# CodeGrep: An RL-Trained Retrieval Agent for LLM Coding Agents

arXiv: https://arxiv.org/abs/2608.05886 · HTML: https://mlsanity.com/paper/2608.05886/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Software engineering coding agents, one of 16 directions in LLM agents and tool use.
- Builds on 15 of its 25 references that are in the map.
- First posted on arXiv 6 Aug 2026.

## Placed in

- LLM agents and tool use / [Software engineering coding agents](https://mlsanity.com/agents/software-engineering-coding-agents/)

## Builds on (most cited first, 15 of 15)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [Toolformer: Language Models Can Teach Themselves to Use Tools](https://mlsanity.com/paper/2302.04761/) (2023)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://mlsanity.com/paper/2405.15793/) (2024)
- [Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning](https://mlsanity.com/paper/2503.09516/) (2025)
- [CodeBERT: A Pre-Trained Model for Programming and Natural Languages](https://mlsanity.com/paper/2002.08155/) (2020)
- [OpenHands: An Open Platform for AI Software Developers as Generalist Agents](https://mlsanity.com/paper/2407.16741/) (2024)
- [GraphCodeBERT: Pre-training Code Representations with Data Flow](https://mlsanity.com/paper/2009.08366/) (2020)
- [Agentless: Demystifying LLM-based Software Engineering Agents](https://mlsanity.com/paper/2407.01489/) (2024)
- [RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning](https://mlsanity.com/paper/2504.20073/) (2025)
- [UniXcoder: Unified Cross-Modal Pre-training for Code Representation](https://mlsanity.com/paper/2203.03850/) (2022)
- [SWE-Fixer: Training Open-Source LLMs for Effective and Efficient GitHub Issue Resolution](https://mlsanity.com/paper/2501.05040/) (2025)
- [Learning to Retrieve from Agent Trajectories](https://mlsanity.com/paper/2604.04949/) (2026)

## Built on by (most cited first, 3 of 3)

- [Vulnerability Localization Benchmark: Measuring Agentic Security Analysis at Repository Scale](https://mlsanity.com/paper/2609.15939/) (2026)
- [From Migration to Calibration: Aligning Agents with Deployment Requirements](https://mlsanity.com/paper/2609.35149/) (2026)
- [Small Agents with Semantic Search: Efficient Multilingual Code Localization](https://mlsanity.com/paper/2610.05099/) (2026)

## Abstract (arXiv)

Modern LLM coding agents such as Claude Code and OpenHands share a common inefficiency: they spend much of their token budget finding the file to patch, rather than patching it. On SWE-Bench Verified, a 30B OpenHands agent averages 23 rounds and 631K tokens per resolved issue, with many calls spent on grep, glob, and view_file during repository exploration. We introduce CodeGrep, a 14B retrieval agent trained end-to-end with GRPO to issue multi-turn parallel grep, glob, and read tool calls and return candidate files to a frozen downstream coding agent. On all 500 SWE-Bench Verified instances, CodeGrep preserves resolve rate while substantially improving efficiency: 27.0% versus 25.8% for the no-retrieval baseline, with 15% fewer rounds and 19% fewer tokens on resolved instances. Across retrievers, downstream utility follows a precision threshold: BM25 with precision 0.375 degrades the agent, Jina with precision 0.445 is neutral, and CodeGrep with precision 0.677 crosses the threshold at which retrieval begins to reduce rollout cost. To enable this study, we mine supervision from 67K open-source agent trajectories using CATM and build a Git-worktree environment for multi-turn agent RL. In our setting, applying the efficiency signal at the advantage layer rather than the reward layer reduces KL drift and translates cleanly into downstream efficiency. We will release the model, training pipeline, RL environment, and evaluation harnesses.