Research map / Vision-language and multimodal / Multimodal LLMs
RePlan: Reasoning-Guided Region Planning for Complex Instruction-Based Image Editing
Tianyuan Qu, Lei Ke, Xiaohang Zhan, Longxiang Tang, Yuqi Liu, Bohao PENG, Bei Yu, Dong Yu, Jiaya Jia
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 RePlan planning and attention injection | Flux.1-Kontext-dev · IV-Edit | Overall: 3.22 -> 3.46 | Improved |
| add RePlan | Qwen-Image-Edit · IV-Edit | Overall: 3.05 -> 3.51 | Improved |
| add RePlan | Flux.2-Klein-9B · IV-Edit | Overall: 3.45 -> 3.92 | Improved |
| add RePlan | Flux.1-Kontext-dev · IV-Edit | Effect: 2.73 -> 2.59 | Hurt |
| add RePlan | Flux.1-Kontext-dev · IV-Edit | Consistency: 2.88 -> 3.64 | Improved |
| add RePlan | Qwen-Image-Edit · IV-Edit | Consistency: 1.79 -> 3.24 | Improved |
| add RL training to the planner (vs zero-shot Gemini2.5-pro) (ablation) | RePlan planner · IV-Edit | Overall: 2.95 -> 3.46 | Improved |
| add RL training to the planner (vs zero-shot Qwen2.5-VL 7B) (ablation) | RePlan planner · IV-Edit | Overall: 2.60 -> 3.46 | Improved |
| add chain-of-thought reasoning to the planner (ablation) | RePlan planner · IV-Edit | Overall: 3.31 -> 3.46 | Improved |
| use two-stage RL instead of single-stage (ablation) | RePlan planner · IV-Edit | Overall: 3.42 -> 3.46 | Improved |
| add rule-group attention acceleration (ablation) | RePlan · IV-Edit latency 5 regions | latency (s): 113.8 -> 54.1 | Improved |
| replace multi-turn Kontext editing with RePlan single pass | Flux.1-Kontext-dev · IV-Edit latency 5 regions | latency (s): 221.0 -> 54.1 | Improved |
Closest papers by meaning
- Reasoning to Edit: Hypothetical Instruction-Based Image Editing with Visual Reasoning (ICML 2026 · similarity 0.88)
- Balancing Preservation and Modification: A Region and Semantic Aware Metric for Instruction-Based Image Editing (ICML 2025 · similarity 0.88)
- RetouchIQ: MLLM Agents for Instruction-Based Image Retouching with Generalist Reward (CVPR 2026 · similarity 0.88)
- FireEdit: Fine-grained Instruction-based Image Editing via Region-aware Vision Language Model (CVPR 2025 · similarity 0.88)
- Text-Vision Co-Instructed Image Editing (NeurIPS 2026 · similarity 0.88)
- Learning Action and Reasoning-Centric Image Editing from Videos and Simulation (NeurIPS 2024 · 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.
