pith. sign in

arxiv: 2606.30248 · v1 · pith:BFJVXWMBnew · submitted 2026-06-29 · 💻 cs.CV · cs.LG

Your Data Manifold is Secretly a Reward Model: Shell-LCC for Text-to-Video Generation

Shihao Zhang , Yuguang Yan , Junzhe Zhang , Wei Zhao , Bohan Wang , Hanwang Zhang This is my paper

Pith reviewed 2026-06-30 06:05 UTC · model grok-4.3

classification 💻 cs.CV cs.LG
keywords text-to-video generationdiffusion modelsmanifold learningreward modelinglocal coordinate codingsupervised fine-tuningvideo quality
0
0 comments X

The pith

The manifold of high-quality SFT data can serve as a dense, cost-free reward model for text-to-video diffusion.

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

The paper claims that high-quality supervised fine-tuning data already encodes a geometric manifold whose structure can be turned into reward signals for aligning generated videos with human preferences. By fitting this manifold with Shell Local Coordinate Coding, an extension of standard LCC, the method pulls video latents onto the high-density surface without training separate reward networks or collecting new annotations. This produces dense, differentiable guidance that targets low-level distortions and fine details more effectively than existing auxiliary signals. A reader would care because current alignment approaches add heavy computation and annotation costs while still struggling with local artifacts.

Core claim

By explicitly modeling the manifold structure of high-quality SFT data and encouraging video latents to lie on this manifold, we derive dense, differentiable, and nearly cost-free reward signals that significantly improve video quality, particularly in mitigating low-level distortions. The modeling builds upon Local Coordinate Coding but extends it to Shell-LCC, which represents the manifold surface as an isotropic shell to avoid mean regression and preserve high-frequency details.

What carries the argument

Shell-LCC, an extension of Local Coordinate Coding that models the manifold as an isotropic shell surface to supply differentiable alignment signals from existing SFT data.

If this is right

  • Generated videos exhibit improved realism, sharper high-frequency details, and reduced over-smoothing and motion blur.
  • Alignment occurs without training separate reward models or incurring annotation costs.
  • The reward signals integrate directly into existing diffusion pipelines as an auxiliary loss term.
  • The approach targets low-level distortions that standard reward models often miss.

Where Pith is reading between the lines

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

  • Similar manifold modeling could supply alignment signals for text-to-image or audio generation tasks that also rely on SFT data.
  • Iteratively updating the fitted shell with newly generated high-quality samples might create a self-improving loop.
  • The shell-surface representation may combine with other latent-space regularizers already used in diffusion training.

Load-bearing premise

The geometric manifold fitted to SFT data points aligns with human aesthetic preferences so that pulling latents onto its surface improves fine details without new artifacts.

What would settle it

Blind human preference tests or perceptual metrics on videos generated with versus without Shell-LCC guidance show no gain in realism or detail sharpness, or show introduced artifacts.

Figures

Figures reproduced from arXiv: 2606.30248 by Bohan Wang, Hanwang Zhang, Junzhe Zhang, Shihao Zhang, Wei Zhao, Yuguang Yan.

Figure 1
Figure 1. Figure 1: Distribution of the manifold distance Rdist for Ground Truth and generated videos. Generated videos show a clear distribution shift, indicating manifold drift [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Qualitative comparison of generated videos. Methods from top to bottom: SFT baseline, +LCC, +Shell-LCC, and +DPO. As shown, adding Shell-LCC significantly improves video quality by mitigat￾ing low-level distortions. Specifically, it yields cleaner backgrounds with richer high-frequency details (red boxes) and recovers realistic, fine-grained textures instead of the over-smoothed “plastic” look common in AI… view at source ↗
Figure 3
Figure 3. Figure 3: Evolution of motion deblurring and training stability. While Shell-LCC reduces the motion blur observed in the baseline, prolonged training (e.g., at iter. 4999) triggers a regression towards the mean (zˆ), leading to model collapse. This trade-off highlights the inherent conflict between eliminating out-of-manifold information (blur) and preserving high-frequency details, necessitating controlled optimiza… view at source ↗
Figure 4
Figure 4. Figure 4: Radial reconstruction from zˆ. Videos transition from mean-like blur at zˆ to sharper structures as the radius increases, while excessive deviation introduces distortions, revealing a shell-shaped latent manifold. direction: z(r) = \hat z + r \cdot \frac {z - \hat z}{\|z - \hat z\|}. As shown in [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison on open-source models, complementing the quantitative results in Tab. 3. From top to bottom: Wan-T2V-1.3B, Wan-T2V-1.3B + Shell-LCC, UltraWan-T2V-1.3B, and UltraWan￾T2V-1.3B + Shell-LCC, for the prompt “A 3D model of an 1800s Victorian house”. Adding Shell-LCC sharpens fine structures (e.g., window lattices and facade details) and suppresses the over-smoothing of the baselines, confi… view at source ↗
read the original abstract

