AnchorDiff: Training-Free Concept Grounding for MM-DiTs via Anchor-Based Graph Propagation 事件

AnchorDiff: Training-Free Concept Grounding for MM-DiTs via Anchor-Based Graph Propagation arXiv:2605.26460v1 Announce Type: new Abstract: Multi-Modal Diffusion Transformers (MM-DiTs) encode rich representations for training-free concept grounding, but existing attention-based methods often produce overlapping activations on visually confusable concepts, a failure mode we call concept leakage, where target responses spill over to non-target objects. To address this issue, we propose AnchorDiff,