Research map / Image generation / Editing and style transfer
T-Edit: Triple-Branch Diffusion Anchoring for Consistent Editing
Linsong Shan, Laurence Tianruo Yang, Zecan Yang, Shijie Lian, shijie lv, Qilin Yang
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 |
|---|---|---|---|
| add reconstruction-based error correction (ablation) | T-Edit base · PIE-Bench | LPIPS: 0.2270 -> 0.1540 | Improved |
| add reconstruction-based error correction (ablation) | T-Edit base · PIE-Bench | CLIP score: 31.12 -> 30.75 | Hurt |
| add reconstruction-based error correction (ablation) | T-Edit base · PIE-Bench | PSNR: 18.54 -> 19.22 | Improved |
| add feature injection at fixed vital layers (ablation) | T-Edit with error correction · PIE-Bench | PSNR: 19.22 -> 20.95 | Improved |
| add feature injection at fixed vital layers (ablation) | T-Edit with error correction · PIE-Bench | LPIPS: 0.1540 -> 0.1325 | Improved |
| replace fixed vital layers with dynamic vital layer localization (ablation) | T-Edit · PIE-Bench | PSNR: 20.95 -> 22.88 | Improved |
| replace fixed vital layers with dynamic vital layer localization (ablation) | T-Edit · PIE-Bench | LPIPS: 0.1325 -> 0.1237 | Improved |
| replace fixed vital layers with dynamic vital layer localization (ablation) | T-Edit · PIE-Bench | FID: 64.12 -> 64.15 | No clear effect |
| add TA frequency-aware injection in non-vital layers (ablation) | T-Edit · PIE-Bench | PSNR: 22.88 -> 23.89 | Improved |
| add TA frequency-aware injection in non-vital layers (ablation) | T-Edit · PIE-Bench | FID: 64.15 -> 62.22 | Improved |
| add TA frequency-aware injection in non-vital layers (ablation) | T-Edit · PIE-Bench | LPIPS: 0.1237 -> 0.1088 | Improved |
| use T-Edit instead of RF-Edit on video | HunyuanVideo · Video editing (Ditto-1M) | FID: 68.45 -> 65.87 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- Stable Flow — Stable Flow: Vital Layers for Training-Free Image Editing (CVPR 2025) · better on 5 of 5 shared results
- KV-Edit — KV-Edit: Training-Free Image Editing for Precise Background Preservation (ICCV 2025) · better on 2 of 5 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
- Why Do DiT Editors Drift? Plug-and-Play Low Frequency Alignment in VAE Latent Space (NeurIPS 2026 · similarity 0.91)
- Inversion-Free Image Editing with Language-Guided Diffusion Models (CVPR 2024 · similarity 0.88)
- Invertible Consistency Distillation for Text-Guided Image Editing in Around 7 Steps (NeurIPS 2024 · similarity 0.88)
- Stable Flow: Vital Layers for Training-Free Image Editing (CVPR 2025 · similarity 0.87)
- h-Edit: Effective and Flexible Diffusion-Based Editing via Doob's h-Transform (CVPR 2025 · similarity 0.87)
- DiT4Edit: Diffusion Transformer for Image Editing (AAAI 2025 · similarity 0.87)
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.
