Pith. sign in

REVIEW 3 major objections 6 minor 81 references

Test-time tuning of a diffusion model on a single video can destroy its generative prior; ElasticTTT prevents this with three targeted interventions and achieves state-of-the-art one-shot video editing.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-03 01:48 UTC pith:UTZW2L3U

load-bearing objection Solid empirical paper on a real problem; the Contrastive CFG theory is a dressed-up heuristic and the Async-NS hyperparameters are fragile, but it deserves a serious referee. the 3 major comments →

arxiv 2607.21529 v2 pith:UTZW2L3U submitted 2026-07-23 cs.CV cs.AI

ElasticTTT: Prior-Preserving Test-Time Tuning for Video Editing

Yueyi Liu , Chi Zhang , Sen Cui , Miao Liu This is my paper
classification cs.CV cs.AI
keywords test-time tuningvideo editingdiffusion modelsprior collapseclassifier-free guidancenoise schedulingflow matchingone-shot generation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper identifies a failure mode it calls prior collapse: when a diffusion model is fine-tuned on one source video, it stops following text prompts and spatial structure, degenerating into regenerating the source. The authors argue this is inherent to single-point optimization against a distribution-mapping model, and that three changes—adding zero-mean noise to the training target, using the source prompt as a negative in classifier-free guidance, and running separate noise schedules for edited versus preserved regions—restore generative flexibility. If right, this makes test-time tuning a viable one-shot video editing paradigm, and the benefits grow with base-model capacity.

Core claim

The central claim is that prior collapse—not ordinary overfitting—explains why test-time tuning fails on video editing, and it can be cured without changing the expected optimization objective. The paper shows that adding zero-mean Gaussian noise to the regression target leaves the expected velocity field unchanged while injecting gradient covariance proportional to the Jacobian norm, ejecting the optimizer from sharp memorization minima. It then shows that standard classifier-free guidance actually amplifies the residual source-attractor bias, while a contrastive term using the source prompt isolates and cancels that bias. A third mechanism, asynchronous noise scheduling, decouples the deno

What carries the argument

Three mechanisms: (1) Target Distribution Regularization (TDR), which perturbs the reconstruction target with zero-mean Gaussian noise, keeping the learned vector field unbiased while adding a Gauss-Newton covariance penalty that forbids sharp memorization minima; (2) Contrastive CFG, which extends standard classifier-free guidance with an extra term contrasting target and source prompts, algebraically canceling the source-attractor bias introduced by tuning; (3) Asynchronous Noise Scheduling (Async-NS), which assigns different noise levels and timestep embeddings to masked edited regions versus preserved regions, desynchronizing their denoising paths and preventing spatial entanglement.

Load-bearing premise

The cancellation of the source-attractor bias in Contrastive CFG rests on the assumption that the TTT-induced bias points in the same direction for all prompts, so a single source-versus-target contrast can cancel it; if the bias direction varies with the prompt, the mechanism loses its theoretical guarantee.

What would settle it

Measure the gradient-bias vector ∇B(xt) after TTT by comparing score predictions with and without the tuned parameters across many prompts; if the bias direction changes substantially across prompts (low cosine similarity between bias vectors), the single-direction cancellation in Proposition 8 fails. Alternatively, empirically test the predicted covariance injection of TDR: if adding zero-mean target noise does not increase gradient covariance as the Jacobian-norm term predicts, the explanation for why TDR prevents prior collapse is wrong.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • If prior collapse is a real failure mode, every test-time-tuning-based editor should include a target-regularization term; the paper shows editing quality degrades as TTT steps grow without it.
  • Contrastive CFG provides a general recipe for removing a fine-tuning-induced bias: use the training condition as a negative prompt during inference.
  • Async-NS shows region-localized editing can be achieved without invasive attention-map surgery, purely by decoupling noise schedules across regions.
  • The gains scale with base model capacity (Wan2.2-5B improved more than Wan2.1-1.3B), suggesting the method unlocks stronger priors rather than merely patching a small model.

