🛰️ Daily AI Frontier
‹ back to 2026-08-07

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

Research LLM Agents

Ranking

Overall 71
Content 85
Popularity 39

Observed public metrics from 1 member.

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

Merged summary

TL;DR - CodeGrep is a 14B retrieval agent trained end-to-end with GRPO to find the right files for a frozen downstream LLM coding agent, cutting the exploration overhead that dominates token budgets on SWE-Bench Verified. It matters because it reframes code retrieval as a learned, RL-trained agentic subtask that measurably reduces rollout cost without sacrificing resolve rate.

  • Motivating measurement: a 30B OpenHands agent averages 23 rounds and 631K tokens per resolved SWE-Bench Verified issue, much of it spent on grep/glob/view_file exploration.
  • On all 500 SWE-Bench Verified instances, CodeGrep reaches 27.0% resolve rate vs. 25.8% for the no-retrieval baseline, with 15% fewer rounds and 19% fewer tokens on resolved instances.
  • Downstream utility shows a precision threshold: BM25 (precision 0.375) hurts the agent, Jina (0.445) is neutral, and CodeGrep (0.677) crosses the point where retrieval actually reduces cost.
  • Training infrastructure: supervision mined from 67K open-source agent trajectories via CATM, a Git-worktree environment for multi-turn agent RL, and an efficiency signal applied at the advantage layer (not the reward layer) to limit KL drift; model, pipeline, environment, and harnesses to be released.

Sources (1)

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

arXiv cs.SE Wuya Chen, Yihao yang, Yang Cao, Yue Lin 2026-08-06 arXiv:2608.05886
Public signals Semantic Scholar citations 0 · Semantic Scholar influential citations 0
Providers: Hugging Face · N/A OpenAlex · N/A Publisher · N/A Semantic Scholar · Citations 0 · Influential citations 0 X · N/A Fetched 2026-08-23 14:18:27.658342 UTC

TL;DR - CodeGrep is a 14B retrieval agent trained end-to-end with GRPO to find the right files for a frozen downstream LLM coding agent, cutting the exploration overhead that dominates token budgets on SWE-Bench Verified. It matters because it reframes code retrieval as a learned, RL-trained agentic subtask that measurably reduces rollout cost without sacrificing resolve rate.

  • Motivating measurement: a 30B OpenHands agent averages 23 rounds and 631K tokens per resolved SWE-Bench Verified issue, much of it spent on grep/glob/view_file exploration.
  • On all 500 SWE-Bench Verified instances, CodeGrep reaches 27.0% resolve rate vs. 25.8% for the no-retrieval baseline, with 15% fewer rounds and 19% fewer tokens on resolved instances.
  • Downstream utility shows a precision threshold: BM25 (precision 0.375) hurts the agent, Jina (0.445) is neutral, and CodeGrep (0.677) crosses the point where retrieval actually reduces cost.
  • Training infrastructure: supervision mined from 67K open-source agent trajectories via CATM, a Git-worktree environment for multi-turn agent RL, and an efficiency signal applied at the advantage layer (not the reward layer) to limit KL drift; model, pipeline, environment, and harnesses to be released.
item →