Recent text-to-video (T2V) diffusion models rely heavily on auxiliary reward signals (e.g., via reward models or DPO) to align generated content with human aesthetics and improve realism. These signals, however, incur substantial computational overhead, require costly human annotations, and often yield limited improvement in fine-grained local details. In this paper, we argue that your data manifold is secretly a reward model. By explicitly modeling the manifold structure of high-quality Supervised Fine-Tuning (SFT) data and encouraging video latents to lie on this manifold, we derive dense, differentiable, and nearly cost-free reward signals that significantly improve video quality, particularly in mitigating low-level distortions. Our modeling builds upon Local Coordinate Coding (LCC), which captures the `skeleton' of the manifold. However, directly applying LCC suffers from mean regression, pulling latents toward the geometric mean and losing high-frequency details. We therefore extend it to Shell Local Coordinate Coding (Shell-LCC), which models the manifold `surface' as an isotropic shell to align with the true high-density region. Experiments demonstrate that our approach improves realism, enhances high-frequency details, reduces over-smoothing artifacts, and alleviates motion blur.

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

3 major / 0 minor

Summary. The paper claims that the manifold structure of high-quality SFT data can be used as a proxy reward model for text-to-video diffusion models. It extends Local Coordinate Coding to Shell-LCC by modeling an isotropic shell around the manifold surface (rather than the geometric mean) to derive dense, differentiable, nearly cost-free reward signals that encourage generated video latents to lie on this surface, thereby improving realism, high-frequency details, and reducing artifacts such as over-smoothing and motion blur.

Significance. If the central claim holds with supporting evidence, the approach would provide a low-overhead alternative to annotation-based reward models or DPO for T2V alignment, leveraging existing SFT data geometry to mitigate low-level distortions without additional compute or human labels.

major comments (3)
  1. [Abstract] Abstract: the central claim of 'significantly improve video quality' and 'mitigating low-level distortions' is stated without any quantitative metrics, ablation studies, baseline comparisons, or human preference evaluations, rendering the claim impossible to assess.
  2. [Modeling section] The modeling section (Shell-LCC derivation): the interpretation of the derived signal as a 'reward model' rather than a manifold regularizer rests on the untested assumption that the isotropic shell fitted to SFT latents coincides with regions preferred by humans; no human preference data, comparison against standard reward models, or ablation on the shell prior versus geometric mean is supplied to support this alignment.
  3. [Experiments] Experiments section: the abstract asserts improvements in realism, detail, and artifact reduction, yet the provided text contains no tables, figures, or numerical results demonstrating these effects or ruling out mode collapse/new artifacts from the shell constraint.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback, which helps clarify the presentation of our contributions. We respond to each major comment below and indicate planned revisions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of 'significantly improve video quality' and 'mitigating low-level distortions' is stated without any quantitative metrics, ablation studies, baseline comparisons, or human preference evaluations, rendering the claim impossible to assess.

    Authors: We agree that the abstract's claims would benefit from direct reference to supporting evidence. In the revised manuscript we will expand the experiments section to include quantitative metrics, ablation studies, baseline comparisons, and human preference evaluations, and we will update the abstract to cite specific results from those evaluations. revision: yes

  2. Referee: [Modeling section] The modeling section (Shell-LCC derivation): the interpretation of the derived signal as a 'reward model' rather than a manifold regularizer rests on the untested assumption that the isotropic shell fitted to SFT latents coincides with regions preferred by humans; no human preference data, comparison against standard reward models, or ablation on the shell prior versus geometric mean is supplied to support this alignment.