Where Pith is reading between the lines

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

  • TDR's geometry-aware covariance argument should apply to any single-example fine-tuning of a generative model—personalization, style transfer, or image editing—so a natural test is whether TDR reduces language drift in image-based DreamBooth-style tuning.
  • The theoretical derivation assumes the TTT-induced bias is a single shared direction across prompts; if the bias direction is prompt-dependent, the cancellation formula would need a per-prompt or learned coefficient rather than a fixed λ2.
  • Async-NS is sharply sensitive to its two noise thresholds (overall score drops from 6.68 to about 4.6 with small deviations), so a self-tuning schedule derived from local mask content would be a more practical extension.
  • The dependency on an external segmentation model for masks could be removed by extracting region masks from the diffusion model's own cross-attention maps during the initial tuning steps, which the paper lists as future work.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper targets a failure mode of test-time tuning (TTT) for text-to-video editing: when a pretrained diffusion model is fine-tuned on a single source video, it can discard the text condition and the spatial latents and merely reconstruct the source. The authors name this 'prior collapse', split it into conditioning collapse and spatial entanglement, and propose ElasticTTT, a three-component pipeline: (1) Target Distribution Regularization (TDR), which adds zero-mean Gaussian noise to the flow-matching regression target while leaving the network input on the original interpolation path, preserving the expected velocity field but inflating the gradient covariance by σ²E[JᵀJ]; (2) Contrastive CFG, a tri-directional guidance that uses the source prompt Csrc as an extra negative to cancel a hypothesised 'source attractor bias'; and (3) Asynchronous Noise Scheduling (Async-NS), which fuses two masked noise schedules with a spatially mixed timestep embedding so that edited and preserved regions follow separate integration trajectories. The manuscript includes derivations of the TDR unbiasedness and covariance injection, a factorized-distribution argument for Contrastive CFG, extensive experiments on Wan2.1-1.3B against eight baselines on a self-selected 125-pair testset, ablations of each component, a 10-participant human study, VLM-human correlation analysis, and a scaling study on Wan2.2-5B.

Significance. The empirical core is solid and useful. The ablation study (Table 4) shows that every component contributes; the main comparison (Table 1) is corroborated by a blinded human study (Table 2) and a head-to-head against the strongest re-implemented TTT baseline (Supp. Fig. 8); and the scaling experiment (Table 3) supports transfer to a larger backbone. The paper is commendably transparent: Supp. B.2 discloses that hyperparameters were chosen on five validation videos; Supp. I lists limitations; Table 6 reports the Async-NS sensitivity in full. The algebraic heart of the TDR analysis — that zero-mean target noise leaves the minimizer of the expected flow-matching objective unchanged (Eqs. 3–5) while adding a Gauss–Newton covariance term (Prop. 3) — is a genuine, parameter-free observation. If the empirical results hold, this is a practical step forward for one-shot video editing. However, two load-bearing points need work: the Contrastive CFG 'exact cancellation' theorem rests on an unverified rank-1/shared-direction ansatz for the TTT bias, and the paper's own Table 6 contradicts the robustness claim for Async-NS. The contribution is therefore better described as a well-engineered and

major comments (3)
  1. [Supp. A.2.1–A.2.3; Eqs. (21)–(29); Sec. 3.3] The cancellation proof for Contrastive CFG rests on Def. 3, which assumes the TTT score residual is one fixed direction ∇B(xt) activated by a scalar α(C)≈sim(C,Csrc) for all prompts. This shared-direction ansatz is asserted, not derived or measured: universal approximation licenses representation, not this factorization, and since the residual is processed through cross-attention conditioned on C, direction independence across prompts is not plausible a priori. Without it, Eq. (29) does not follow, so 'exact cancellation' is unsupported. Moreover, no estimate of α(Ctrg) or β is given; with the chosen λ1=6, λ2=2, cancellation would require α(Ctrg)=α(Csrc)/4, a quantitative relation the paper neither derives nor tests. Finally, the empirical gain is modest (Table 4: OVL 6.52→6.68), consistent with a stronger negative-prompt CFG rather than the claimed mechanism. Please either measure Δs(xt
  2. [Table 6; Supp. B.2; Sec. 4.1] Table 6 shows Async-NS is an isolated operating point: changing (Te,Tp) from (0.97,0.55) to (0.95,0.55) or (0.97,0.60) drops OVL from 6.68 to ≈4.6, and the default was selected on only five validation videos (Supp. B.2). Yet Sec. 4.1 promises a sensitivity analysis 'that verifies the robustness of our method.' The reported data verify the opposite for the component central to the method, so the headline score rests on a narrow island in hyperparameter space. Since Te and Tp directly define the mechanism, this is a reproducibility concern for the central quantitative claim. I recommend a more systematic selection (larger validation grid, per-setting variance on the test set) or a substantial softening of the robustness language in Sec. 4.1 and the abstract.
  3. [Supp. A.1.1–A.1.2; Sec. 3.2] The TDR algebra is correct: Eqs. (3)–(5) and Prop. 2 establish that the expected gradient (and population minimizer) is unchanged by zero-mean target noise, and Prop. 3's covariance decomposition is valid. The conclusions drawn from these identities, however, are stronger than the algebra supports. Prop. 1 infers an 'irreversibly trapped' sharp singularity from Σ→0, but vanishing gradient covariance holds at any converged minimum, sharp or flat. Props. 3–4 plus the SDE heuristic claim that σ²_regE||Jθ||²_F 'violently ejects' the optimizer from memorizing states; this requires the unverified assumptions that sharp memorization implies exploding ||Jθ||_F and that the pretrained prior is flat with small Jacobian. No Jacobian/sharpness measurements are reported, and the 2D toy (Fig. 3) offered as support is never described. Since the empirical TDR evidence (Tables 4, 9; Fig. 9) is convincing
minor comments (6)
  1. [Sec. 3.1, Fig. 3] The 2D toy experiment motivates the paper's central concept but is not reproducible: no architecture, data, or TTT protocol for the toy is given in the main text or Supplementary.
  2. [Table 4 vs. Table 6] The no-TDR configuration is reported as OVL 6.64 in Table 4 but 6.59 in Table 6. Clarify whether these come from different evaluation subsets, seeds, or step counts.
  3. [Supp. G, Listings 2–3] The printed VLM prompts are garbled and self-repetitive (identical preamble lines repeat; 'VLM prompt for Video' is truncated). As printed they are not exactly reproducible.
  4. [Sec. 4.1.2] Please specify whether the re-implemented Tune-A-Video and VidTTA use ElasticTTT's full configuration (λ1=6, Euler/50 steps, 100 TTT steps) or their own optimal settings; running baselines at ElasticTTT's operating point may understate them.
  5. [Sec. 3.4, Eqs. (8)–(9)] The notation for the asynchronous schedules and the fused timestep t^i = M̃⊙t_e^i + (1−M̃)⊙t_p^i is hard to parse; please define the per-step schedule and the fused update more carefully.
  6. [Sec. 4.1.3] The GPT-5 scores are single-run; given VLM stochasticity, please report multiple runs or variance for at least the headline OVL, or note stability.

