In plain words: Instead of prompting a model to look things up, this trains it with rewards to write its own search queries mid-thought and fold results into its answers. On seven question-answering tests it beat the usual retrieve-then-answer approach by 41% with the larger of two models.
Abstract · Search-R1: Training LLMs to Reason and Leverage Search Engines with Reinforcement Learning
Efficiently acquiring external knowledge and up-to-date information is essential for effective reasoning and text generation in large language models (LLMs). Prompting advanced LLMs with reasoning capabilities to use search engines during inference is often suboptimal, as the LLM might not fully possess the capability on how to interact optimally with the search engine. This paper introduces Search-R1, an extension of reinforcement learning (RL) for reasoning frameworks where the LLM learns to autonomously generate (multiple) search queries during step-by-step reasoning with real-time retrieval. Search-R1 optimizes LLM reasoning trajectories with multi-turn search interactions, leveraging retrieved token masking for stable RL training and a simple outcome-based reward function. Experiments on seven question-answering datasets show that Search-R1 improves performance by 41% (Qwen2.5-7B) and 20% (Qwen2.5-3B) over various RAG baselines under the same setting. This paper further provides empirical insights into RL optimization methods, LLM choices, and response length dynamics in retrieval-augmented reasoning. The code and model checkpoints are available at https://github.com/PeterGriffinJin/Search-R1.
Bowen Jin, Hansi Zeng, Zhenrui Yue, Jinsung Yoon, Sercan Arik, Dong Wang, Hamed Zamani, Jiawei Han
arXiv:2503.09516 · cs.CL, cs.AI, cs.IR · submitted Mar 12, 2025 · updated Aug 5, 2025
abstract · pdf · html · 31 pages
When OpenAI was launched this is what I thought it was going to be like. Something, something for the betterment of man kind.