AINeutralarXiv – CS AI · 15h ago6/10
🧠
AnchorDiff: Training-Free Concept Grounding for MM-DiTs via Anchor-Based Graph Propagation
Researchers propose AnchorDiff, a training-free method for improving concept grounding in Multi-Modal Diffusion Transformers by addressing 'concept leakage' where attention activations overlap on visually similar objects. The approach uses anchor-based graph propagation to better localize and distinguish between confusable concepts, with evaluation on a newly introduced Multi-Concept Confusion Dataset.