Pith. sign in

REVIEW 5 major objections 6 minor 226 references

Distilling Physical Priors into Streaming World Models

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims that temporal credit routing—applying window-level physical rewards to temporally aligned denoising actions—is what makes online RL improve the physical plausibility of few-step causal video generators.

desk verdict A solid extensional result: the PhyS-120K dataset and the TCR credit-routing tweak are real contributions, but the reward-proxy validation is the load-bearing weak spot. read the letter →

arxiv 2608.07981 v1 pith:Q6DSFI2I submitted 2026-08-08 cs.CV

Liangliang Zhao , Junying Wang , Danni Yang , Yifan Chang , Bin Fu , Yu Qiao , Bowen Zhou , Yihao Liu This is my paper · ORCID
classification cs.CV
keywords worldmodelsvideogenerationphysicalplausibilitytemporalcreditassignmentreinforcementlearningdistributionmatchingdistillationphysics-awaredatasetstreaming
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 claims that the physical plausibility of streaming video world models can be substantially improved by a three-stage training pipeline. First, a new dataset of 120,804 real-world physical-interaction videos, annotated with object properties and causal event sequences, is used to fine-tune a 14B bidirectional diffusion transformer. That teacher is distilled into a 1.3B few-step causal diffusion transformer, and the student is then aligned by online reinforcement learning. The decisive optimization idea is Temporal Credit Routing: instead of giving every denoising step the same video-level reward, the reward is computed on overlapping temporal windows and routed to the denoising actions that produced those frames. If the results are right, this raises PhysicsIQ scores by 18.2% over the teacher and by 23.7%, 14.8%, and 31.4% over the three forcing-based distillation baselines, with additional gains on VideoPhy, VideoPhy2, and PhyGenBench.

What carries the argument

The mechanism that carries the argument is Temporal Credit Routing (TCR), a three-step procedure applied to each group of rollouts. Windows $W_j$ slide over the generated frames; each window is scored by a composite reward $R_{i,j} = R^{\mathrm{VJ}}_{i,j} + \lambda_{\mathrm{flow}} R^{\mathrm{Flow}}_{i,j}$, combining V-JEPA2 representation agreement with a RAFT optical-flow term; rewards are normalized per window across the group; and the resulting advantages are summed with overlap weights $O_{m,j} = |B_m \cap W_j| / \sum_k |B_m \cap W_k|$ to give each causal block a block-level advantage. This block advantage is then used in the Flow-GRPO policy objective with clipping and KL penalty. The paper's argument is that this localization—overlapping windows rather than one global score—is what lets the policy update correct local physical errors without being diluted by easy or hard temporal regions.

What would settle it

Run the full TCR pipeline with the V-JEPA2 reward replaced by a fixed random score or an inverted score and measure PhysicsIQ; if the model still improves, the gains are not caused by the physical-reward signal. Complement this by measuring V-JEPA2's pairwise agreement with human ratings on held-out PhysicsIQ prompts; a drop toward chance would indicate the training-distribution agreement does not transfer.

Watch

Extended reading notes

Core claim

The paper's central claim is that the main bottleneck in physically plausible streaming generation is not the generator architecture but temporal credit assignment in the alignment stage. It argues that a globally broadcast reward cannot tell a causal model which of its blocks produced a physical violation, so it treats the physical-consistency score as a window-level signal: a frozen V-JEPA2 predictor compares predicted and actual future-frame representations within overlapping windows, supplemented by a RAFT optical-flow term, and the resulting group-relative advantages are routed to temporally aligned denoising transitions. The authors claim this produces more physically coherent and causally consistent long-horizon rollouts than video-level reward broadcasting (Flow-GRPO) across PhysicsIQ, VideoPhy, VideoPhy2, and PhyGenBench, and that the effect holds for all three forcing-based causal distillation baselines.

Load-bearing premise

The load-bearing premise is that the frozen V-JEPA2-plus-RAFT reward actually measures physical plausibility; the paper supports this with a 0.65 pairwise agreement with human ratings on 1,000 training prompts, and if that agreement does not hold on the benchmark distribution, the reported RL gains would not reflect genuine physics understanding.

Editorial extensions

If this is right

  • If the central claim is correct, physics-aware supervised fine-tuning on real-world interaction videos is enough to improve a bidirectional teacher on all four physics benchmarks, so data curation can inject physical priors without changing the architecture.
  • TCR improves each forcing-based causal baseline by 8.8% to 17.5% on PhysicsIQ over its Flow-GRPO counterpart, which means the routing, not just the reward, drives the gain.
  • The alignment gains appear without increasing inference cost: the 1.3B causal student still uses 4 to 5 network function evaluations.
  • Long-horizon rollouts at 15 and 20 seconds improve on temporal quality, dynamic degree, visual stability, and VideoPhy physical commonsense, suggesting the benefit does not fade as the rollout extends.

Reading between the lines

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

  • A testable extension is to swap in other window-level scorers, such as physics simulators or human-preference models, inside the same TCR loop; the framework is agnostic to the scorer, so the comparison would isolate whether V-JEPA2's representation agreement is the best proxy for physical plausibility.
  • The reported 0.65 pairwise agreement with human ratings is measured on training prompts, so a natural next experiment is to measure reward-model agreement on the benchmark distributions themselves; a drop toward chance would suggest part of the PhysicsIQ gain is optimizing the reward rather than genuine physical understanding.
  • The uneven PhyGenBench category results point to a dataset-coverage bottleneck: expanding PhyS-120K with optics and material-interaction clips could directly close those gaps, a prediction the paper's own regression discussion supports but does not test.
  • Because TCR only assumes blockwise causal generation and window-level rewards, it could transfer to next-frame autoregressive video models and interactive world models beyond diffusion transformers, though the paper does not demonstrate this.
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 / 6 minor

Summary. The paper proposes PhyS, a three-stage framework for injecting physical priors into streaming video world models. Stage 1 constructs PhyS-120K, a dataset of 120,804 real-world physical-interaction videos with structured annotations from Qwen3-VL, and performs physics-aware SFT on a Wan2.1-14B bidirectional DiT teacher. Stage 2 distills the teacher into a 1.3B causal DiT using DMD under Self, Rolling, or Causal Forcing. Stage 3 applies online RL with a composite V-JEPA2 + RAFT reward and Temporal Credit Routing (TCR), which computes per-window group-normalized advantages and routes them to temporally aligned denoising actions. The paper reports that PhyS improves PhysicsIQ by 18.2% over the teacher and by 23.7%, 14.8%, and 31.4% over the three forcing baselines, with additional gains on VideoPhy, VideoPhy2, and PhyGenBench, and on long-horizon rollouts.