Circularity Check

0 steps flagged

No significant circularity: the method is evaluated against external baselines, and the one conditional cancellation argument is a free-parameter construction, not a fit called a prediction.

full rationale

Walking the derivation chain, ElasticTTT's components are not equivalent to their inputs. The TDR unbiasedness result (Sec. 3.2, Supp. A.1.2) is self-contained algebra: adding zero-mean noise to the target leaves the expected gradient unchanged and adds a covariance term, so the claim that the global minimum remains aligned with the original vector field is derived, not assumed. The Contrastive CFG analysis (Sec. 3.3, Supp. A.2) is conditional on an explicit factorization pθ*(xt|Csrc) ∝ p_base(xt|Csrc)p_mem(xt) and a rank-1 bias ansatz. Given that ansatz, Eq. (29) shows the residual coefficient is λ1α(Ctrg) − λ2β and states that an appropriate λ2 can zero it. This is a construction with a free hyperparameter, not a parameter fitted to the testset and then reported as a prediction; the ablation (Table 4) and the main comparisons use independent VLM/human scores against external baselines. The only overlapping-author citation ([71]) appears in the introductory context of personalized AI systems and is not load-bearing, and no uniqueness theorem is imported from the authors' prior work. The unverified factorization and the fact that exact cancellation is parameterized by λ2 are correctness/robustness risks, but they do not make the derivation equal to its input. Therefore no circular step is identified.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 2 invented entities

The central claim rests on five hand-picked hyperparameters, an assumed factorization of the post-TTT distribution, and an asserted relationship between memorization and Jacobian norm. None are derived from first principles or validated in isolation.

