Research map / Multi-agent systems and games / LLM agents
One Tool Is Enough: Reinforcement Learning of LLM Agents for Repository-Level Code Navigation
Zhaoxi Zhang, Yitong Duan, Yanzhi Zhang, Yiming Xu, Zhixiang Wang, Kun Liang, Weikang Li, Jiahui Liang, Deguo Xia, Jizhou Huang, Jiyan He, Shuxin Zheng and 1 more
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes RepoNavigator, an LLM agent with a single execution-aware tool trained with reinforcement learning for repository-level code localization.
Problem
Locating files and functions that need modification in large software repositories is hard; existing LLM methods treat it as retrieval and rely on multiple auxiliary tools that ignore execution logic.
Approach
An LLM agent equipped with a single execution-aware tool, trained via reinforcement learning to navigate repositories, replacing multi-tool retrieval pipelines.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add all three extra tools (GetClass, GetFunc, GetStruc) to jump (ablation) | RepoNavigator · SWE-bench Verified function-level (7B) | IoU: 24.28 -> 13.71 | Hurt |
| add GetClass and GetFunc to jump (ablation) | RepoNavigator · SWE-bench Verified function-level (7B) | IoU: 24.28 -> 21.44 | Hurt |
| add GetStruc to jump (ablation) | RepoNavigator · SWE-bench Verified function-level (7B) | IoU: 24.28 -> 24.00 | Hurt |
| train with GRPO | RepoNavigator 7B · SWE-bench Verified function-level | Sample-F1: 16.19 -> 27.49 | Improved |
| train with GRPO | RepoNavigator 14B · SWE-bench Verified function-level | Sample-F1: 25.58 -> 29.23 | Improved |
| train with GRPO | RepoNavigator 32B · SWE-bench Verified function-level | Sample-F1: 27.12 -> 34.09 | Improved |
| train with GRPO | RepoNavigator 7B · SWE-bench Pro function-level | Sample-F1: 10.67 -> 14.29 | Improved |
| add tool-calling success rate to the reward (hybrid reward) (ablation) | GRPO RepoNavigator 7B · SWE-bench Verified | localization metrics | Improved |
| use RL (GRPO) instead of SFT (ablation) | RepoNavigator 7B · SWE-bench Verified | localization metrics | Improved |
| allow more tool-calling turns (ablation) | RepoNavigator · SWE-bench Verified | DICE and IoU: avg slope 0.51 / turn to 1.08 / turn | Improved |
| replace Agentless localization with RepoNavigator | Agentless repair · SWE-bench Verified (14B) | resolved (%): 10.12 -> 14.74 | Improved |
| add RL training to RepoNavigator localization | RepoNavigator · SWE-bench Verified (14B) | resolved (%): 14.74 -> 15.03 | Improved |
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.
- OrcaLoca — OrcaLoca: An LLM Agent Framework for Software Issue Localization (ICML 2025) · better on 6 of 6 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
- Improving Code Localization with Repository Memory (ICLR 2026 · similarity 0.87)
- RepoAudit: An Autonomous LLM-Agent for Repository-Level Code Auditing (ICML 2025 · similarity 0.87)
- RepoMirage: Probing Repository Context Reasoning in Code Agents with Perturbations (NeurIPS 2026 · similarity 0.87)
- RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning (ICML 2025 · similarity 0.87)
- Build-and-Find: An Effort-Aware Protocol for Evaluating Agent-Managed Codebases (NeurIPS 2026 · similarity 0.87)
- SWE-RL: Advancing LLM Reasoning via Reinforcement Learning on Open Software Evolution (NeurIPS 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.
