Research map / Multi-agent systems and games / LLM agents
Self-Evolving Agentic Image Restoration via Deliberate Planning and Intuitive Execution
Shuang Cui, Fan Ji, Guanglong Sun, Yufei Guo, Xiongxin Tang, Jiangmeng Li, Fanjiang Xu
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 self-evolving episodic memory (ablation) | SEAR · MiO100 Group B | tool calls per image: 16.75 -> 8.15 | Improved |
| add self-evolving episodic memory (ablation) | SEAR · MiO100 Group B | PSNR: 22.0222 -> 22.1332 | Improved |
| add P-MCTS instead of greedy heuristics (ablation) | SEAR · MiO100 Group B | PSNR: 20.8395 -> 22.1332 | Improved |
| add P-MCTS instead of greedy heuristics (ablation) | SEAR · MiO100 Group B | tool calls per image: 4.08 -> 8.15 | Hurt |
| add macro-task agenda planning (ablation) | SEAR · MiO100 Group B | PSNR: 19.5378 -> 22.1332 | Improved |
| add macro-task agenda planning (ablation) | SEAR · MiO100 Group B | tool calls per image: 71.76 -> 8.15 | Improved |
| add pruning-aware MCTS instead of vanilla MCTS (ablation) | SEAR · MiO100 Group B | tool calls per image: 9.93 -> 8.15 | Improved |
| add pruning-aware MCTS instead of vanilla MCTS (ablation) | SEAR · MiO100 Group B | PSNR: 22.1252 -> 22.1332 | No clear effect |
| add MLLM semantic state fingerprint instead of IQA-only descriptors (ablation) | SEAR · MiO100 Group B | PSNR: 21.0774 -> 22.1332 | Improved |
| SEAR vs AgenticIR (agentic baseline) | AgenticIR · MiO100 Group B | PSNR: 21.72 -> 22.13 | Improved |
| SEAR vs 4KAgent in execution time | 4KAgent · MiO100 Group B | inference time (min): 2.55 -> 1.98 | Improved |
| exploration constant c=0.8 vs 0.4 | SEAR · Real-world dataset (100 pairs) | PSNR: 24.16 -> 24.41 | Improved |
Closest papers by meaning
- An Intelligent Agentic System for Complex Image Restoration Problems (ICLR 2025 · similarity 0.89)
- OPERA: An Agent for Image Restoration with End-to-End Joint Planning–Execution Optimization (NeurIPS 2026 · similarity 0.89)
- ViLoMem: Agentic Learner with Grow-and-Refine Multimodal Semantic Memory (CVPR 2026 · similarity 0.86)
- Agent0-VL: Exploring Self-Evolving Agent for Tool-Integrated Vision-Language Reasoning (ICML 2026 · similarity 0.86)
- MemEvolve: Meta-Evolution of Agent Memory Systems (ICML 2026 · similarity 0.86)
- SenseSearch: Empowering Vision-Language Models with High-Resolution Agentic Search-Reasoning via Reinforcement Learning (CVPR 2026 · 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.
