CARVE: Concept Avoidance via Reward-shaped Visual Erasure
This work treats diffusion-based image generation as a sequential decision process and uses Diffusion Denoising Policy Optimization (DDPO) to learn a policy that steers generation away from a chosen target concept - in this case, "teddy bear."
Chris Stanulet, F. Ibanez, Febie Lin
· 0 citations