Significance. If the reported results are robust, the paper makes a valuable contribution: it introduces a large real-world physics-interaction dataset, a concrete three-stage pipeline for transferring physical priors from a bidirectional teacher to a lightweight causal generator, and a temporally localized credit assignment method that is well motivated and supported by ablations showing an advantage over global reward broadcasting. The experimental design is generally sound in that the final evaluation metrics are external benchmarks rather than the training reward, so the headline gains are not circular by construction. However, the empirical validation has several gaps that currently prevent full confidence in the central claims: missing uncertainty estimates, no quantitative comparison to prior physics-aware methods, and insufficient evidence that the Stage 3 reward is a valid proxy for physical plausibility on the evaluation distributions.

major comments (5)
  1. [Tables 1 and 2] No error bars, confidence intervals, or multiple-seed results are reported for any metric in Tables 1 and 2. Several comparisons involve small absolute differences (e.g., Self Forcing PhysicsIQ 17.4 vs 17.9 with SFT teacher; VideoPhy2 SA 23.6 vs 23.9), which cannot be distinguished from training noise without variance estimates. Please report means and standard deviations over at least three independent runs (or equivalent) for the main comparisons, or justify why single runs are reliable.
  2. [Section 4.3 and Fig. 7(a)] The only validation of the reward model is a 0.65 pairwise ranking agreement with human preferences on 1,000 training prompts, with no chance baseline, confidence interval, or evaluation on benchmark distributions. Because the headline RL gains are optimized against this reward, the paper should report reward-human agreement on held-out benchmark prompts (e.g., from PhysicsIQ, VideoPhy, VideoPhy2, PhyGenBench) and analyze the correlation between the reward and the final evaluation metrics. Without this, the Stage 3 improvements could reflect optimization of a proxy that does not track physical plausibility on the evaluation distribution.
  3. [Section 5.2] The main results compare only against forcing baselines and the teacher model; no quantitative comparison is made to existing physics-aware video generation methods such as WISA, VideoREPA, PhysMaster, or ProPhy, which are discussed in the Related Work. To support the claim that PhyS is a state-of-the-art physics-aware streaming world model, at least one shared-benchmark quantitative comparison with a prior physics-aware method is needed; the qualitative examples in Fig. 5 are not sufficient.
  4. [Table 1, PhyGenBench rows] The category-level PhyGenBench results show substantial regressions: Rolling Forcing + TCR Material drops from 0.305 to 0.197, and Optics regresses for both Self Forcing (0.451 to 0.402) and Rolling Forcing (0.493 to 0.440). The text attributes this to limited category coverage in PhyS-120K, but this explanation is speculative and the Material regression is large. Please analyze whether the reward or routing induces category-specific failure modes and report per-category results with variance.
  5. [Section 5.4 and Table 4] Key hyperparameters of the composite reward and routing—lambda_flow=0.4, window length and stride, group size G=8, and KL coefficient beta=0.015—are not ablated. Since the reported gains depend on these choices, a sensitivity analysis or at least a discussion of their impact would strengthen the claim that TCR is robust rather than tuned to the specific configuration.
minor comments (6)
  1. [Abstract] The percentages such as 18.2%, 23.7%, 14.8%, and 31.4% are relative improvements, but the abstract does not state this explicitly; please clarify to avoid ambiguity.
  2. [Section 4.3, Eq. (9)] The expectation in Eq. (9) is not explicitly defined; please specify that it is taken over prompts, the group of G rollouts, and the stochastic denoising transitions.
  3. [Fig. 7(a)] Please include a chance-level baseline (0.5) and confidence intervals for the pairwise agreement values in Fig. 7(a), and state how many human raters and prompts were used.
  4. [Table 1] The Wan2.1-1.3B row at 40 NFE has a dash for PhysicsIQ; please explain why this model is not evaluated on that benchmark.
  5. [Section 3] The text mentions 'random human inspection' of the VLM-generated annotations but gives no numbers; please report how many clips were audited and the resulting agreement rate.
  6. [Section 5.1] The sentence 'its window length matches the causal block size, with a half-window stride' leaves the actual frame counts unspecified; please give the concrete number of frames per block and window stride.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; all headline claims are evaluated on external benchmarks rather than on the training reward or fitted values.

full rationale

PhyS's derivation chain is self-contained with respect to circularity. Stage 1 fine-tunes a bidirectional Wan2.1-14B teacher on the externally constructed PhyS-120K dataset, and the reported 18.2% PhysicsIQ improvement is measured on the external PhysicsIQ benchmark, not on PhyS-120K. Stage 2 distills the teacher into a 1.3B causal student using DMD, and the student's gains over Self, Rolling, and Causal Forcing are again reported on external benchmarks (PhysicsIQ, VideoPhy, VideoPhy2, PhyGenBench). Stage 3 optimizes a composite V-JEPA2-plus-RAFT window reward with Temporal Credit Routing; the paper does not present benchmark gains as reward values, but instead reports external VBench-Long and physics-benchmark metrics in Tables 1, 2, and 4. The only in-distribution reward validation is Fig. 7a, where V-JEPA2 is selected over VideoScore2, VideoPhy2, and PhyDetEx based on 0.65 pairwise agreement with human raters on 1,000 training prompts; this is a reward-model selection step, not a fitted parameter later renamed as a prediction. No load-bearing step invokes a self-citation or a uniqueness theorem from the authors' prior work. The skeptical concern about in-distribution reward validation is an external-validity or robustness issue, not a circular reduction, because the central empirical claims are adjudicated by held-out benchmarks. The paper's own noted regressions on PhyGenBench Optics and Material are data-coverage limitations, not evidence of circularity.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

No new physical entities, forces, or particles are introduced; TCR and PhyS-120K are methods and data, not invented entities. The free parameters are training and reward-design choices that the central results depend on.

free parameters (4)
  • lambda_flow = 0.4
    Weight balancing V-JEPA2 consistency reward and RAFT optical-flow reward in the composite reward of Eq. (3); chosen by hand and no sensitivity ablation is reported.
  • beta = 0.015
    KL regularization coefficient in the policy objective of Eq. (9); chosen for training stability without reported sensitivity analysis.
  • rollout_group_size G = 8
    Number of rollouts used for group-relative advantage normalization in Eq. (4); fixed in Section 5.1 and not ablated.
  • window length and stride = window length equals block size, half-window stride
    Temporal window geometry in Section 4.3; chosen without ablation across other window sizes.
assumptions (3)
  • domain assumption Real-world videos from public platforms and open-source datasets, filtered and annotated by Qwen3-VL with random human audit, provide accurate physical-process supervision.
    Stage 1 SFT relies entirely on PhyS-120K annotations for teaching physical priors; annotation errors would propagate to teacher and student.
  • domain assumption V-JEPA2 representation consistency between predicted and encoded future frames is a valid reward proxy for physical plausibility.
    Stage 3 RL reward uses V-JEPA2; validity is supported only by a 0.65 pairwise agreement with human ratings on 1,000 prompts (Fig. 7a), which is moderate.
  • domain assumption DMD distribution matching between noised student-rollout and teacher distributions preserves enough physical priors for the causal student.
    Distillation objective Eq. (2) assumes the teacher distribution contains the physical priors; the paper itself notes regressions that indicate imperfect transfer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distilling Physical Priors into Streaming World Models." pith.science (2026). https://pith.science/paper/Q6DSFI2I