    Authors: The modeling section derives the signal from the geometry of high-quality SFT data as a proxy. We acknowledge that the current version lacks direct human-preference validation and explicit comparisons. In revision we will add an ablation of the isotropic shell versus the geometric mean and comparisons to standard reward models to better substantiate the alignment claim. revision: yes

  3. Referee: [Experiments] Experiments section: the abstract asserts improvements in realism, detail, and artifact reduction, yet the provided text contains no tables, figures, or numerical results demonstrating these effects or ruling out mode collapse/new artifacts from the shell constraint.

    Authors: We agree that the experiments section must supply tables, figures, and numerical results to demonstrate the claimed effects and to address potential new artifacts. The revised manuscript will include these elements along with ablations that rule out mode collapse or introduced distortions. revision: yes

Circularity Check

0 steps flagged

No circularity detected; reward signal defined from external SFT manifold without reduction to fitted inputs or self-citations

full rationale

The provided abstract and context contain no equations or derivation steps that reduce a claimed prediction or result to its own inputs by construction. The core argument—that modeling the SFT data manifold yields a reward signal—is presented as an interpretive claim supported by experiments on quality metrics, not as a mathematical identity or fitted parameter renamed as prediction. No self-citation load-bearing steps, uniqueness theorems, or ansatzes smuggled via prior work are quoted. The method extends LCC to Shell-LCC on the manifold surface, but this is an independent modeling choice evaluated against external benchmarks rather than tautological. Per rules, absent explicit quotes showing reduction (e.g., Eq. X = Eq. Y by fit), the finding is no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

Abstract-only review; ledger populated from stated assumptions in the abstract. The approach rests on the existence of a learnable manifold in SFT latents that coincides with quality and on the shell geometry avoiding mean regression.

axioms (2)
  • domain assumption High-quality SFT data lies on a low-dimensional manifold whose local geometry can be captured by coordinate coding.
    Invoked in the first paragraph to justify using the data manifold as reward.
  • ad hoc to paper An isotropic shell around the manifold surface aligns with the true high-density region better than the geometric mean.
    Introduced to address mean regression in standard LCC.
invented entities (1)
  • Shell Local Coordinate Coding (Shell-LCC) no independent evidence
    purpose: To model the manifold surface rather than its interior mean for reward signal derivation.
    New modeling choice presented as the key technical extension.

pith-pipeline@v0.9.1-grok · 5764 in / 1305 out tokens · 27377 ms · 2026-06-30T06:05:40.817544+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

49 extracted references · 20 canonical work pages · 15 internal anchors

  1. [1]

    Latent space oddity: on the curvature of deep generative models.arXiv preprint arXiv:1710.11379, 2017

    Arvanitidis, G., Hansen, L.K., Hauberg, S.: Latent space oddity: on the curvature of deep generative models. arXiv preprint arXiv:1710.11379 (2017)

    work page arXiv 2017

  2. [2]

    VideoPhy: Evaluating Physical Commonsense for Video Generation

    Bansal, H., Lin, Z., Xie, T., Zong, Z., Yarom, M., Bitton, Y., Jiang, C., Sun, Y., Chang, K.W., Grover, A.: Videophy: Evaluating physical commonsense for video generation. arXiv preprint arXiv:2406.03520 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  3. [3]

    In: SIGGRAPH Asia 2024 Conference Papers

    Bar-Tal, O., Chefer, H., Tov, O., Herrmann, C., Paiss, R., Zada, S., Ephrat, A., Hur, J., Liu, G., Raj, A., et al.: Lumiere: A space-time diffusion model for video generation. In: SIGGRAPH Asia 2024 Conference Papers. pp. 1–11 (2024)

    2024

  4. [4]

    Neural computation15(6), 1373–1396 (2003)

    Belkin, M., Niyogi, P.: Laplacian eigenmaps for dimensionality reduction and data representation. Neural computation15(6), 1373–1396 (2003)

    2003

  5. [5]

    Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets

    Blattmann, A., Dockhorn, T., Kulal, S., Mendelevitch, D., Kilian, M., Lorenz, D., Levi, Y., English, Z., Voleti, V., Letts, A., et al.: Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  6. [6]

    OpenAI Blog1(8), 1 (2024)

    Brooks, T., Peebles, B., Holmes, C., DePue, W., Guo, Y., Jing, L., Schnurr, D., Taylor, J., Luhman, T., Luhman, E., et al.: Video generation models as world simulators. OpenAI Blog1(8), 1 (2024)

    2024

  7. [7]

    Advances in neural information processing systems30(2017)

