Research map / Image generation / Text-to-image diffusion
Circuit Mechanisms for Spatial Relation Generation in Diffusion Transformers
Binxu Wang, Jingxuan Fan, Xu Pan
This paper tried a common change: Add positional encoding
add positional encoding to random token encoder · Synthetic relations · binding and relation accuracy Improved
What 646 accepted papers report for this change
- 525 improved
- 58 no clear effect
- 7 mixed
- 56 hurt
- Learning-Augmented Scalable Linear Assignment Problem Optimization via Neural Dual Warm-Starts ICML 2026 · add positional encodings to features (d=13 -> d=21) · Block-structured, N=4096 · speedupImproved
- Mesh Based Simulations with Spatial and Temporal awareness ICML 2026 · use learned positional embedding, learnable relative bias or weighted adjacency instead of RoPE · Aneurysm · RMSEHurt
- Sparse ActionGen: Accelerating Diffusion Policy with Real-time Pruning ICML 2026 · add sinusoidal positional encoding to the pruner · Robomimic Tool Hang · success rate (%): 28 -> 54Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| ablate relation cross-attention head L2H8 (ablation) | RTE-DiT-B · Synthetic relations | relation accuracy: 67% -> 33% when removed | Improved |
| ablate object head L4H3 (ablation) | RTE-DiT-B · Synthetic relations | shape accuracy: 90% -> 76% when removed | Improved |
| mask relation words in prompt (ablation) | T5-DiT-B · Synthetic relations | relation accuracy | No clear effect |
| mask shape2 text token (ablation) | T5-DiT-B · Synthetic relations | shape/binding/relation accuracy: about 50% drop when masked | Improved |
| add positional encoding to random token encoder (ablation) | RTE-DiT-B · Synthetic relations | binding and relation accuracy | Improved |
| use smallest model (DiT-nano) with RTE (ablation) | RTE-DiT-nano · Synthetic relations | spatial relation accuracy: 5% | Hurt |
| replace random token embeddings with T5 embeddings | DiT-B · Synthetic relations | in-distribution relation accuracy | No clear effect |
| insert filler word or reverse sentence in prompt | T5-DiT vs RTE-DiT · Synthetic relations | relation accuracy | Hurt |
| vector arithmetic on shape2 T5 embedding to swap relation vector (ablation) | T5-DiT · Synthetic relations | generated object position | Improved |
Closest papers by meaning
- Relation Rectification in Diffusion Model (CVPR 2024 · similarity 0.89)
- DifFRACT: Diffusion Feature Reconstruction and Attribution for Circuit Tracing (NeurIPS 2026 · similarity 0.89)
- GrounDiT: Grounding Diffusion Transformers via Noisy Patch Transplantation (NeurIPS 2024 · similarity 0.88)
- Rare Text Semantics Were Always There in Your Diffusion Transformer (NeurIPS 2025 · similarity 0.88)
- Spatial Transport Optimization by Repositioning Attention Map for Training-Free Text-to-Image Synthesis (CVPR 2025 · similarity 0.87)
- R&B: Region and Boundary Aware Zero-shot Grounded Text-to-image Generation (ICLR 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