@misc{pith2026260807981,
  author       = {Pith},
  title        = {Pith review of: Distilling Physical Priors into Streaming World Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q6DSFI2I}},
  note         = {Machine review of arXiv:2608.07981}
}
read the original abstract

Streaming world models predict future visual states online while maintaining physically coherent dynamics over long horizons. However, their rollouts often violate basic physical constraints. A common approach distills pretrained bidirectional DiTs into few-step causal generators. However, this paradigm suffers from two fundamental limitations: generic bidirectional teachers acquire limited physical priors from visually oriented pretraining, and the limited priors suffer further loss during bidirectional-to-causal distillation. We present PhyS, a three-stage framework for distilling physical priors into streaming world models. To acquire physical priors from real-world interactions, we construct PhyS-120K, a dataset of 120K real-world physical-interaction videos spanning rigid-body dynamics, soft-body deformation, fluid phenomena, and phase transitions. Each video is annotated with structured descriptions of object properties and causal state transitions. Physics-aware supervised fine-tuning injects the physical priors into a bidirectional 14B DiT teacher, which we then distill into a lightweight 1.3B causal DiT for few-step autoregressive streaming generation. Finally, we use online reinforcement learning to incentivize the distilled model to generate physically plausible rollouts and further propose Temporal Credit Routing (TCR) to address temporal credit assignment. TCR evaluates physical consistency over overlapping temporal windows and routes the resulting group-relative advantages to temporally aligned denoising actions. On PhysicsIQ, PhyS improves the Wan2.1-14B teacher by 18.2\% and the Self Forcing, Rolling Forcing, and Causal Forcing by 23.7\%, 14.8\%, and 31.4\%, respectively. Results also improve the physics-aware video benchmarks VideoPhy, VideoPhy2, and PhyGenBench. The dataset, code, and more sample videos are available on our Project Page.

Figures

Figures reproduced from arXiv: 2608.07981 by the authors.

Figure 1
Figure 1. Representative gains of PhyS on three physics-aware [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the PhyS-120K construction pipeline and dataset. (a) Raw clips from YouTube, Pexels, and WISA are [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Representative scenarios from PhyS-120K. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Overview of PhyS. Stages 1 and 2 transfer physical priors from a bidirectional teacher to a few-step causal generator. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Stage-1 qualitative comparison [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Long-horizon causal rollout comparison. Compared with forcing-based baselines, PhyS better preserves physical [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Reward-model agreement with human pairwise [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

226 extracted references · 13 canonical work pages

  1. [1]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Efficient diffusion models: A comprehensive survey from principles to practices , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  2. [2]

    Advances in Neural Information Processing Systems , volume=

    Diffusion forcing: Next-token prediction meets full-sequence diffusion , author=. Advances in Neural Information Processing Systems , volume=

  3. [3]

    2506.08009 , archivePrefix=

    Self forcing: Bridging the train-test gap in autoregressive video diffusion , author=. 2506.08009 , archivePrefix=

    arXiv

  4. [4]

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

    From slow bidirectional to fast autoregressive video diffusion models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  5. [5]

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

    One-step diffusion with distribution matching distillation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  6. [6]

    Yan, Wilson and Zhang, Yunzhi and Abbeel, Pieter and Srinivas, Aravind , eprint=

  7. [7]

    Proceedings of the 41st International Conference on Machine Learning , year=

    Kondratyuk, Dan and Yu, Lijun and Gu, Xiuye and Lezama, Jos. Proceedings of the 41st International Conference on Machine Learning , year=

  8. [8]

    Advances in Neural Information Processing Systems , year=

    Chen, Boyuan and Mons. Advances in Neural Information Processing Systems , year=

Show all 226 references
  1. [9]

    Hu, Jinyi and Hu, Shengding and Song, Yuxuan and Huang, Yufei and Wang, Mingxuan and Zhou, Hao and Liu, Zhiyuan and Ma, Wei-Ying and Sun, Maosong , eprint=

  2. [10]

    International Conference on Learning Representations , year=

    Pyramidal Flow Matching for Efficient Video Generative Modeling , author=. International Conference on Learning Representations , year=

  3. [11]

    Gao, Kaifeng and Shi, Jiaxin and Zhang, Hanwang and Wang, Chunping and Xiao, Jun and Chen, Long , eprint=

  4. [12]

    2505.23884 , archivePrefix=

    Test-Time Training Done Right , author=. 2505.23884 , archivePrefix=

    arXiv

  5. [13]

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

    From slow bidirectional to fast autoregressive video diffusion models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

  6. [14]

    Huang, Xun and Li, Zhengqi and He, Guande and Zhou, Mingyuan and Shechtman, Eli , eprint=

  7. [15]

    Teng, Hansi and others , eprint=

  8. [16]

    Proceedings of the 42nd International Conference on Machine Learning , series=

    History-Guided Video Diffusion , author=. Proceedings of the 42nd International Conference on Machine Learning , series=

  9. [17]

    2503.19325 , archivePrefix=

    Long-Context Autoregressive Video Modeling with Next-Frame Prediction , author=. 2503.19325 , archivePrefix=

    arXiv

  10. [18]

    Proceedings of the 41st International Conference on Machine Learning , pages=

    Rolling diffusion models , author=. Proceedings of the 41st International Conference on Machine Learning , pages=

  11. [19]

    Chen, Guibin and Lin, Dixuan and Yang, Jiangping and Lin, Chunze and Liu, Junchen and Fan, Mingyuan and Zhang, Hao and Chen, Sheng and Chen, Zheng and Ma, Chengcheng and others , eprint=

  12. [20]

    Liu, Kunhao and Hu, Wenbo and Xu, Jiale and Shan, Ying and Lu, Shijian , eprint=

  13. [21]

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

    One-step diffusion with distribution matching distillation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  14. [22]

    2504.08388 , archivePrefix=

    Mineworld: a real-time and open-source interactive world model on minecraft , author=. 2504.08388 , archivePrefix=

    arXiv

  15. [23]

    2511.09057 , archivePrefix=

    Pan: A world model for general, interactable, and long-horizon world simulation , author=. 2511.09057 , archivePrefix=

  16. [24]

    CV](2025) , author=

    Yan: Foundational Interactive Video Generation, arXiv [cs. CV](2025) , author=

    2025

  17. [25]

