Search-R1: An Efficient, Scalable RL Training Framework for Reasoning & Search Engine Calling interleaved LLM based on veRL
active 2025-02-28 → 2026-08-07 (UTC)
Activity over time
Daily event counts in the loaded window
Line chart, 526 days from 2025-02-28 to 2026-08-07. Pushes: 58 total, peak 4 in a day. Pull requests: 14 total, peak 4 in a day. Issues: 226 total, peak 9 in a day. Comments: 307 total, peak 16 in a day. Stars: 3,099 total, peak 350 in a day.
- Pushes
- Pull requests
- Issues
- Comments
- Stars
Stars, PRs, issues and forks are under-captured in the later part of this window. GH Archive progressively stopped capturing non-push events during 2026 — −95% or worse by the end of the window. Every series here except Pushes fades for that reason, so a decline above reflects the archive, not this repository. Pushes stay reliable throughout, so read them, and the contributor counts derived from them, as the real signal. Data health has the measurements.
Top contributors
Pushes, PRs, issues, reviews and comments — stars and forks excluded, so this is contribution rather than popularity
| Contributor | Contributions | Pushes | PRs | Comments |
|---|---|---|---|---|
| PeterGriffinJin | 262 | 57 | 4 | 118 |
| Danee-wawawa | 25 | 0 | 0 | 15 |
| Zhuofeng-Li | 10 | 0 | 0 | 6 |
| xjtupy | 9 | 0 | 0 | 8 |
| codingforwhat | 8 | 0 | 0 | 5 |
| Stardust-y | 6 | 0 | 0 | 3 |
| GiantAILab | 6 | 0 | 0 | 4 |
| JunzheShen | 6 | 0 | 0 | 3 |
| tehranixyz | 6 | 0 | 0 | 1 |
| yazdanbakhsh | 5 | 0 | 0 | 2 |
| BaleChen | 5 | 0 | 0 | 2 |
| yangluo23 | 5 | 0 | 0 | 3 |
| wannanfeng | 5 | 0 | 0 | 2 |
| YouthLiuYS | 5 | 0 | 0 | 2 |
| xiaobo-yang | 5 | 0 | 4 | 1 |
| JiwenJ | 5 | 0 | 0 | 4 |
| coscotuff | 4 | 0 | 0 | 2 |
| ghh1125 | 4 | 0 | 0 | 3 |
| ZihuaZhao | 4 | 0 | 0 | 1 |
| Necolizer | 4 | 0 | 0 | 2 |
Recent activity
Latest issues, pull requests and releases
- Issue#177PeterGriffinJin2026-05-31 04:29About benchmark
- Pull request#183dashitongzhi2026-05-19 00:26
- Issue comment#182jorel6662026-05-10 05:39请问 sm_120框架的 blackwell gpu可以使用吗
- Issue comment#167PeterGriffinJin2026-04-10 03:41Why is {do_search} False in the last turn
- Issue#175hyntdlemon19002026-03-07 11:54训练稳定性与信用分配：针对长检索序列实现“跳过检索内容”的 GAE 机制
- Issue comment#129JamesSand2026-02-16 07:58verl/third_party/vllm modifications and support for newer vLLM versions
- Issue comment#127JamesSand2025-12-31 03:28OBSERVATION TOO LONG, CONSIDER CHANGING YOUR CONFIG
- Issue#171happyzzj22025-12-21 06:33Test dataset
- Issue comment#151ccccccchuang2025-12-06 04:19Why not use </search> for early stopping during rollout?
- Issue comment#169heavenrain992025-11-28 08:25Inference using the VLLM-based OpenAI interface fails to align with the output of the offline interface.
- Issue#169heavenrain992025-11-27 07:26基于vllm的openai接口推理，输出无法与离线接口对齐
- Pull request#168BFCmath2025-11-26 15:13
- Issue comment#98Fdioa2025-11-12 03:37Qwen3 support
- Issue#166WanFan112025-11-08 16:45Urgent, training encountered an error: batch size must be positive
- Issue comment#165PeterGriffinJin2025-11-07 14:01Some questions about retrieval services
- Issue#165yaxundai2025-11-07 09:51Some questions about retrieval services
- Issue#162Miracleme7782025-11-03 18:04Response become garbled in later stages of GRPO training
- Issue comment#79lihua88482025-11-01 06:24reward question
- Issue comment#133jiangix-paper2025-10-29 08:58Newer WIKI?
- Issue comment#118swordlidev2025-10-27 03:31Questions About use_kl_loss Logic
- Issue comment#155Fdioa2025-10-23 02:22Top-k Parameter Set to 3, but 5 Documents Returned for Some Queries During Training
- Issue comment#157Ymmm3302025-10-22 10:44Retrieval Service Fails with Faiss CUBLAS Error During PPO Training：CUBLAS_STATUS_EXECUTION_FAILED
- Issue comment#157JingLu11132025-10-22 10:41Retrieval Service Fails with Faiss CUBLAS Error During PPO Training：CUBLAS_STATUS_EXECUTION_FAILED
- Issue comment#153Fdioa2025-10-21 04:09Upgrade Search-R1 to support latest vLLM and Verl
- Issue#156AlbertQiSun2025-10-17 16:16model support issues
Totals cover only the window loaded into ClickHouse and count events, not GitHub's lifetime totals — 3,099 stars here means stars gained during the window, not the repo's star count.