Mechanical Turk

[Mechanical Turk @MechMathTurk]

@BrianCChao Thanks, Brian

[Mechanical Turk @MechMathTurk]

@BrianCChao Could you please elaborate? Does A.3 have any proof that one should rely on the distance between velocity and noise (in freq. domain) to define the proper transition time? And not the distance between spectrum of x_t and noise, for example?

[Mechanical Turk @MechMathTurk]

@howard_xhc @KumbongHermann @GordonWetzstein there is some evidence that this has been tried https://t.co/wk0KOwbHpv But indeed, not adopted massively

[Mechanical Turk @MechMathTurk]

@BrianCChao I wonder what is the impact of relying on velocity vs input? Does it matter in practice? I'd be glad to know your opinion

[Mechanical Turk @MechMathTurk]

@BrianCChao Your derivation from Sec. 4.2 with the Assumption A1 is indeed remarkable! I have a question about it: choosing the transition time based on noise-dominated *velocity* is a design choice. In SwD work the authors did the same analysis but relied on noise-dominated *input* instead

[Mechanical Turk @MechMathTurk]

@GordonWetzstein Interestingly, in PyramidalWan paper, a similar approach of injecting noise into high frequencies is described. Before that, PyramidFlow used it implicitly with Haar wavelet
https://t.co/UtMg2Cz9SS
https://t.co/GWi2LK9mHv

[Mechanical Turk @MechMathTurk]

Interestingly, papers from @NeurIPSConf 2025 have incorrect bibtex on openreview: they show the year of 2026, not 2025.

[Mechanical Turk @MechMathTurk]

@yoshitomo_cs @CVPR Thanks! Emailed.

[Mechanical Turk @MechMathTurk]

@jm_alexia @oguzhannercan We've tried JEDi in our Mobile Video Diffusion project https://t.co/Bdq4FhoVII (see Appendix) and it indeed showed itself more aligned with the process of model pruning than FVD. Thanks for your research!

[Mechanical Turk @MechMathTurk]

@A_K_Nain @_akhaliq

[Mechanical Turk @MechMathTurk]

@LinBin46984 Amazing work! Many thanks for sharing it with the community!
May I ask about the video generation evaluation? Has Latte-L been initialized with the same PixArt-α despite the latent space had changed? Or have pretrained weights for the UCF and SkyTimelapse datasets been used?

[Mechanical Turk @MechMathTurk]

@jiang_zhengkai @_akhaliq Hey, congrats! Cool work! Do you have any plans to release the supplementary materials mentioned in the text?

[Mechanical Turk @MechMathTurk]

@_akhaliq Am I missing something or have the authors indeed forgotten to include the assessment of visual results in their manuscript?

[Mechanical Turk @MechMathTurk]

@jm_alexia To be honest, I do not really understand what the difference of the proposed CMMD vs well-known KID is (except for the different kernel used). The authors mention KID in the Related Works section but do not evaluate against it

[Mechanical Turk @MechMathTurk]

@AmazonHelp Sure, I have checked it, with no result

[Mechanical Turk @MechMathTurk]

Hello @AmazonHelp, my Amazon DE account is currently on hold, and I cannot log in on the website/in the app. I received an email from account-resolution(AT)amazon(DOT)de and replied with the requested document, but after 4 days still haven't received any response

[Mechanical Turk @MechMathTurk]

@_akhaliq have you stopped updating Sigmoid(dot)social?