    2512.08931 , archivePrefix=

    Astra: General Interactive World Model with Autoregressive Denoising , author=. 2512.08931 , archivePrefix=

  18. [26]

    Zhu, Hongzhou and Zhao, Min and He, Guande and Su, Hang and Li, Chongxuan and Zhu, Jun , booktitle=

  19. [27]

    2509.22622 , archivePrefix=

    Longlive: Real-time interactive long video generation , author=. 2509.22622 , archivePrefix=

    arXiv

  20. [28]

    2507.17744 , archivePrefix=

    Yume: An interactive world generation model , author=. 2507.17744 , archivePrefix=

    arXiv

  21. [29]

    2512.22096 , archivePrefix=

    Yume-1.5: A Text-Controlled Interactive World Generation Model , author=. 2512.22096 , archivePrefix=

  22. [30]

    2601.20540 , archivePrefix=

    Advancing Open-source World Models , author=. 2601.20540 , archivePrefix=

    arXiv

  23. [31]

    Bruce, Jake and Dennis, Michael D and Edwards, Ashley and Parker-Holder, Jack and Shi, Yuge and Hughes, Edward and Lai, Matthew and Mavalankar, Aditi and Steigerwald, Richie and Apps, Chris and others , booktitle=

  24. [32]

    URL: https://deepmind

    Genie 2: A large-scale foundation world model , author=. URL: https://deepmind. google/discover/blog/genie-2-a-large-scale-foundation-world-model , volume=

  25. [33]

    2506.09985 , archivePrefix=

    Assran, Mido and Bardes, Adrien and Fan, David and Garrido, Quentin and Howes, Russell and Komeili, Mojtaba and others , year=. 2506.09985 , archivePrefix=

    arXiv

  26. [34]

    2406.15252 , archivePrefix=

    VideoScore: Building Automatic Metrics to Simulate Fine-Grained Human Feedback for Video Generation , author=. 2406.15252 , archivePrefix=

    arXiv

  27. [35]

    He, Xuan and Jiang, Dongfu and Nie, Ping and Liu, Minghao and Jiang, Zhengxuan and Su, Mingyi and Ma, Wentao and Lin, Junru and Ye, Chun and Lu, Yi and Wu, Keming and Schneider, Benjamin and Do, Quy Duc and Li, Zhuofeng and Jia, Yiming and Zhang, Yuxuan and Cheng, Guo and Wang...

  28. [36]

    Wang, Zeqing and Wang, Keze and Zhang, Lei , eprint=

  29. [37]

    2406.09246 , archivePrefix=

    Openvla: An open-source vision-language-action model , author=. 2406.09246 , archivePrefix=

    arXiv

  30. [38]

    2406.16862 , archivePrefix=

    Dreamitate: Real-world visuomotor policy learning via video generation , author=. 2406.16862 , archivePrefix=

    arXiv

  31. [39]

    2505.12705 , archivePrefix=

    Dreamgen: Unlocking generalization in robot learning through video world models , author=. 2505.12705 , archivePrefix=

    arXiv

  32. [40]

    2512.15692 , archivePrefix=

    mimic-video: Video-action models for generalizable robot control beyond vlas , author=. 2512.15692 , archivePrefix=

    arXiv

  33. [41]

    2602.15922 , archivePrefix=

    World action models are zero-shot policies , author=. 2602.15922 , archivePrefix=

    arXiv

  34. [42]

    2024 , howpublished=

    Video generation models as world simulators , author=. 2024 , howpublished=

    2024

  35. [43]

    Polyak, Adam and Zohar, Amit and Brown, Andrew and Tjandra, Andros and Sinha, Animesh and Lee, Ann and Vyas, Apoorv and Shi, Bowen and Ma, Chih-Yao and Chuang, Ching-Yao and others , eprint=

  36. [44]

    Ma, Guoqing and Huang, Haoyang and Yan, Kun and Chen, Liangyu and Duan, Nan and Yin, Shengming and Wan, Changyi and Ming, Ranchen and Song, Xiaoniu and Chen, Xing and others , eprint=

  37. [45]

    2501.03575 , archivePrefix=

    Cosmos world foundation model platform for physical ai , author=. 2501.03575 , archivePrefix=

    arXiv

  38. [46]

    1967 , publisher=

    The nature of explanation , author=. 1967 , publisher=

    1967

  39. [47]

    Advances in neural information processing systems , volume=

    Recurrent world models facilitate policy evolution , author=. Advances in neural information processing systems , volume=

  40. [48]

    ACM Sigart Bulletin , volume=

    Dyna, an integrated architecture for learning, planning, and reacting , author=. ACM Sigart Bulletin , volume=. 1991 , publisher=

    1991

  41. [49]

    2509.20328 , archivePrefix=

    Video models are zero-shot learners and reasoners , author=. 2509.20328 , archivePrefix=

    arXiv

  42. [50]

    2025 , howpublished =

    Genie 3: A New Frontier for World Models , author =. 2025 , howpublished =

    2025

  43. [51]

    2511.19861 , archivePrefix=

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

  44. [52]

    2507.00990 , archivePrefix=

    Robotic Manipulation by Imitating Generated Videos Without Physical Demonstrations , author=. 2507.00990 , archivePrefix=

    arXiv

  45. [53]

    2601.16163 , archivePrefix=

    Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planning , author=. 2601.16163 , archivePrefix=

    arXiv

  46. [54]

    2503.09871 , archivePrefix=

    LuciBot: Automated Robot Policy Learning from Generated Videos , author=. 2503.09871 , archivePrefix=

    arXiv

  47. [55]

    2409.16283 , archivePrefix=

    Gen2act: Human video generation in novel scenarios enables generalizable robot manipulation , author=. 2409.16283 , archivePrefix=

    arXiv

  48. [56]

    2410.10076 , archivePrefix=

    Videoagent: Self-improving video generation , author=. 2410.10076 , archivePrefix=

    arXiv

  49. [57]

    2507.04447 , archivePrefix=

    Dreamvla: a vision-language-action model dreamed with comprehensive world knowledge , author=. 2507.04447 , archivePrefix=

    arXiv

  50. [58]

    2025 , eprint=

    Unified Video Action Model , author=. 2025 , eprint=

    2025

  51. [59]

    2510.10125 , archivePrefix=

    Ctrl-world: A controllable generative world model for robot manipulation , author=. 2510.10125 , archivePrefix=

    arXiv

  52. [60]

    2511.11520 , archivePrefix=

    Scalable policy evaluation with video world models , author=. 2511.11520 , archivePrefix=

  53. [61]

    Acm siggraph 2016 courses , pages=

    The material point method for simulating continuum materials , author=. Acm siggraph 2016 courses , pages=. 2016 , publisher=

    2016

  54. [62]