free parameters (5)
  • σ_reg (TDR noise variance) = 0.2
    Chosen empirically on a 5-video validation set (Supp. B.2). Sensitivity shown in Table 6: OVL varies <0.05 across 0.1–0.3.
  • λ1 (standard CFG scale) = 6
    Chosen empirically; sensitivity band OVL 6.63–6.75 across tested combinations.
  • λ2 (contrastive CFG scale) = 2
    Chosen empirically; no independent prediction of its value from the theory.
  • T_e (edited-region noise level) = 0.97
    Chosen empirically; Async-NS is highly sensitive to this value (Table 6: OVL 4.59–6.68 across nearby settings).
  • T_p (preserved-region noise level) = 0.55
    Chosen empirically; critical for preserving unedited regions. Nearby values (0.50, 0.60) cause large OVL drops.
axioms (6)
  • standard math Flow-matching objective with linear interpolation path and L2 velocity loss (Eqs. 1–3)
    Standard formulation, not proven in the paper.
  • ad hoc to paper Factorized post-TTT distribution pθ*(xt|Csrc) = (1/Z) p_base(xt|Csrc)·p_mem(xt) with p_mem approaching Dirac
    Supplementary A.2.1, Eq. (21); central to the Contrastive CFG analysis but not established for actual networks.
  • ad hoc to paper Condition-dependent activation α(C) ≈ sim(C,Csrc) of the source attractor bias
    Supplementary A.2.2, Proposition 6; no empirical verification beyond the claim.
  • domain assumption SDE model of SGD: dθ = −ηḡ dt + √(ηΣ) dW
    Supplementary A.1.1; standard approximation, but not exact for AdamW/LoRA.
  • ad hoc to paper Sharp memorization implies exploding Jacobian norm ||J_θ||_F
    Supplementary A.1.2; asserted without proof.
  • ad hoc to paper Pretrained prior corresponds to a flat minimum with small Jacobian norm
    Supplementary A.1.2; required for TDR to drive optimization back to the base prior.
invented entities (2)
  • Prior collapse no independent evidence
    purpose: Label for the degenerate state where TTT causes conditioning collapse and spatial entanglement, motivating the framework.
    Demonstrated only via toy experiments and ablations within this paper; no external validation.
  • Source attractor bias ∇B(xt) no independent evidence
    purpose: Mathematical device in the Contrastive CFG analysis to model TTT-induced distribution shift.
    Introduced in Supplementary A.2.1 as the gradient of log p_mem; not directly measured.

pith-pipeline@v1.3.0-alltime-deepseek · 26273 in / 16180 out tokens · 157562 ms · 2026-08-03T01:48:20.187574+00:00 · methodology

0 comments
read the original abstract

Test-Time Tuning (TTT) on pretrained diffusion models has emerged as a powerful paradigm for video editing. However, there exists a foundational mismatch between the distribution-mapping nature of generative models and the single-point optimization of standard TTT. In this paper, we demonstrate that this mismatch triggers \textit{Prior Collapse}, a degenerate state where the model discards the text conditions and spatial latents, collapsing generations to the source video, or entangling the features of distinct regions. To resolve this, we propose \textbf{ElasticTTT}, a novel framework that preserves the prior generative distribution and rescues generative elasticity. Specifically, we propose \textit{Target Distribution Regularization} to prevent sharp memorization minima, \textit{Contrastive CFG} to guide inference away from source biases, and \textit{Asynchronous Noise Schedule} to preserve unedited regions. Extensive evaluations, supported by theoretical analysis, demonstrate that ElasticTTT successfully preserves the generative prior of the base model, achieving state-of-the-art performance on one-shot video editing.

Review history (2 revisions) →

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

81 extracted references · 28 linked inside Pith

  1. [1]

    J. Bai, T. He, Y. Wang, J. Guo, H. Hu, Z. Liu, and J. Bian. Uniedit: A unified tuning-free framework for video motion and appearance editing, 2025

    2025

  2. [2]

    Q. Bai, Q. Wang, H. Ouyang, Y. Yu, H. Wang, W. Wang, K. L. Cheng, S. Ma, Y. Zeng, Z. Liu, Tsinghua University | 14 Y. Xu, Y. Shen, and Q. Chen. Scaling instruction-based video editing with a high-quality synthetic dataset.arXiv preprint arXiv:2510.15742, 2025

    arXiv 2025

  3. [3]

    S. Bai, Y. Cai, R. Chen, K. Chen, X. Chen, Z. Cheng, L. Deng, W. Ding, C. Gao, C. Ge, W. Ge, Z. Guo, Q. Huang, J. Huang, F. Huang, B. Hui, S. Jiang, Z. Li, M. Li, M. Li, K. Li, Z. Lin, J. Lin, X. Liu, J. Liu, C. Liu, Y. Liu, D. Liu, S. Liu, D. Lu, R. Luo, C. Lv, R. Men, L. Meng, X. Ren, X. Ren, S. Song, Y. Sun, J. Tang, J. Tu, J. Wan, P. Wang, P. Wang, Q....

