AnchorDiff: Training-Free Concept Grounding for MM-DiTs via Anchor-Based Graph Propagation
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917533641080832 |
|---|---|
| author | Zhang, Jian Zhang, Zhijun |
| author_facet | Zhang, Jian Zhang, Zhijun |
| contents | 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, a training-free grounding method that decouples semantic localization from structural refinement. AnchorDiff selects a high-confidence anchor from concept-to-image attention map and propagates it as a one-hot seed over a hybrid graph derived from image-to-image self-attention. The graph uses output-space similarity for dense within-object propagation and a row-wise attention gate to suppress cross-object connections. Additionally, we introduce the Multi-Concept Confusion Dataset, which contains images with multiple visually similar concepts and separate masks, enabling explicit evaluation of concept leakage. Experiments show that AnchorDiff achieves strong grounding performance on ImageNet-Segmentation and PascalVOC, while substantially reducing concept leakage on our Multi-Concept Confusion Dataset. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_26460 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | AnchorDiff: Training-Free Concept Grounding for MM-DiTs via Anchor-Based Graph Propagation Zhang, Jian Zhang, Zhijun Computer Vision and Pattern Recognition Artificial Intelligence 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, a training-free grounding method that decouples semantic localization from structural refinement. AnchorDiff selects a high-confidence anchor from concept-to-image attention map and propagates it as a one-hot seed over a hybrid graph derived from image-to-image self-attention. The graph uses output-space similarity for dense within-object propagation and a row-wise attention gate to suppress cross-object connections. Additionally, we introduce the Multi-Concept Confusion Dataset, which contains images with multiple visually similar concepts and separate masks, enabling explicit evaluation of concept leakage. Experiments show that AnchorDiff achieves strong grounding performance on ImageNet-Segmentation and PascalVOC, while substantially reducing concept leakage on our Multi-Concept Confusion Dataset. |
| title | AnchorDiff: Training-Free Concept Grounding for MM-DiTs via Anchor-Based Graph Propagation |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2605.26460 |