    2024 , organization=

    Liu, Shaowei and Ren, Zhongzheng and Gupta, Saurabh and Wang, Shenlong , booktitle=. 2024 , organization=

    2024

  55. [63]

    Journal of Computational physics , volume=

    Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , author=. Journal of Computational physics , volume=. 2019 , publisher=

    2019

  56. [64]

    Chen, Chuhao and Dou, Zhiyang and Wang, Chen and Huang, Yiming and Chen, Anjun and Feng, Qiao and Gu, Jiatao and Liu, Lingjie , booktitle=

  57. [65]

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

    Physgen3d: Crafting a miniature interactive world from a single image , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  58. [66]

    Li, Zizhang and Yu, Hong-Xing and Liu, Wei and Yang, Yin and Herrmann, Charles and Wetzstein, Gordon and Wu, Jiajun , booktitle=

  59. [67]

    Liu, Wei and Chen, Ziyu and Li, Zizhang and Wang, Yue and Yu, Hong-Xing and Wu, Jiajun , eprint=

  60. [68]

    Lv, Jiaxi and Huang, Yi and Yan, Mingfu and Huang, Jiancheng and Liu, Jianzhuang and Liu, Yifan and Wen, Yafei and Chen, Xiaoxin and Chen, Shifeng , booktitle=

  61. [69]

    Xie, Tianyi and Zhao, Yiwei and Jiang, Ying and Jiang, Chenfanfu , booktitle=

  62. [70]

    ACM Transactions on Graphics (TOG) , volume=

    A material point method for snow simulation , author=. ACM Transactions on Graphics (TOG) , volume=. 2013 , publisher=

    2013

  63. [71]

    Li, Xuan and Qiao, Yi-Ling and Chen, Peter Yichen and Jatavallabhula, Krishna Murthy and Lin, Ming and Jiang, Chenfanfu and Gan, Chuang , eprint=

  64. [72]

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

    Physgaussian: Physics-integrated 3d gaussians for generative dynamics , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  65. [73]

    2507.07982 , archivePrefix=

    Geometry forcing: Marrying video diffusion and 3d representation for consistent world modeling , author=. 2507.07982 , archivePrefix=

    arXiv

  66. [74]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Wonderworld: Interactive 3d scene generation from a single image , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  67. [75]

    2509.21657 , archivePrefix=

    Fantasyworld: Geometry-consistent world modeling via unified video and 3d prediction , author=. 2509.21657 , archivePrefix=

  68. [76]

    Lu, Haoran and Wu, Shang and Liu, Songling and Zhang, Jianshu and Su, Maojiang and Ye, Guo and others , eprint=

  69. [77]

    Zhou, Siyuan and Wang, Hejun and Cheng, Hu and Li, Jinxi and Wang, Dongsheng and Jiang, Junwei and Jin, Yixiao and Huang, Jiayue and Mao, Shiwei and Liu, Shangjia and others , eprint=

  70. [78]

    Xue, Qiyao and Yin, Xiangyu and Yang, Boyuan and Gao, Wei , booktitle=

  71. [79]

    2025 , eprint=

    Think Before You Diffuse: Infusing Physical Rules into Video Diffusion , author=. 2025 , eprint=

    2025

  72. [80]

    2509.24702 , archivePrefix=

    Enhancing physical plausibility in video generation by reasoning the implausibility , author=. 2509.24702 , archivePrefix=

    arXiv

  73. [81]

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

    Chain of Event-Centric Causal Thought for Physically Plausible Video Generation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  74. [82]

    Wang, Jing and Ma, Ao and Cao, Ke and Zheng, Jun and Zhang, Zhanjie and Feng, Jiasong and Liu, Shanyuan and Ma, Yuhang and Cheng, Bo and Leng, Dawei and others , eprint=

  75. [83]

    2025 , url=

    Li, Chenyu and Michel, Oscar and Pan, Xichen and Liu, Sainan and Roberts, Mike and Xie, Saining , booktitle=. 2025 , url=

    2025

  76. [84]

    Pathak, Saurabh and Arani, Elahe and Pechenizkiy, Mykola and Zonooz, Bahram , booktitle=

  77. [85]

    Wang, Zijun and Hu, Panwen and Wang, Jing and Zhang, Terry Jingchen and Cheng, Yuhao and Chen, Long and others , booktitle=

  78. [86]

    Ji, Sihui and Chen, Xi and Tao, Xin and Wan, Pengfei and Zhao, Hengshuang , eprint=

  79. [87]

    2504.15932 , archivePrefix=

    Reasoning physical video generation with diffusion timestep tokens via reinforcement learning , author=. 2504.15932 , archivePrefix=

    arXiv

  80. [88]

    Gillman, Nate and Zhou, Yinghua and Tang, Zitian and Luo, Evan and Chakravarthy, Arjan and Aggarwal, Daksh and Freeman, Michael and Sun, Chen , booktitle=

  81. [89]

    Zhang, Qiyuan and Gong, Biao and Tan, Shuai and Zhang, Zheng and Shen, Yujun and Zhu, Xing and Li, Yuyuan and Yao, Kelu and Shen, Chunhua and Zou, Changqing , eprint=

  82. [90]

    Advances in Neural Information Processing Systems , volume=

    Motioncraft: Physics-based zero-shot video generation , author=. Advances in Neural Information Processing Systems , volume=

  83. [91]

    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=

  84. [92]

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

    Diffusion model alignment using direct preference optimization , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  85. [93]

    Advances in Neural Information Processing Systems , volume=

    Training language models to follow instructions with human feedback , author=. Advances in Neural Information Processing Systems , volume=

  86. [94]

    2402.03300 , archivePrefix=

    DeepSeekMath: Pushing the limits of mathematical reasoning in open language models , author=. 2402.03300 , archivePrefix=

    arXiv

  87. [95]

    Liu, Jie and Liu, Gongye and Liang, Jiajun and Li, Yangguang and Liu, Jiaheng and Wang, Xintao and Wan, Pengfei and Zhang, Di and Ouyang, Wanli , eprint=

  88. [96]

    2505.07818 , archivePrefix=

    DanceGRPO: Unleashing GRPO on visual generation , author=. 2505.07818 , archivePrefix=

    arXiv

  89. [97]

    2603.17051 , archivePrefix=

    Astrolabe: Steering Forward-Process Reinforcement Learning for Distilled Autoregressive Video Models , author=. 2603.17051 , archivePrefix=

  90. [98]

    2605.14278 , archivePrefix=

    KVPO: ODE-Native GRPO for Autoregressive Video Alignment via KV Semantic Exploration , author=. 2605.14278 , archivePrefix=

    arXiv

  91. [99]

    2210.02747 , archivePrefix=

    Flow matching for generative modeling , author=. 2210.02747 , archivePrefix=

    arXiv

  92. [100]

