Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
October 20, 2025Open Access

T2VUnlearning: A Concept Erasing Method for Text-to-Video Diffusion Models

View Full Paper
Ask AI
Bookmark
Share

Authors

XYXuekui YeSCSusan ChengYWYongtao Wang

Discussion

Loading...

Member takes

Overview

This method demonstrates effective concept erasing in text-to-video models, suggesting improvements in controlling generated content.

Key Points

  • Our method effectively erases specific concepts in text-to-video models while maintaining overall generation capabilities.
  • Using a combination of negative velocity prediction and prompt augmentation, we ensure robust performance against refined prompts.
  • Incorporating mask-based localization and concept preservation regularizations further enhances precise unlearning.
  • Extensive experiments show our approach outperforms existing solutions in managing explicit or harmful video content.

Cite This Study

Ye et al. (2025) studied this question.

synapsesocial.com/papers/68f6196ee0bbbc94fac361f0https://doi.org/10.48550/arxiv.2505.17550
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Unlearning Concepts from Text-to-Video Diffusion Models2024
  2. 2Defensive Unlearning with Adversarial Training for Robust Concept Erasure in Diffusion Models2024 · 4 citations
  3. 3Erasing Concepts from Text-to-Image Diffusion Models with Few-shot Unlearning2024 · 2 citations
  4. 4Unlearning Concepts in Diffusion Model via Concept Domain Correction and Concept Preserving Gradient2024 · 3 citations
  5. 5Concept Unlearning by Modeling Key Steps of Diffusion Process2025