Research map / Image generation / Editing and style transfer
Anchor Token Matching: Implicit Structure Locking for Training-free AR Image Editing
Taihang Hu, Linxuan Li, Kai Wang, Yaxing Wang, Jian Yang, Ming-Ming Cheng
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| replace naive prompt modification with anchor token matching | LlamaGen editing · PIE-Bench subset (5 editing types) | Structure Distance: 113.95 -> 31.79 | Improved |
| replace attention-map injection (PnP-AR) with anchor token matching | LlamaGen editing · PIE-Bench subset (5 editing types) | Structure Distance: 103.94 -> 31.79 | Improved |
| replace naive prompt modification with anchor token matching | LlamaGen editing · PIE-Bench subset (5 editing types) | PSNR: 12.14 -> 19.75 | Improved |
| anchor token matching vs naive modification on whole-image CLIP | LlamaGen editing · PIE-Bench subset (5 editing types) | CLIP similarity (whole): 24.71 -> 24.19 | Hurt |
| use dynamic window instead of fixed |W|=50 (ablation) | Anchor Token Matching · PIE-Bench subset (5 editing types) | Structure Distance: 60.83 -> 31.79 | Improved |
| use dynamic window instead of fixed |W|=150 (ablation) | Anchor Token Matching · PIE-Bench subset (5 editing types) | CLIP similarity: 22.18 -> 24.19 | Improved |
| use dynamic window instead of fixed |W|=150 (ablation) | Anchor Token Matching · PIE-Bench subset (5 editing types) | Structure Distance: 30.39 -> 31.79 | Hurt |
| use dynamic window instead of fixed windows (ablation) | Anchor Token Matching · PIE-Bench subset (5 editing types) | Structure Distance / CLIP ratio: 1.37 -> 1.31 | Improved |
| increase fixed window size from 50 to 150 (ablation) | Anchor Token Matching · PIE-Bench subset (5 editing types) | Structure Distance: 60.83 -> 30.39 | Improved |
| increase fixed window size from 50 to 150 (ablation) | Anchor Token Matching · PIE-Bench subset (5 editing types) | CLIP similarity: 24.79 -> 22.18 | Hurt |
| lower threshold tau in adaptive constraint relaxation (ablation) | Anchor Token Matching · qualitative edits | editing strength | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- InstructPix2Pix — InstructPix2Pix: Learning To Follow Image Editing Instructions (CVPR 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- EditAR: Unified Conditional Generation with Autoregressive Models (CVPR 2025 · similarity 0.89)
- Semantic Context Matters: Improving Conditioning for Autoregressive Models (CVPR 2026 · similarity 0.89)
- Toward Guidance-Free AR Visual Generation via Condition Contrastive Alignment (ICLR 2025 · similarity 0.88)
- PAIR Diffusion: A Comprehensive Multimodal Object-Level Image Editor (CVPR 2024 · similarity 0.87)
- Energy-Guided Optimization for Personalized Image Editing with Pretrained Text-to-Image Diffusion Models (AAAI 2025 · similarity 0.86)
- Towards Understanding Cross and Self-Attention in Stable Diffusion for Text-Guided Image Editing (CVPR 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.