    The Eleventh International Conference on Learning Representations , year=

    Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow , author=. The Eleventh International Conference on Learning Representations , year=

  93. [101]

    2024 , organization=

    DynoSurf: Neural Deformation-based Temporally Consistent Dynamic Surface Reconstruction , author=. 2024 , organization=

    2024

  94. [102]

    Cadex: Learning canonical deformation coordinate space for dynamic surface representation via neural homeomorphism , author=

  95. [103]

    Motion2vecsets: 4d latent vector set diffusion for non-rigid shape reconstruction and tracking , author=

  96. [104]

    2405.15056 , archivePrefix=

    ElastoGen: 4D Generative Elastodynamics , author=. 2405.15056 , archivePrefix=

  97. [105]

    2412.05161 , archivePrefix=

    DNF: Unconditional 4D Generation with Dictionary-based Neural Fields , author=. 2412.05161 , archivePrefix=

    arXiv

  98. [106]

    2406.17988 , archivePrefix=

    DICE: End-to-end Deformation Capture of Hand-Face Interactions from a Single Image , author=. 2406.17988 , archivePrefix=

    arXiv

  99. [107]

    Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

    SMPL: A skinned multi-person linear model , author=. Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

  100. [108]

    3D menagerie: Modeling the 3D shape and pose of animals , author=

  101. [109]

    2017 , publisher=

    Embodied Hands: Modeling and Capturing Hands and Bodies Together , author=. 2017 , publisher=

    2017

  102. [110]

    Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

    A morphable model for the synthesis of 3D faces , author=. Seminal Graphics Papers: Pushing the Boundaries, Volume 2 , pages=

  103. [111]

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

    Driving into the future: Multiview visual forecasting and planning with world model for autonomous driving , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  104. [112]

    2211.13221 , archivePrefix=

    Latent video diffusion models for high-fidelity long video generation , author=. 2211.13221 , archivePrefix=

    arXiv

  105. [113]

    2024 , organization=

    Dynamicrafter: Animating open-domain images with video diffusion priors , author=. 2024 , organization=

    2024

  106. [114]

    Oasis: A universe in a transformer , author=

  107. [115]

    2408.14837 , archivePrefix=

    Diffusion models are real-time game engines , author=. 2408.14837 , archivePrefix=

    arXiv

  108. [116]

    2310.19512 , archivePrefix=

    Videocrafter1: Open diffusion models for high-quality video generation , author=. 2310.19512 , archivePrefix=

    arXiv

  109. [117]

    2411.00769 , archivePrefix=

    Gamegen-x: Interactive open-world game video generation , author=. 2411.00769 , archivePrefix=

    arXiv

  110. [118]

    2024 , organization=

    Champ: Controllable and consistent human image animation with 3d parametric guidance , author=. 2024 , organization=

    2024

  111. [119]

    Proceedings of the 9th International Conference on Motion in Games , pages=

    XPBD: position-based simulation of compliant constrained dynamics , author=. Proceedings of the 9th International Conference on Motion in Games , pages=

  112. [120]

    2024 , publisher=

    Simplicits: Mesh-free, geometry-agnostic elastic simulation , author=. 2024 , publisher=

    2024

  113. [121]

    , author=

    Learning a model of facial shape and expression from 4D scans. , author=

  114. [122]

    2404.12379 , archivePrefix=

    Dynamic Gaussians Mesh: Consistent Mesh Reconstruction from Monocular Videos , author=. 2404.12379 , archivePrefix=

    arXiv

  115. [123]

    Physdiff: Physics-guided human motion diffusion model , author=

  116. [124]

    2024 , organization=

    Emdm: Efficient motion diffusion model for fast and high-quality motion generation , author=. 2024 , organization=

    2024

  117. [125]

    2208.15001 , archivePrefix=

    Motiondiffuse: Text-driven human motion generation with diffusion model , author=. 2208.15001 , archivePrefix=

    arXiv

  118. [126]

    Human Motion Diffusion Model , author=

  119. [127]

    Mofusion: A framework for denoising-diffusion-based motion synthesis , author=

  120. [128]

    2024 , organization=

    Tlcontrol: Trajectory and language control for human motion synthesis , author=. 2024 , organization=

    2024

  121. [129]

    Denoising diffusion probabilistic models , author=

  122. [130]

    Denoising diffusion implicit models , author=

  123. [131]

    Generative modeling by estimating gradients of the data distribution , author=

  124. [132]

    Score-Based Generative Modeling through Stochastic Differential Equations , author=

  125. [133]

    ICLR Workshop on Deep Generative Models for Highly Structured Data , year=

    Video Diffusion Models , author=. ICLR Workshop on Deep Generative Models for Highly Structured Data , year=

  126. [134]

    High-resolution image synthesis with latent diffusion models , author=

  127. [135]

    Photorealistic text-to-image diffusion models with deep language understanding , author=

  128. [136]

    Diffuseq: Sequence to sequence text generation with diffusion models , author=

  129. [137]

    2023 , organization=

    Nerfdiff: Single-image view synthesis with nerf-guided distillation from 3d-aware diffusion , author=. 2023 , organization=

    2023

  130. [138]

    2210.02303 , archivePrefix=

    Imagen video: High definition video generation with diffusion models , author=. 2210.02303 , archivePrefix=

    arXiv

  131. [139]

    Zero-1-to-3: Zero-shot one image to 3d object , author=

  132. [140]

    2024 , publisher=

    CLAY: A Controllable Large-scale Generative Model for Creating High-quality 3D Assets , author=. 2024 , publisher=

    2024

  133. [141]

    3d neural field generation using triplane diffusion , author=

  134. [142]

    2024 , organization=

    Physdreamer: Physics-based interaction with 3d objects via video generation , author=. 2024 , organization=

    2024

  135. [143]

    Npms: Neural parametric models for 3d deformable shapes , author=

  136. [144]

    H4d: Human 4d modeling by learning neural compositional representation , author=

  137. [145]

    Hyperdiffusion: Generating implicit neural fields with weight-space diffusion , author=

  138. [146]

    Neural ordinary differential equations , author=

  139. [147]

    Learning parallel dense correspondence from spatio-temporal descriptors for efficient and robust 4d reconstruction , author=

  140. [148]

    Occupancy flow: 4d reconstruction by learning particle dynamics , author=

  141. [149]

    Occupancy networks: Learning 3d reconstruction in function space , author=

  142. [150]

    ACM SIGGRAPH 2024 Conference Papers , pages=

    Vr-gs: A physical dynamics-aware interactive gaussian splatting system in virtual reality , author=. ACM SIGGRAPH 2024 Conference Papers , pages=

    2024

  143. [151]

    Pie-nerf: Physics-based interactive elastodynamics with nerf , author=

  144. [152]

    2022 , publisher=

    Physics informed neural fields for smoke reconstruction with sparse data , author=. 2022 , publisher=

