Push-Pull Attentional Anchoring for Diffusion Concept Erasure

Key points

• Rapid advances in diffusion models have raised concerns about privacy, harmful content, and copyright violations.1
• Concept erasure methods aim to address these risks by removing undesirable concepts from pretrained models without full retraining.1
• By using a relative similarity ratio instead of an absolute difference, our adaptive threshold represents a meaningful percentage of semantic retention, facilitating a single value to be applied uniformly across layers.1
• Our method achieves favorable trade-offs against prior approaches on Stable Diffusion v1.4 and demonstrates its applicability to modern architectures such as FLUX.1

Sources (1)

• [1]Push-Pull Attentional Anchoring for Diffusion Concept Erasure
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 06:38 AM

  To address this, we propose Push-Pull Attentional Anchoring (PPAA), a mechanism in the cross-attention space that displaces target representations while anchoring the scale of erasure via bounded cosine-similarity constraints.

  Rapid advances in diffusion models have raised concerns about privacy, harmful content, and copyright violations.

Extractive summary: sentences quoted from the sources.