Research map / Image generation / Editing and style transfer
LayerVerse: Finding the Sweet Spot for KV-Injection in Training-Free Image Editing
Mikhail Zhirnov, Arsen Kuzhamuratov, Andrey Kuznetsov, Ivan Oseledets, Konstantin Sobolev
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 |
|---|---|---|---|
| apply LayerVerse layer allocation to FireFlow | FireFlow (FLUX.1-dev) · PIE-Bench | LPIPS: 64.59 -> 37.82 | Improved |
| apply LayerVerse to FireFlow | FireFlow (FLUX.1-dev) · PIE-Bench | EditScore: 3.60 -> 5.05 | Improved |
| apply LayerVerse to FireFlow | FireFlow (FLUX.1-dev) · PIE-Bench | Structure distance: 9.48 -> 9.36 | Improved |
| apply LayerVerse to FireFlow | FireFlow (FLUX.1-dev) · PIE-Bench | CLIP whole: 23.68 -> 24.39 | Mixed |
| apply LayerVerse to RF-Solver | RF-Solver (FLUX.1-dev) · PIE-Bench | avg rank: 5.88 -> 3.38 | Improved |
| use only global KV injection instead of combined (ablation) | LayerVerse · PIE-Bench FLUX.1-dev | Structure distance: 9.36 -> 11.01 | Hurt |
| use only masked KV injection instead of combined (ablation) | LayerVerse · PIE-Bench FLUX.1-dev | Structure distance: 9.36 -> 19.03 | Hurt |
| use random block selection with same budget (ablation) | LayerVerse (FireFlow) · PIE-Bench FLUX.1-dev | Structure distance: 9.36 -> 60.54 | Hurt |
| use random block selection with same budget (ablation) | LayerVerse (Euler) · PIE-Bench FLUX.1-dev | Structure distance: 12.24 -> 64.16 | Hurt |
| use quadratic pairwise error model instead of linear (ablation) | LayerVerse MILP · PIE-Bench FLUX.1-dev | reconstruction PSNR | Improved |
| reduce calibration set from 115 to 10 images (ablation) | LayerVerse · FLUX.1-dev calibration | cost / performance: ~348 -> ~30 GPU hours | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- FireFlow — FireFlow: Fast Inversion of Rectified Flow for Image Semantic Editing (ICML 2025) · better on 5 of 5 shared results
- DNAEdit — DNAEdit: Direct Noise Alignment for Text-Guided Rectified Flow Editing (NeurIPS 2025) · better on 3 of 4 shared results
- FlowEdit — FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models (ICCV 2025) · better on 2 of 3 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
- Stable Flow: Vital Layers for Training-Free Image Editing (CVPR 2025 · similarity 0.88)
- Streamlining Image Editing with Layered Diffusion Brushes (ICCV 2025 · similarity 0.88)
- Masked Region Transformer for Layered Image Generation and Editing at Scale (CVPR 2026 · similarity 0.88)
- Improving Editability in Image Generation with Layer-wise Memory (CVPR 2025 · similarity 0.86)
- DesignEdit: Unify Spatial-Aware Image Editing via Training-free Inpainting with a Multi-Layered Latent Diffusion Framework (AAAI 2025 · similarity 0.86)
- h-Edit: Effective and Flexible Diffusion-Based Editing via Doob's h-Transform (CVPR 2025 · 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.