    2022

  145. [153]

    Inferring hybrid neural fluid fields from videos , author=

  146. [154]

    Nerf: Representing scenes as neural radiance fields for view synthesis , author=

  147. [155]

    ACM SIGGRAPH 2024 Conference Papers , pages=

    Physics-Informed Learning of Characteristic Trajectories for Smoke Reconstruction , author=. ACM SIGGRAPH 2024 Conference Papers , pages=

    2024

  148. [156]

    1977 , publisher=

    The finite element method , author=. 1977 , publisher=

    1977

  149. [157]

    2013 , publisher=

    Fast simulation of mass-spring systems , author=. 2013 , publisher=

    2013

  150. [158]

    , author=

    3d gaussian splatting for real-time radiance field rendering. , author=

  151. [159]

    Wonder3d: Single image to 3d using cross-domain diffusion , author=

  152. [160]

    2403.13307 , archivePrefix=

    Laserhuman: language-guided scene-aware human motion generation in free environment , author=. 2403.13307 , archivePrefix=

    arXiv

  153. [161]

    2411.17189 , archivePrefix=

    PhysMotion: Physics-Grounded Dynamics From a Single Image , author=. 2411.17189 , archivePrefix=

    arXiv

  154. [162]

    2408.06072 , archivePrefix=

    Cogvideox: Text-to-video diffusion models with an expert transformer , author=. 2408.06072 , archivePrefix=

    arXiv

  155. [163]

    2311.15127 , archivePrefix=

    Stable video diffusion: Scaling latent video diffusion models to large datasets , author=. 2311.15127 , archivePrefix=

    arXiv

  156. [164]

    2412.03603 , archivePrefix=

    Hunyuanvideo: A systematic framework for large video generative models , author=. 2412.03603 , archivePrefix=

    arXiv

  157. [165]

    Objaverse: A universe of annotated 3d objects , author=

  158. [166]

    2024 , organization=

    Sv3d: Novel multi-view synthesis and 3d generation from a single image using latent video diffusion , author=. 2024 , organization=

    2024

  159. [167]

    2024 , organization=

    Lgm: Large multi-view gaussian model for high-resolution 3d content creation , author=. 2024 , organization=

    2024

  160. [168]

    Segment anything , author=

  161. [169]

    2501.03847 , archivePrefix=

    Diffusion as Shader: 3D-aware Video Diffusion for Versatile Video Generation Control , author=. 2501.03847 , archivePrefix=

    arXiv

  162. [170]

    2412.02700 , archivePrefix=

    Motion prompting: Controlling video generation with motion trajectories , author=. 2412.02700 , archivePrefix=

    arXiv

  163. [171]

    Animate anyone: Consistent and controllable image-to-video synthesis for character animation , author=

  164. [172]

    2404.02101 , archivePrefix=

    Cameractrl: Enabling camera control for text-to-video generation , author=. 2404.02101 , archivePrefix=

    arXiv

  165. [173]

    2024 , organization=

    Reconstruction and simulation of elastic objects with spring-mass 3d gaussians , author=. 2024 , organization=

    2024

  166. [174]

    Videocrafter2: Overcoming data limitations for high-quality video diffusion models , author=

  167. [175]

    2412.07759 , archivePrefix=

    3DTrajMaster: Mastering 3D Trajectory for Multi-Entity Motion in Video Generation , author=. 2412.07759 , archivePrefix=

    arXiv

  168. [176]

    2024 , organization=

    DriveDreamer: Towards Real-World-Drive World Models for Autonomous Driving , author=. 2024 , organization=

    2024

  169. [177]

    2411.02385 , archivePrefix=

    How far is video generation from world model: A physical law perspective , author=. 2411.02385 , archivePrefix=

    arXiv

  170. [178]

    2501.09038 , archivePrefix=

    Do generative video models learn physical principles from watching videos? , author=. 2501.09038 , archivePrefix=

    arXiv

  171. [179]

    2502.11831 , archivePrefix=

    Intuitive physics understanding emerges from self-supervised pretraining on natural videos , author=. 2502.11831 , archivePrefix=

    arXiv

  172. [180]

    2024 , organization=

    Draganything: Motion control for anything using entity representation , author=. 2024 , organization=

    2024

  173. [181]

    2412.07721 , archivePrefix=

    ObjCtrl-2.5 D: Training-free Object Control with Camera Poses , author=. 2412.07721 , archivePrefix=

    arXiv

  174. [182]

    ACM Transactions on Graphics (TOG) , volume=

    The affine particle-in-cell method , author=. ACM Transactions on Graphics (TOG) , volume=. 2015 , publisher=

    2015

  175. [183]

    Proceedings of the 14th ACM SIGGRAPH/Eurographics Symposium on Computer Animation , pages=

    A material point method for viscoelastic fluids, foams and sponges , author=. Proceedings of the 14th ACM SIGGRAPH/Eurographics Symposium on Computer Animation , pages=

  176. [184]

    ACM Transactions on Graphics (TOG) , volume=

    Drucker-prager elastoplasticity for sand animation , author=. ACM Transactions on Graphics (TOG) , volume=. 2016 , publisher=

    2016

  177. [185]

    ACM Transactions on Graphics (TOG) , volume=

    Anisotropic elastoplasticity for cloth, knit and hair frictional contact , author=. ACM Transactions on Graphics (TOG) , volume=. 2017 , publisher=

    2017

  178. [186]

    ACM Transactions on Graphics (TOG) , volume=

    A moving least squares material point method with displacement discontinuity and two-way rigid body coupling , author=. ACM Transactions on Graphics (TOG) , volume=. 2018 , publisher=

    2018

  179. [187]

    Advances in Neural Information Processing Systems , volume=

    Objaverse-xl: A universe of 10m+ 3d objects , author=. Advances in Neural Information Processing Systems , volume=

  180. [188]

    2412.01506 , archivePrefix=

    Structured 3d latents for scalable and versatile 3d generation , author=. 2412.01506 , archivePrefix=

    arXiv

  181. [189]

    NVIDIA GPU Technology Conference (GTC) , year=

    Warp: A high-performance python framework for gpu simulation and graphics , author=. NVIDIA GPU Technology Conference (GTC) , year=

  182. [190]

    Computer Animation and Simulation/Springer Vienna , year=

    Smoothed particles: A new paradigm for animating highly deformable bodies , author=. Computer Animation and Simulation/Springer Vienna , year=

  183. [191]

    Computer Graphics Forum , volume=

    An implicit SPH formulation for incompressible linearly elastic solids , author=. Computer Graphics Forum , volume=. 2018 , organization=

    2018

  184. [192]