    Pith/arXiv arXiv 2025

  4. [4]

    X. Bi, J. Lu, B. Liu, X. Cun, Y. Zhang, W. Li, and B. Xiao. Customttt: Motion and appearance customized video generation via test-time training. InProceedings of the AAAI Conference on Artificial Intelligence, volume 39, pages 1871–1879, 2025

    2025

  5. [5]

    Y. Bian, Z. Zhang, X. Ju, M. Cao, L. Xie, Y. Shan, and Q. Xu. Videopainter: Any-length video inpainting and editing with plug-and-play context control, 2025

    2025

  6. [6]

    bingshuai liu, A. Wang, Z. Min, C. Lyu, L. Wang, and J. Su. VEBench: Towards comprehensive and automatic evaluation for text-guided video editing, 2025

    2025

  7. [7]

    Caelles, A

    S. Caelles, A. Montes, K.-K. Maninis, Y. Chen, L. Van Gool, F. Perazzi, and J. Pont-Tuset. The 2018 davis challenge on video object segmentation.arXiv preprint arXiv:1803.00557, 2018

    Pith/arXiv arXiv 2018

  8. [8]

    Chung, S

    J. Chung, S. Hyun, and J.-P. Heo. Style injection in diffusion: A training-free approach for adapting large-scale diffusion models for style transfer. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8795–8805, 2024

    2024

  9. [9]

    Cohen.Statistical power analysis for the behavioral sciences (2nd ed.Statistical power analysis for the behavioral sciences, 1988

    J. Cohen.Statistical power analysis for the behavioral sciences (2nd ed.Statistical power analysis for the behavioral sciences, 1988

    1988

  10. [10]

    Cohen, V

    N. Cohen, V. Kulikov, M. Kleiner, I. Huberman-Spiegelglas, and T. Michaeli. Slicedit: Zero-shot video editing with text-to-image diffusion models using spatio-temporal slices.arXiv preprint arXiv:2405.12211, 2024

    Pith/arXiv arXiv 2024

  11. [11]

    Foret, A

    P. Foret, A. Kleiner, H. Mobahi, and B. Neyshabur. Sharpness-aware minimization for efficiently improving generalization.arXiv preprint arXiv:2010.01412, 2020

    Pith/arXiv arXiv 2010

  12. [12]

    R. Gal, Y. Alaluf, Y. Atzmon, O. Patashnik, A. H. Bermano, G. Chechik, and D. Cohen-Or. An image is worth one word: Personalizing text-to-image generation using textual inversion. arXiv preprint arXiv:2208.01618, 2022

    Pith/arXiv arXiv 2022

  13. [13]

    C. Gao, L. Ding, X. Cai, Z. Huang, Z. Wang, and T. Xue. Lora-edit: Controllable first- frame-guided video editing via mask-aware lora fine-tuning.arXiv preprint arXiv:2506.10082, 2025

    arXiv 2025

  14. [14]

    C. F. Gauss.Theoria motus corporum coelestium in sectionibus conicis solem ambientium, volume 7. FA Perthes, 1877

  15. [15]

    Geyer, O

    M. Geyer, O. Bar-Tal, S. Bagon, and T. Dekel. Tokenflow: Consistent diffusion features for consistent video editing. InThe Twelfth International Conference on Learning Representations, 2024. Tsinghua University | 15

    2024

  16. [16]

    S. S. Harsha, A. Revanur, D. Agarwal, and S. Agrawal. Genvideo: One-shot target-image and shape aware video editing using t2i diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pages 7559–7568, June 2024

    2024

  17. [17]

    H. He, J. Wang, J. Zhang, Z. Xue, X. Bu, Q. Yang, S. Wen, and L. Xie. Openve-3m: A large-scale high-quality dataset for instruction-guided video editing.arXiv preprint arXiv:2512.07826, 2025

    arXiv 2025

  18. [18]

    J. Ho, A. Jain, and P. Abbeel. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    2020

  19. [19]

    Ho and T

    J. Ho and T. Salimans. Classifier-free diffusion guidance.arXiv preprint arXiv:2207.12598, 2022

    Pith/arXiv arXiv 2022

  20. [20]

    Hochreiter and J

    S. Hochreiter and J. Schmidhuber. Flat minima.Neural computation, 9(1):1–42, 1997

    1997

  21. [21]

    W. Hong, M. Ding, W. Zheng, X. Liu, and J. Tang. Cogvideo: Large-scale pretraining for text-to-video generation via transformers.arXiv preprint arXiv:2205.15868, 2022

    Pith/arXiv arXiv 2022

  22. [22]

    Hotelling

    H. Hotelling. Analysis of a complex of statistical variables into principal components.Journal of educational psychology, 24(6):417, 1933

    1933

  23. [23]

    E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, W. Chen, et al. Lora: Low-rank adaptation of large language models.Iclr, 1(2):3, 2022

    2022

  24. [24]

    Huang, W

    Y. Huang, W. Xiong, H. Zhang, C. Chen, J. Liu, M. Yan, and S. Chen. Dive: Taming dino for subject-driven video editing. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 16004–16014, 2025

    2025

  25. [25]

    Huang, Y

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

    2024

  26. [26]

