Research map / Image generation / Fast diffusion sampling
Free Lunch for Stabilizing Rectified Flow Inversion
Chenru Wang, Beier Zhu, Chi Zhang
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 PMI | Euler · PIE-Bench reconstruction (conditional) | PSNR: 22.10 -> 22.56 | Improved |
| add PMI | Heun · PIE-Bench reconstruction (conditional) | PSNR: 29.16 -> 30.38 | Improved |
| add PMI | RF-Solver · PIE-Bench reconstruction (conditional) | PSNR: 29.17 -> 30.72 | Improved |
| add PMI | Euler · PIE-Bench reconstruction (conditional) | LPIPS x10^3: 168.48 -> 179.10 | Hurt |
| add PMI | Euler · PIE-Bench reconstruction (unconditional) | PSNR: 21.96 -> 23.26 | Improved |
| add PMI | FireFlow · PIE-Bench reconstruction (unconditional) | PSNR: 28.87 -> 29.73 | Improved |
| add PMI and mimic-CFG | RF-Solver · PIE-Bench editing | background PSNR: 21.29 -> 22.18 | Improved |
| add PMI and mimic-CFG | RF-Inversion · PIE-Bench editing | CLIP similarity (whole): 25.02 -> 25.44 | Improved |
| add mimic-CFG on top of PMI (ablation) | FireFlow · PIE-Bench editing | CLIP similarity (whole): 25.34 -> 25.50 | Improved |
| add L1 proximity term versus none (ablation) | PMI editing · PIE-Bench editing | PSNR: 23.66 -> 23.67 | No clear effect |
| use L2 norm instead of L1 (ablation) | PMI editing · PIE-Bench editing | PSNR: 23.67 -> 23.65 | Hurt |
| replace EMA velocity averaging with integral averaging (ablation) | FireFlow editing · PIE-Bench editing | SSIM x10^2: 82.97 -> 81.76 | Hurt |
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.
- FireFlow — FireFlow: Fast Inversion of Rectified Flow for Image Semantic Editing (ICML 2025) · better on 4 of 4 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
- Taming Rectified Flow for Inversion and Editing (ICML 2025 · similarity 0.91)
- Semantic Image Inversion and Editing using Rectified Stochastic Differential Equations (ICLR 2025 · similarity 0.91)
- Text-to-Image Rectified Flow as Plug-and-Play Priors (ICLR 2025 · similarity 0.89)
- Root-Selecting Fixed-Point Inversion for Rectified Flows via Trajectory Straightness (NeurIPS 2026 · similarity 0.89)
- Solving Inverse Problems with FLAIR (NeurIPS 2025 · similarity 0.89)
- FireFlow: Fast Inversion of Rectified Flow for Image Semantic Editing (ICML 2025 · similarity 0.89)
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.