    Proceedings of the ACM on Computer Graphics and Interactive Techniques , volume=

    Fast corotated elastic SPH solids with implicit zero-energy mode control , author=. Proceedings of the ACM on Computer Graphics and Interactive Techniques , volume=. 2021 , publisher=

    2021

  185. [193]

    2411.18613 , archivePrefix=

    Cat4d: Create anything in 4d with multi-view video diffusion models , author=. 2411.18613 , archivePrefix=

    arXiv

  186. [194]

    2022 , note=

    Warp: A High-performance Python Framework for GPU Simulation and Graphics , author =. 2022 , note=

    2022

  187. [195]

    2503.20314 , archivePrefix=

    Wan: Open and Advanced Large-Scale Video Generative Models , author=. 2503.20314 , archivePrefix=

    arXiv

  188. [196]

    ACM Transactions on Graphics (TOG) , volume=

    Taichi: a language for high-performance computation on spatially sparse data structures , author=. ACM Transactions on Graphics (TOG) , volume=. 2019 , publisher=

    2019

  189. [197]

    2406.03520 , archivePrefix=

    Videophy: Evaluating physical commonsense for video generation , author=. 2406.03520 , archivePrefix=

    arXiv

  190. [198]

    Bansal, Hritik and Peng, Clark and Bitton, Yonatan and Goldenberg, Roman and Grover, Aditya and Chang, Kai-Wei , eprint=

  191. [199]

    Huang, Ziqi and He, Yinan and Yu, Jiashuo and Zhang, Fan and Si, Chenyang and Jiang, Yuming and Zhang, Yuanhan and Wu, Tianxing and Jin, Qingyang and Chanpaisit, Nattapol and Wang, Yaohui and Chen, Xinyuan and Wang, Limin and Lin, Dahua and Qiao, Yu and Liu, Ziwei , booktitle=

  192. [200]

    Zhang, Xiangdong and Liao, Jiaqi and Zhang, Shaofeng and Meng, Fanqing and Wan, Xiangpeng and Yan, Junchi and Cheng, Yu , booktitle=

  193. [201]

    Journal of Visual Communication and Image Representation , volume=

    Position based dynamics , author=. Journal of Visual Communication and Image Representation , volume=. 2007 , publisher=

    2007

  194. [202]

    ACM Transactions on Graphics (TOG) , volume=

    Position based fluids , author=. ACM Transactions on Graphics (TOG) , volume=. 2013 , publisher=

    2013

  195. [203]

    2201.11976 , archivePrefix=

    Learning to simulate unseen physical systems with graph neural networks , author=. 2201.11976 , archivePrefix=

    arXiv

  196. [204]

    Proceedings of Robotics: Science and Systems (RSS) , year=

    AdaptiGraph: Material-Adaptive Graph-Based Neural Dynamics for Robotic Manipulation , author=. Proceedings of Robotics: Science and Systems (RSS) , year=

  197. [205]

    2509.21625 , archivePrefix=

    Guiding Audio Editing with Audio Language Model , author=. 2509.21625 , archivePrefix=

  198. [206]

    2009.09761 , archivePrefix=

    Diffwave: A versatile diffusion model for audio synthesis , author=. 2009.09761 , archivePrefix=

    2009 arXiv

  199. [207]

    International Conference on Machine Learning , pages=

    Make-an-audio: Text-to-audio generation with prompt-enhanced diffusion models , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  200. [208]

    International conference on machine learning , pages=

    Learning to simulate complex physics with graph networks , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  201. [209]

    The International Journal of Robotics Research , volume=

    Robocraft: Learning to see, simulate, and shape elasto-plastic objects in 3d with graph networks , author=. The International Journal of Robotics Research , volume=. 2024 , publisher=

    2024

  202. [210]

    2505.19386 , archivePrefix=

    Force Prompting: Video Generation Models Can Learn and Generalize Physics-based Control Signals , author=. 2505.19386 , archivePrefix=

  203. [211]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Go-with-the-flow: Motion-controllable video diffusion models using real-time warped noise , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  204. [212]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Vggt: Visual geometry grounded transformer , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  205. [213]

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

    Sc-gs: Sparse-controlled gaussian splatting for editable dynamic scenes , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  206. [214]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    Mosca: Dynamic gaussian fusion from casual videos via 4d motion scaffolds , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  207. [215]

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

    Do generative video models understand physical principles? , author=. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages=

  208. [216]

    2410.05363 , archivePrefix=

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

    arXiv

  209. [217]

    , booktitle=

    Yi, Kexin and Gan, Chuang and Li, Yunzhu and Kohli, Pushmeet and Wu, Jiajun and Torralba, Antonio and Tenenbaum, Joshua B. , booktitle=

  210. [218]

    and Wang, Elias and Mrowca, Damian and Binder, Felix and Tung, Hsiao-Yu Fish and Pramod, R

    Bear, Daniel M. and Wang, Elias and Mrowca, Damian and Binder, Felix and Tung, Hsiao-Yu Fish and Pramod, R. T. and Holdaway, Cameron and Tao, Sirui and Smith, Kevin and Sun, Fan-Yun and Fei-Fei, Li and Kanwisher, Nancy and Tenenbaum, Joshua B. and Yamins, Daniel L. K. and Fan,...

  211. [219]

    and Yamins, Daniel L

    Tung, Hsiao-Yu and Ding, Mingyu and Chen, Zhenfang and Bear, Daniel and Gan, Chuang and Tenenbaum, Joshua B. and Yamins, Daniel L. K. and Fan, Judith E. and Smith, Kevin A. , booktitle=

  212. [220]

    Bansal, Hritik and Lin, Zongyu and Xie, Tianyi and Zong, Zeshun and Yarom, Michal and Bitton, Yonatan and Jiang, Chenfanfu and Sun, Yizhou and Chang, Kai-Wei and Grover, Aditya , booktitle=

  213. [221]

    International Conference on Learning Representations , year=

    NewtonGen: Physics-Consistent and Controllable Text-to-Video Generation via Neural Newtonian Dynamics , author=. International Conference on Learning Representations , year=

  214. [222]

    Proceedings of the 32nd ACM International Conference on Multimedia , pages=

    Soucek, Tom. Proceedings of the 32nd ACM International Conference on Multimedia , pages=

  215. [223]

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

    Exploring video quality assessment on user generated contents from aesthetic and technical perspectives , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  216. [224]

    2020 , organization=

    Teed, Zachary and Deng, Jia , booktitle=. 2020 , organization=

    2020

  217. [225]

    Bai, Shuai and Cai, Yuxuan and Chen, Ruizhe and Chen, Keqin and Chen, Xionghui and Cheng, Zesen and Deng, Lianghao and Ding, Wei and Gao, Chang and Ge, Chunjiang and others , eprint=

  218. [226]

    2021 , howpublished =

    Beik-Mohammadi, Hadi , title =. 2021 , howpublished =

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.