CSVC optimizes text prompts via vision-language-model feedback to steer frozen video diffusion editors toward causally consistent facial counterfactuals such as aging, gender change, beard addition, and baldness.
young" should be replaced with a more precise age or age range to guide the model more effectively. For instance, specifying
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Causally Steered Diffusion for Automated Video Counterfactual Generation
CSVC optimizes text prompts via vision-language-model feedback to steer frozen video diffusion editors toward causally consistent facial counterfactuals such as aging, gender change, beard addition, and baldness.