    Jeong and J

    H. Jeong and J. C. Ye. Ground-a-video: Zero-shot grounded video editing using text-to-image diffusion models. InThe Twelfth International Conference on Learning Representations, 2024

    2024

  27. [27]

    Kahatapitiya, A

    K. Kahatapitiya, A. Karjauv, D. Abati, F. Porikli, Y. M. Asano, and A. Habibian. Object-centric diffusion for efficient video editing. InEuropean Conference on Computer Vision, pages 91–108. Springer, 2024

    2024

  28. [28]

    N. S. Keskar, D. Mudigere, J. Nocedal, M. Smelyanskiy, and P. T. P. Tang. On large-batch train- ing for deep learning: Generalization gap and sharp minima.arXiv preprint arXiv:1609.04836, 2016

    Pith/arXiv arXiv 2016

  29. [29]

    J. Kim, Y. Hong, J. Park, and J. C. Ye. Flowalign: Trajectory-regularized, inversion-free flow-based image editing.arXiv preprint arXiv:2505.23145, 2025

    Pith/arXiv arXiv 2025

  30. [30]

    J. Kim, Y. Hong, J. Park, and J. C. Ye. Flowalign: Trajectory-regularized, inversion-free flow- based image editing. InThe Fourteenth International Conference on Learning Representations, 2026. Tsinghua University | 16

    2026

  31. [31]

    D. P. Kingma and J. Ba. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980, 2014

    Pith/arXiv arXiv 2014

  32. [32]

    W. Kong, Q. Tian, Z. Zhang, R. Min, Z. Dai, J. Zhou, J. Xiong, X. Li, B. Wu, J. Zhang, et al. Hunyuanvideo: A systematic framework for large video generative models.arXiv preprint arXiv:2412.03603, 2024

    Pith/arXiv arXiv 2024

  33. [33]

    M. Ku, C. Wei, W. Ren, H. Yang, and W. Chen. Anyv2v: A tuning-free framework for any video-to-video editing tasks.Transactions on Machine Learning Research, 2024. Reproducibility Certification

    2024

  34. [34]

    Kulikov, M

    V. Kulikov, M. Kleiner, I. Huberman-Spiegelglas, and T. Michaeli. Flowedit: Inversion-free text-based editing using pre-trained flow models. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 19721–19730, 2025

    2025

  35. [35]

    Kumari, B

    N. Kumari, B. Zhang, R. Zhang, E. Shechtman, and J.-Y. Zhu. Multi-concept customization of text-to-image diffusion. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1931–1941, 2023

    1931

  36. [36]

    G. Kwon, J. Park, and J. C. Ye. Unified editing of panorama, 3d scenes, and videos through disentangled self-attention injection.arXiv preprint arXiv:2405.16823, 2024

    Pith/arXiv arXiv 2024

  37. [37]

    LeCun, S

    Y. LeCun, S. Chopra, R. Hadsell, M. Ranzato, F. Huang, et al. A tutorial on energy-based learning.Predicting structured data, 1(0), 2006

    2006

  38. [38]

    M. Li, L. Lin, Y. Liu, Y. Zhu, and Y. Li. Qffusion: Controllable portrait video editing via quadrant-grid attention learning.IEEE Transactions on Visualization and Computer Graphics, 2025

    2025

  39. [39]

    Liang, A

    F. Liang, A. Kodaira, C. Xu, M. Tomizuka, K. Keutzer, and D. Marculescu. Looking backward: Streaming video-to-video translation with feature banks. InThe Thirteenth International Conference on Learning Representations, 2025

    2025

  40. [40]

    Liang, B

    F. Liang, B. Wu, J. Wang, L. Yu, K. Li, Y. Zhao, I. Misra, J.-B. Huang, P. Zhang, P. Vajda, et al. Flowvid: Taming imperfect optical flows for consistent video-to-video synthesis. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8207–8216, 2024

    2024

  41. [41]

    Lipman, R

    Y. Lipman, R. T. Q. Chen, H. Ben-Hamu, M. Nickel, and M. Le. Flow matching for generative modeling. InThe Eleventh International Conference on Learning Representations, 2023

    2023

  42. [42]

    C. Liu, R. Li, K. Zhang, Y. Lan, and D. Liu. Stablev2v: Stabilizing shape consistency in video-to-video editing.IEEE Transactions on Circuits and Systems for Video Technology, 2025

    2025

  43. [43]

    S. Liu, Y. Zhang, W. Li, Z. Lin, and J. Jia. Video-p2p: Video editing with cross-attention control. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8599–8608, 2024

    2024

  44. [44]

    Loshchilov and F

    I. Loshchilov and F. Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    Pith/arXiv arXiv 2017

  45. [45]

