Research map / Reinforcement learning / Visual and representation RL
Functional Cache Grafting: Robust and Rapid Code-Policy Synthesis for Embodied Agents
Saehun Chun, Wonje Choi, Sera Choi, Sanghyun Ahn, Honguk Woo
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 Function-Interface tier I to the code cache (ablation) | FCGraft · RLBench Open-Composition | SR: 38.64 -> 45.91 | Improved |
| add Function-Code tier C to the code cache (ablation) | FCGraft · RLBench Open-Composition | SR: 34.37 -> 45.91 | Improved |
| add the whole hierarchical code cache H (ablation) | FCGraft · RLBench Open-Composition | SR: 34.64 -> 45.91 | Improved |
| add semantic locality term l_sema (ablation) | FCGraft · RLBench Open-Composition | SR: 35.02 -> 45.91 | Improved |
| add semantic locality term l_sema (ablation) | FCGraft · RLBench Open-Composition | HR: 48.61 -> 73.65 | Improved |
| use semantic scoring instead of retrieval-based cache selection (ablation) | FCGraft · RLBench Open-Composition | HR: 43.06 -> 73.65 | Improved |
| use semantic scoring instead of retrieval-based cache selection (ablation) | FCGraft · RLBench Open-Composition | PSL: 4.42 -> 3.24 | Improved |
| use localized cache-patching instead of full regeneration (ablation) | FCGraft · RLBench Open-Composition | PSL: 6.89 -> 3.24 | Improved |
| use localized cache-patching instead of full regeneration (ablation) | FCGraft · RLBench Open-Composition | SR: 35.77 -> 45.91 | Improved |
| add Function-Interface tier I to the code cache (ablation) | FCGraft · RLBench Open-Composition | PSL: 4.12 -> 3.24 | Improved |
| use a CodeLLM instead of a general-purpose LLM (ablation) | FCGraft (Qwen2.5) · RLBench (model-choice ablation) | SR: 26.38 -> 29.62 | Improved |
| use a CodeLLM instead of a general-purpose LLM (ablation) | FCGraft (Llama2) · RLBench (model-choice ablation) | SR: 25.33 -> 31.12 | Improved |
Closest papers by meaning
- Towards Reliable Code-as-Policies: A Neuro-Symbolic Framework for Embodied Task Planning (NeurIPS 2025 · similarity 0.89)
- CaP-X: A Framework for Benchmarking and Improving Coding Agents for Robot Manipulation (ICML 2026 · similarity 0.88)
- Towards Direct Latent-Space Synthesis for Parallel Branches in LLM-Agent Workflows (NeurIPS 2026 · similarity 0.87)
- ECHO: Terminal Agents Learn World Models for Free (NeurIPS 2026 · similarity 0.87)
- Efficient Skill Grounding via Code Refactoring with Small Language Models (ICML 2026 · similarity 0.87)
- Generalizable End-to-End Tool-Use RL with Synthetic CodeGym (ICLR 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.
