PeterGriffinJin/Search-R1

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

5328 stars · 483 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Search-R1 is an open-source reinforcement learning framework for training LLMs that interleave reasoning with search engine calls, built on top of veRL. It supports multiple RL methods (PPO, GRPO, REINFORCE), LLMs (Llama, Qwen), and search backends (local sparse/dense retrievers and online engines).

Use cases

• train an LLM to reason and call a search engine with reinforcement learning
• build an open-source alternative to OpenAI DeepResearch
• fine-tune Qwen or Llama models with PPO or GRPO for tool-augmented reasoning
• research RL training for interleaved reasoning and retrieval
• train models to interleave chain-of-thought with retrieval calls

When to choose

• you want to RL-train LLMs to interleave reasoning with search engine calls
• you need an open-source pipeline replicating DeepSeek-R1-style training with retrieval
• you are doing research on tool-augmented LLM reasoning

When to avoid

• you only need inference-time RAG without training
• you lack GPU resources for RL fine-tuning
• you need a production search agent rather than a research training framework

Facets

framework · maturity active

llm-training reinforcement-learning rag agent-framework machine-learning large-language-models reinforcement-learning machine-learning python rl-training tool-calling search-engine-calling verl ppo grpo deep-research llm-reasoning retrieval-augmented-generation ai-agents gpu linux docker

Member repositories

For agents

markdown · JSON · MCP: product_card(name="PeterGriffinJin/Search-R1")