    N. Ma, S. Tong, H. Jia, H. Hu, Y.-C. Su, M. Zhang, X. Yang, Y. Li, T. Jaakkola, X. Jia, et al. Inference-time scaling for diffusion models beyond scaling denoising steps.arXiv preprint arXiv:2501.09732, 2025. Tsinghua University | 17

    Pith/arXiv arXiv 2025

  46. [46]

    Mandt, M

    S. Mandt, M. D. Hoffman, and D. M. Blei. Stochastic gradient descent as approximate bayesian inference.Journal of Machine Learning Research, 18(134):1–35, 2017

    2017

  47. [47]

    C. Meng, Y. He, Y. Song, J. Song, J. Wu, J.-Y. Zhu, and S. Ermon. Sdedit: Guided image synthesis and editing with stochastic differential equations.arXiv preprint arXiv:2108.01073, 2021

    Pith/arXiv arXiv 2021

  48. [48]

    T. H. S. Meral, H. Yesiltepe, C. Dunlop, and P. Yanardag. Motionflow: Attention-driven motion transfer in video diffusion models.arXiv preprint arXiv:2412.05275, 2024

    Pith/arXiv arXiv 2024

  49. [49]

    Mokady, A

    R. Mokady, A. Hertz, K. Aberman, Y. Pritch, and D. Cohen-Or. Null-text inversion for editing real images using guided diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6038–6047, 2023

    2023

  50. [50]

    C. Mou, M. Cao, X. Wang, Z. Zhang, Y. Shan, and J. Zhang. Revideo: Remake a video with motionandcontentcontrol.Advances in Neural Information Processing Systems, 37:18481–18505, 2024

    2024

  51. [51]

    Gpt-5 system card, 2025

    OpenAI. Gpt-5 system card, 2025

    2025

  52. [52]

    Ouyang, Y

    W. Ouyang, Y. Dong, L. Yang, J. Si, and X. Pan. I2vedit: First-frame-guided video editing via image-to-video diffusion models.CoRR, abs/2405.16537, 2024

    Pith/arXiv arXiv 2024

  53. [53]

    Peebles and S

    W. Peebles and S. Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF international conference on computer vision, pages 4195–4205, 2023

    2023

  54. [54]

    Perazzi, J

    F. Perazzi, J. Pont-Tuset, B. McWilliams, L. Van Gool, M. Gross, and A. Sorkine-Hornung. A benchmark dataset and evaluation methodology for video object segmentation. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 724–732, 2016

    2016

  55. [55]

    Pont-Tuset, F

    J. Pont-Tuset, F. Perazzi, S. Caelles, P. Arbeláez, A. Sorkine-Hornung, and L. Van Gool. The 2017 davis challenge on video object segmentation.arXiv preprint arXiv:1704.00675, 2017

    Pith/arXiv arXiv 2017

  56. [56]

    Radford, J

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, et al. Learning transferable visual models from natural language supervision. InInternational conference on machine learning, pages 8748–8763. PmLR, 2021

    2021

  57. [57]

    N. Ravi, V. Gabeur, Y.-T. Hu, R. Hu, C. Ryali, T. Ma, H. Khedr, R. Rädle, C. Rolland, L. Gustafson, E. Mintun, J. Pan, K. V. Alwala, N. Carion, C.-Y. Wu, R. Girshick, P. Dol- lár, and C. Feichtenhofer. Sam 2: Segment anything in images and videos.arXiv preprint arXiv:2408.00714, 2024

    Pith/arXiv arXiv 2024

  58. [58]

    N. Ruiz, Y. Li, V. Jampani, Y. Pritch, M. Rubinstein, and K. Aberman. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 22500–22510, 2023

    2023

  59. [59]

    K. Sohn, N. Ruiz, K. Lee, D. C. Chin, I. Blok, H. Chang, J. Barber, L. Jiang, G. Entis, Y. Li, et al. Styledrop: Text-to-image generation in any style.arXiv preprint arXiv:2306.00983, 2023

    Pith/arXiv arXiv 2023

  60. [60]

    J. Song, C. Meng, and S. Ermon. Denoising diffusion implicit models.arXiv preprint arXiv:2010.02502, 2020

    Pith/arXiv arXiv 2010

  61. [61]

    Song and S

    Y. Song and S. Ermon. Generative modeling by estimating gradients of the data distribution. Advances in neural information processing systems, 32, 2019. Tsinghua University | 18

    2019

  62. [62]

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

    Pith/arXiv arXiv 2011

  63. [63]

    Y. Sun, X. Wang, Z. Liu, J. Miller, A. Efros, and M. Hardt. Test-time training with self- supervision for generalization under distribution shifts. InInternational conference on machine learning, pages 9229–9248. PMLR, 2020

