FlowErase-RL is the first GRPO-based reward optimization framework for concept erasure in flow matching models, using a dynamic dual-path reward mechanism to suppress target concepts while preserving generative quality.
To generate or not? safety-driven unlearned diffusion models are still easy to generate unsafe images
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
FlowErase-RL: Rethinking Concept Erasure as Reward Optimization in Flow Matching Models
FlowErase-RL is the first GRPO-based reward optimization framework for concept erasure in flow matching models, using a dynamic dual-path reward mechanism to suppress target concepts while preserving generative quality.