    Christiano, P.F., Leike, J., Brown, T., Martic, M., Legg, S., Amodei, D.: Deep reinforcement learning from human preferences. Advances in neural information processing systems30(2017)

    2017

  8. [8]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Dai, M., Hang, H.: Manifold matching via deep metric learning for generative modeling. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 6587–6597 (2021)

    2021

  9. [9]

    Advances in neural information processing systems35, 2406–2422 (2022)

    De Bortoli, V., Mathieu, E., Hutchinson, M., Thornton, J., Teh, Y.W., Doucet, A.: Riemannian score- based generative modelling. Advances in neural information processing systems35, 2406–2422 (2022)

    2022

  10. [10]

    In: Forty-first international conference on machine learning (2024)

    Esser, P., Kulal, S., Blattmann, A., Entezari, R., Müller, J., Saini, H., Levi, Y., Lorenz, D., Sauer, A., Boesel, F., et al.: Scaling rectified flow transformers for high-resolution image synthesis. In: Forty-first international conference on machine learning (2024)

    2024

  11. [11]

    Advances in Neural Information Processing Systems36, 79858–79885 (2023)

    Fan, Y., Watkins, O., Du, Y., Liu, H., Ryu, M., Boutilier, C., Abbeel, P., Ghavamzadeh, M., Lee, K., Lee, K.: Dpok: Reinforcement learning for fine-tuning text-to-image diffusion models. Advances in Neural Information Processing Systems36, 79858–79885 (2023)

    2023

  12. [12]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Han, H., Li, S., Chen, J., Yuan, Y., Wu, Y., Deng, Y., Leong, C.T., Du, H., Fu, J., Li, Y., et al.: Video-bench: Human-aligned video generation benchmark. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 18858–18868 (2025)

    2025

  13. [13]

    Ad- vances in neural information processing systems35, 8633–8646 (2022)

    Ho, J., Salimans, T., Gritsenko, A., Chan, W., Norouzi, M., Fleet, D.J.: Video diffusion models. Ad- vances in neural information processing systems35, 8633–8646 (2022)

    2022

  14. [14]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Huang, Z., He, Y., Yu, J., Zhang, F., Si, C., Jiang, Y., Zhang, Y., Wu, T., Jin, Q., Chanpaisit, N., et al.: Vbench: Comprehensive benchmark suite for video generative models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 21807–21818 (2024)

    2024

  15. [15]

    In: ICML 2024 Workshop on Geometry-grounded Representation Learning and Generative Modeling (2024)

    Humayun, A.I., Amara, I., Schumann, C., Farnadi, G., Rostamzadeh, N., Havaei, M.: On the local ge- ometry of deep generative manifolds. In: ICML 2024 Workshop on Geometry-grounded Representation Learning and Generative Modeling (2024)

    2024

  16. [16]

    arXiv preprint arXiv:2408.08307 (2024)

    Humayun, A.I., Amara, I., Vasconcelos, C., Ramachandran, D., Schumann, C., He, J., Heller, K., Farnadi, G., Rostamzadeh, N., Havaei, M.: What secrets do your manifolds hold? understanding the local geometry of generative models. arXiv preprint arXiv:2408.08307 (2024)

    work page arXiv 2024

  17. [17]

    VideoPoet: A Large Language Model for Zero-Shot Video Generation

    Kondratyuk, D., Yu, L., Gu, X., Lezama, J., Huang, J., Schindler, G., Hornung, R., Birodkar, V., Yan, J., Chiu, M.C., et al.: Videopoet: A large language model for zero-shot video generation. arXiv preprint arXiv:2312.14125 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  18. [18]

    Aligning Text-to-Image Models using Human Feedback

    Lee, K., Liu, H., Ryu, M., Watkins, O., Du, Y., Boutilier, C., Abbeel, P., Ghavamzadeh, M., Gu, S.S.: Aligning text-to-image models using human feedback. arXiv preprint arXiv:2302.12192 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  19. [19]

    Flow Matching for Generative Modeling

    Lipman, Y., Chen, R.T., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. arXiv preprint arXiv:2210.02747 (2022) Abbreviated paper title 17

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  20. [20]

    Flow-GRPO: Training Flow Matching Models via Online RL

    Liu, J., Liu, G., Liang, J., Li, Y., Liu, J., Wang, X., Wan, P., Zhang, D., Ouyang, W.: Flow-grpo: Training flow matching models via online rl. arXiv preprint arXiv:2505.05470 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  21. [21]