    2020

  64. [64]

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

    2025

  65. [65]

    D. Wang, E. Shelhamer, S. Liu, B. Olshausen, and T. Darrell. Tent: Fully test-time adaptation by entropy minimization.arXiv preprint arXiv:2006.10726, 2020

    Pith/arXiv arXiv 2006

  66. [66]

    J. Wang, Y. Chen, Y. He, X. Song, Y. Xin, D. Zhang, Z. Wan, B. Li, and R. Zhang. Low-cost test-time adaptation for robust video editing.arXiv preprint arXiv:2507.21858, 2025

    Pith/arXiv arXiv 2025

  67. [67]

    Wang and J

    Z. Wang and J. Tang. T3v2v: Test time training for domain adaptation in video-to-video editing. InCVPR 2025 Workshop, AI for Creative Visual Content Generation Editing and Understanding

    2025

  68. [68]

    Z. Wang, L. Zhao, and W. Xing. Stylediffusion: Controllable disentangled style transfer via diffusion models. InProceedings of the IEEE/CVF international conference on computer vision, pages 7677–7689, 2023

    2023

  69. [69]

    J. Z. Wu, Y. Ge, X. Wang, S. W. Lei, Y. Gu, Y. Shi, W. Hsu, Y. Shan, X. Qie, and M. Z. Shou. Tune-a-video: One-shot tuning of image diffusion models for text-to-video generation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 7623–7633, 2023

    2023

  70. [70]

    Y. Wu, L. Chen, R. Li, S. Wang, C. Xie, and L. Zhang. Insvie-1m: Effective instruction-based video editing with elaborate dataset construction. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 16692–16701, 2025

    2025

  71. [71]

    Y. Wu, S. Zhang, W. H. Ching, C. Zhang, and M. Liu. Pipbench: A profile-inclusive framework for personalized image generation evaluation.arXiv preprint arXiv:2607.06440, 2026

    Pith/arXiv arXiv 2026

  72. [72]

    S. Yang, Z. Gu, L. Hou, X. Tao, P. Wan, X. Chen, and J. Liao. Mtv-inpaint: Multi-task long video inpainting.CoRR, abs/2503.11412, March 2025

    Pith/arXiv arXiv 2025

  73. [73]

    X. Yang, L. Zhu, H. Fan, and Y. Yang. Eva: Zero-shot accurate attributes and multi-object video editing.CoRR, abs/2403.16111, 2024

    Pith/arXiv arXiv 2024

  74. [74]

    X. Yang, L. Zhu, H. Fan, and Y. Yang. Videograin: Modulating space-time attention for multi- grained video editing. InThe Thirteenth International Conference on Learning Representations, 2025. Tsinghua University | 19

    2025

  75. [75]

    Z. Yang, J. Teng, W. Zheng, M. Ding, S. Huang, J. Xu, Y. Yang, W. Hong, X. Zhang, G. Feng, D. Yin, Yuxuan.Zhang, W. Wang, Y. Cheng, B. Xu, X. Gu, Y. Dong, and J. Tang. Cogvideox: Text-to-video diffusion models with an expert transformer. InThe Thirteenth International Conference on Learning Representations, 2025

    2025

  76. [76]

    S. Yu, D. Liu, Z. Ma, Y. Hong, Y. Zhou, H. Tan, J. Chai, and M. Bansal. Veggie: Instructional editing and reasoning video concepts with grounded generation. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 15147–15158, 2025

    2025

  77. [77]

    Zhang, Z

    C. Zhang, Z. Chen, K. Zheng, and J. Zhu. Voicebridge: Designing latent bridge models for general speech restoration at scale.arXiv preprint arXiv:2509.25275, 2025

    arXiv 2025

  78. [78]

    Zhang, N

    Y. Zhang, N. Huang, F. Tang, H. Huang, C. Ma, W. Dong, and C. Xu. Inversion-based style transfer with diffusion models. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10146–10156, 2023

    2023

  79. [79]

    Zi and et al

    B. Zi and et al. Señorita-2m: A high-quality instruction-based dataset for general video editing by video specialists.NeurIPS, 2026

    2026

  80. [80]

    B. Zi, P. Ruan, M. Chen, X. Qi, S. Hao, S. Zhao, Y. Huang, B. Liang, R. Xiao, and K.-F. Wong. Se\˜ norita-2m: A high-quality instruction-based dataset for general video editing by video specialists.arXiv preprint arXiv:2502.06734, 2025

    Pith/arXiv arXiv 2025

Showing first 80 references.