Certain Gaussian initial noises act as winning tickets that bias motion diffusion toward specific semantics; retrieving and KL-refining them improves text-motion alignment without retraining.
In: CVPR (2023)
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
fields
cs.CV 3years
2026 3representative citing papers
HO-Flow synthesizes realistic hand-object motions from text and canonical 3D objects via an interaction-aware VAE and masked flow matching, reporting SOTA physical plausibility and diversity on GRAB, OakInk, and DexYCB.
MoCHA canonicalizes captions to motion-recoverable semantics before contrastive training, cutting within-motion embedding variance by 11-19% and lifting T2M R@1 by 3.1pp on HumanML3D and 10.3pp on KIT-ML.
citing papers explorer
-
Retrieving and Refining Winning Noise Tickets for Diffusion-Based Motion Generation
Certain Gaussian initial noises act as winning tickets that bias motion diffusion toward specific semantics; retrieving and KL-refining them improves text-motion alignment without retraining.
-
HO-Flow: Generalizable Hand-Object Interaction Generation with Latent Flow Matching
HO-Flow synthesizes realistic hand-object motions from text and canonical 3D objects via an interaction-aware VAE and masked flow matching, reporting SOTA physical plausibility and diversity on GRAB, OakInk, and DexYCB.
-
MoCHA: Denoising Caption Supervision for Motion-Text Retrieval
MoCHA canonicalizes captions to motion-recoverable semantics before contrastive training, cutting within-motion embedding variance by 11-19% and lifting T2M R@1 by 3.1pp on HumanML3D and 10.3pp on KIT-ML.