    Improving Video Generation with Human Feedback

    Liu, J., Liu, G., Liang, J., Yuan, Z., Liu, X., Zheng, M., Wu, X., Wang, Q., Qin, W., Xia, M., Wang, X., Liu, X., Yang, F., Wan, P., Zhang, D., Gai, K., Yang, Y., Ouyang, W.: Improving video generation with human feedback. arXiv preprint arXiv:2501.13918 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  22. [22]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Liu, R., Wu, H., Zheng, Z., Wei, C., He, Y., Pi, R., Chen, Q.: Videodpo: Omni-preference alignment for video diffusion generation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 8009–8019 (2025)

    2025

  23. [23]

    Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow

    Liu, X., Gong, C., Liu, Q.: Flow straight and fast: Learning to generate and transfer data with rectified flow. arXiv preprint arXiv:2209.03003 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  24. [24]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Liu, Y., Cun, X., Liu, X., Wang, X., Zhang, Y., Chen, H., Liu, Y., Zeng, T., Chan, R., Shan, Y.: Eval- crafter: Benchmarking and evaluating large video generation models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 22139–22149 (2024)

    2024

  25. [25]

    Latte: Latent Diffusion Transformer for Video Generation

    Ma, X., Wang, Y., Chen, X., Jia, G., Liu, Z., Li, Y.F., Chen, C., Qiao, Y.: Latte: Latent diffusion transformer for video generation. arXiv preprint arXiv:2401.03048 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  26. [26]

    In: Forty-first International Conference on Machine Learning (2024)

    Munos, R., Valko, M., Calandriello, D., Azar, M.G., Rowland, M., Guo, Z.D., Tang, Y., Geist, M., Mes- nard, T., Fiegel, C., et al.: Nash learning from human feedback. In: Forty-first International Conference on Machine Learning (2024)

    2024

  27. [27]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Ni, Y., Koniusz, P., Hartley, R., Nock, R.: Manifold learning benefits gans. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 11265–11274 (2022)

    2022

  28. [28]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Peebles, W., Xie, S.: Scalable diffusion models with transformers. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 4195–4205 (2023)

    2023

  29. [29]

    Advances in neural information processing systems36, 53728–53741 (2023)

    Rafailov, R., Sharma, A., Mitchell, E., Manning, C.D., Ermon, S., Finn, C.: Direct preference opti- mization: Your language model is secretly a reward model. Advances in neural information processing systems36, 53728–53741 (2023)

    2023

  30. [30]

    science 290(5500), 2323–2326 (2000)

    Roweis, S.T., Saul, L.K.: Nonlinear dimensionality reduction by locally linear embedding. science 290(5500), 2323–2326 (2000)

    2000

  31. [31]

    Journal of machine learning research4(Jun), 119–155 (2003)

    Saul, L.K., Roweis, S.T.: Think globally, fit locally: unsupervised learning of low dimensional manifolds. Journal of machine learning research4(Jun), 119–155 (2003)

    2003

  32. [32]

    Make-A-Video: Text-to-Video Generation without Text-Video Data

    Singer, U., Polyak, A., Hayes, T., Yin, X., An, J., Zhang, S., Hu, Q., Yang, H., Ashual, O., Gafni, O., et al.: Make-a-video: Text-to-video generation without text-video data. arXiv preprint arXiv:2209.14792 (2022)

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  33. [33]

    Score-Based Generative Modeling through Stochastic Differential Equations

    Song, Y., Sohl-Dickstein, J., Kingma, D.P., Kumar, A., Ermon, S., Poole, B.: Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456 (2020)

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  34. [34]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Sun, K., Huang, K., Liu, X., Wu, Y., Xu, Z., Li, Z., Liu, X.: T2v-compbench: A comprehensive bench- mark for compositional text-to-video generation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 8406–8416 (2025)

    2025

  35. [35]

    science290(5500), 2319–2323 (2000)

    Tenenbaum,J.B.,Silva,V.d.,Langford,J.C.:Aglobalgeometricframeworkfornonlineardimensionality reduction. science290(5500), 2319–2323 (2000)

    2000

  36. [36]

    Vershynin, R.: High-dimensional probability: An introduction with applications in data science, vol. 47. Cambridge university press (2018)

