README.md

---
library_name: transformers
license: apache-2.0
language:
- en
base_model: Qwen/Qwen3-4B-Instruct-2507
pipeline_tag: text-generation
tags:
- code-search
- code-localization
- reinforcement-learning
- agent
- software-engineering
- GSPO
- OpenHands
- SWE-Bench
datasets:
- OpenHands/SWE-smith-py-code-search
- OpenHands/SWE-Gym-code-search
- OpenHands/CodeScout_Training_Rollouts
---

# CodeScout-4B

[📄 Paper](https://arxiv.org/abs/2603.17829) • [💻 Code](https://github.com/OpenHands/codescout) • [🤗 Collection](https://huggingface.co/collections/OpenHands/codescout-69b9a6adcf21f348f4db937f)

**Best efficiency–performance trade-off — outperforms 8× larger Qwen3-32B across all benchmarks.**

<p align="center">
  <img src="codescout_overview.png" alt="CodeScout Overview" width="100%">
</p>

CodeScout-4B is part of the **CodeScout** family of open-source RL-trained code search agents.
CodeScout models achieve state-of-the-art repository-level code localization using *nothing more than a standard Unix terminal* — no static analysis, no repository graphs, no language-specific tooling.

## Key Highlights

- Consistently **outperforms 8× larger Qwen3-32B** on all benchmarks
- Surpasses RepoNavigator-14B by **2–10%** in file F1 and **8–11%** in function F1
- Exceeds GPT-5 with RepoNavigator by **9%** in file F1 and **5%** in function F1 on SWE-Bench Verified
- Best efficiency–performance trade-off in the CodeScout family

## Results

Performance on SWE-Bench code localization (instance-averaged F1 scores):


| Benchmark | CodeScout-1.7B | CodeScout-4B | CodeScout-14B |
|---|---|---|---|
| **SWE-Bench Verified** — File F1 | 55.46 | 68.52 | **68.57** |
| **SWE-Bench Verified** — Func F1 | 28.22 | 36.78 | **40.32** |
| **SWE-Bench Pro** — File F1 | 40.96 | 51.77 | **53.63** |
| **SWE-Bench Pro** — Func F1 | 18.24 | **29.03** | 28.74 |
| **SWE-Bench Lite** — File F1 | 56.57 | 67.03 | **71.84** |
| **SWE-Bench Lite** — Func F1 | 27.07 | 39.87 | **44.43** |


<p align="center">
  <img src="f1_vs_params_file.png" alt="File-level F1 vs Model Size" width="48%">
  <img src="f1_vs_params_function.png" alt="Function-level F1 vs Model Size" width="48%">
</p>

<p align="center"><em>Code localization performance on SWE-Bench Verified. CodeScout (⭐) achieves superior or competitive results over larger open-source LLMs and narrows the gap with closed-source frontier models.</em></p>

## Training

CodeScout-4B is trained directly from `Qwen3-4B-Instruct-2507` using GSPO reinforcement learning.

- **Training data:** 1,600 instances from [SWE-Smith](https://huggingface.co/datasets/OpenHands/SWE-smith-py-code-search) (39K filtered, 128 repos)
- **RL steps:** 200
- **Batch size:** 8, with 8 rollouts per instance
- **Max context length:** 40K tokens
- **Max turns per episode:** 6
- **Reward:** Multi-level F1 (file + module + function)
- **Hardware:** 8×H100 GPUs
- **Learning rate:** 1e-6 (constant)

## How It Works

CodeScout uses the **OpenHands-Bash** scaffold — an agent equipped with only a `Terminal` tool (supporting standard Unix commands like `rg`, `find`, `grep`, `ls`) and a `LocalizationFinish` tool for structured output submission. The agent iteratively navigates the repository to identify relevant files, classes, and functions related to a given issue.

The model is trained with **GSPO** (Group Sequence Policy Optimization) using multi-level F1 rewards at the file, module, and function level.

## Intended Use

CodeScout-4B is designed for **repository-level code localization**: given a GitHub issue description and a code repository, it identifies the relevant files, classes, and functions that need to be modified. It is intended to be used as a localization subagent within larger coding agent pipelines.

## Limitations

- Trained and evaluated exclusively on **Python** repositories
- Designed for code *localization*, not code *editing* or issue resolution
- Performance may vary on repositories significantly different from the training distribution
- Requires the OpenHands-Bash scaffold for optimal performance

## Citation

```bibtex
@misc{sutawika2026codescouteffectiverecipereinforcement,
      title={CodeScout: An Effective Recipe for Reinforcement Learning of Code Search Agents}, 
      author={Lintang Sutawika and Aditya Bharat Soni and Bharath Sriraam R R and Apurva Gandhi and Taha Yassine and Sanidhya Vijayvargiya and Yuchen Li and Xuhui Zhou and Yilin Zhang and Leander Melroy Maben and Graham Neubig},
      year={2026},
      eprint={2603.17829},
      archivePrefix={arXiv},
      primaryClass={cs.SE},
      url={https://arxiv.org/abs/2603.17829}, 
}
```