learning from cross-embodiment videos, 2025

    2025

  27. [27]

    Dynam- icrafter: Animating open-domain images with video diffusion priors, 2023

    Jinbo Xing, Menghan Xia, Yong Zhang, Haoxin Chen, Wangbo Yu, Hanyuan Liu, Xintao Wang, Tien-Tsin Wong, and Ying Shan. Dynam- icrafter: Animating open-domain images with video diffusion priors, 2023

    2023

  28. [28]

    Wan: Open and Advanced Large-Scale Video Generative Models

    Team Wan, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, Haiming Zhao, Jianxiao Yang, et al. Wan: Open and advanced large-scale video generative models.arXiv preprint arXiv:2503.20314, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  29. [29]

    Genie: Generative interactive environments, 2024

    Jake Bruce, Michael Dennis, Ashley Edwards, Jack Parker-Holder, Yuge Shi, Edward Hughes, Matthew Lai, Aditi Mavalankar, Richie Steigerwald, Chris Apps, Yusuf Aytar, Sarah Bechtle, Feryal Behbahani, Stephanie Chan, Nicolas Heess, Lucy Gonzalez, Simon Osindero, Sherjil Ozair, Scott Reed, Jingwei Zhang, Konrad Zolna, Jeff Clune, Nando de Freitas, Satinder Si...

    2024

  30. [30]

    ivideogpt: Interactive videogpts are scalable world models, 2024

    Jialong Wu, Shaofeng Yin, Ningya Feng, Xu He, Dong Li, Jianye Hao, and Mingsheng Long. ivideogpt: Interactive videogpts are scalable world models, 2024

    2024

  31. [31]

    Irasim: A fine-grained world model for robot manipulation, 2025

    Fangqi Zhu, Hongtao Wu, Song Guo, Yuxiao Liu, Chilam Cheang, and Tao Kong. Irasim: A fine-grained world model for robot manipulation, 2025

    2025

  32. [32]

    Enerverse-ac: Envisioning embodied environ- ments with action condition, 2025

    Yuxin Jiang, Shengcong Chen, Siyuan Huang, Liliang Chen, Pengfei Zhou, Yue Liao, Xindong He, Chiming Liu, Hongsheng Li, Maoqing Yao, and Guanghui Ren. Enerverse-ac: Envisioning embodied environ- ments with action condition, 2025

    2025

  33. [33]

    Video depth anything: Consistent depth estimation for super-long videos, 2025

    Sili Chen, Hengkai Guo, Shengnan Zhu, Feihu Zhang, Zilong Huang, Jiashi Feng, and Bingyi Kang. Video depth anything: Consistent depth estimation for super-long videos, 2025

    2025

  34. [34]

    Sam 2: Segment anything in images and videos, 2024

    Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu, Ronghang Hu, Chaitanya Ryali, Tengyu Ma, Haitham Khedr, Roman R¨adle, Chloe Rolland, Laura Gustafson, Eric Mintun, Junting Pan, Kalyan Vasudev Alwala, Nicolas Carion, Chao-Yuan Wu, Ross Girshick, Piotr Doll ´ar, and Christoph Feichtenhofer. Sam 2: Segment anything in images and videos, 2024

    2024

  35. [35]

    Motionctrl: A unified and flexible motion controller for video generation, 2024

    Zhouxia Wang, Ziyang Yuan, Xintao Wang, Tianshui Chen, Menghan Xia, Ping Luo, and Ying Shan. Motionctrl: A unified and flexible motion controller for video generation, 2024

    2024

  36. [36]

    AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems

    Qingwen Bu, Jisong Cai, Li Chen, Xiuqi Cui, Yan Ding, Siyuan Feng, Shenyuan Gao, Xindong He, Xuan Hu, Xu Huang, et al. Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems.arXiv preprint arXiv:2503.06669, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  37. [37]

    Ewmbench: Evaluating scene, motion, and semantic quality in embodied world models, 2025

    Hu Yue, Siyuan Huang, Yue Liao, Shengcong Chen, Pengfei Zhou, Liliang Chen, Maoqing Yao, and Guanghui Ren. Ewmbench: Evaluating scene, motion, and semantic quality in embodied world models, 2025

    2025

  38. [38]

    Yolo-world: Real-time open-vocabulary object detection, 2024

    Tianheng Cheng, Lin Song, Yixiao Ge, Wenyu Liu, Xinggang Wang, and Ying Shan. Yolo-world: Real-time open-vocabulary object detection, 2024

    2024