    2018

  37. [37]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Wallace, B., Dang, M., Rafailov, R., Zhou, L., Lou, A., Purushwalkam, S., Ermon, S., Xiong, C., Joty, S., Naik, N.: Diffusion model alignment using direct preference optimization. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 8228–8238 (2024)

    2024

  38. [38]

    Wan: Open and Advanced Large-Scale Video Generative Models

    Wan, T., Wang, A., Ai, B., Wen, B., Mao, C., Xie, C.W., Chen, D., Yu, F., Zhao, H., Yang, J., Zeng, J., Wang, J., Zhang, J., Zhou, J., Wang, J., Chen, J., Zhu, K., Zhao, K., Yan, K., Huang, L., Feng, M., Zhang, N., Li, P., Wu, P., Chu, R., Feng, R., Zhang, S., Sun, S., Fang, T., Wang, T., Gui, T., Weng, T., Shen, T., Lin, W., Wang, W., Wang, W., Zhou, W.,...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  39. [39]

    Wang, B., Yue, Z., Zhang, F., Chen, S., Bi, L., Zhang, J., Song, X., Chan, K.Y., Pan, J., Wu, W., Zhou, M., Lin, W., Pan, K., Zhang, S., Jia, L., Hu, W., Zhao, W., Zhang, H.: Discrete visual tokens of autoregression, by diffusion, and for reasoning (2025),https://arxiv.org/abs/2505.07538

    work page arXiv 2025

  40. [40]

    International Journal of Computer Vision133(5), 3059–3078 (2025)

    Wang, Y., Chen, X., Ma, X., Zhou, S., Huang, Z., Wang, Y., Yang, C., He, Y., Yu, J., Yang, P., et al.: Lavie: High-quality video generation with cascaded latent diffusion models. International Journal of Computer Vision133(5), 3059–3078 (2025)

    2025

  41. [41]

    DanceGRPO: Unleashing GRPO on Visual Generation

    Xue, Z., Wu, J., Gao, Y., Kong, F., Zhu, L., Chen, M., Liu, Z., Liu, W., Guo, Q., Huang, W., et al.: Dancegrpo: Unleashing grpo on visual generation. arXiv preprint arXiv:2505.07818 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  42. [42]

    arXiv preprint arXiv:2506.13691 (2025)

    Xue, Z., Zhang, J., Hu, T., He, H., Chen, Y., Cai, Y., Wang, Y., Wang, C., Liu, Y., Li, X., Tao, D.: Ul- travideo: High-quality uhd video dataset with comprehensive captions. arXiv preprint arXiv:2506.13691 (2025)

    work page arXiv 2025

  43. [43]

    VideoGPT: Video Generation using VQ-VAE and Transformers

    Yan, W., Zhang, Y., Abbeel, P., Srinivas, A.: Videogpt: Video generation using vq-vae and transformers. arXiv preprint arXiv:2104.10157 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  44. [44]

    Advances in neural information processing systems22(2009)

    Yu, K., Zhang, T., Gong, Y.: Nonlinear learning using local coordinate coding. Advances in neural information processing systems22(2009)

    2009

  45. [45]

    In: Interna- tional Conference on Machine Learning (ICML) (2024)

    Zhang, S., Kawaguchi, K., Yao, A.: Deep regression representation learning with topology. In: Interna- tional Conference on Machine Learning (ICML) (2024)

    2024

  46. [46]

    ICLR (2025)

    Zhang, S., Yan, Y., Yao, A.: Improving deep regression with tightness. ICLR (2025)

    2025

  47. [47]

    ICLR (2023)

    Zhang, S., Yang, L., Mi, M.B., Zheng, X., Yao, A.: Improving deep regression with ordinal entropy. ICLR (2023)

    2023

  48. [48]

    arXiv preprint arXiv:2305.10425 (2023)

    Zhao, Y., Joshi, R., Liu, T., Khalman, M., Saleh, M., Liu, P.J.: Slic-hf: Sequence likelihood calibration with human feedback. arXiv preprint arXiv:2305.10425 (2023)

    work page arXiv 2023

  49. [49]

    Open-Sora: Democratizing Efficient Video Production for All

    Zheng, Z., Peng, X., Yang, T., Shen, C., Li, S., Liu, H., Zhou, Y., Li, T., You, Y.: Open-sora: Democ- ratizing efficient video production for all. arXiv preprint arXiv:2412.20404 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024