ReAct: Synergizing Reasoning and Acting in Language Models
by Shunyu Yao et al. (Princeton / Google Research)
The paper that taught LLMs to interleave thinking and acting.
Overview
ReAct: Synergizing Reasoning and Acting in Language Models, by Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao (arXiv:2210.03629, submitted Oct 6, 2022; ICLR 2023), explores using LLMs to generate both reasoning traces and task-specific actions in an interleaved manner. Reasoning traces help the model induce, track, and update action plans and handle exceptions, while actions let it interface with external sources such as knowledge bases or environments to gather additional information. The Thought-Action-Observation loop it introduces is widely regarded as a founding pattern for agentic LLMs, delivering improvements on question-answering and fact-verification tasks and greater interpretability than baselines that lack an explicit reasoning component.
At a Glance
- Topic
- Skills
- Level
- Advanced
- Format
- Paper
- Cost
- Free
- Provider
- Shunyu Yao et al. (Princeton / Google Research)
- Hands-on
- No
- Certificate
- None
What You’ll Learn
- ✓The Thought - Action - Observation loop underpinning modern agents
- ✓How interleaving reasoning with tool actions improves reliability
- ✓How reasoning traces help track plans and handle exceptions
- ✓Why grounding actions in external sources reduces hallucination
- ✓The interpretability benefits of explicit reasoning traces
Highlights
- •Foundational, widely cited basis for agentic LLM patterns
- •ICLR 2023 paper with public code and project page
- •Directly informs how frameworks structure tool-using loops today
Who It’s For
Best For
- ✓Engineers who want the primary source behind agent loops
- ✓Researchers and practitioners studying reasoning + tool use
Prerequisites
- •Familiarity with LLMs and prompting
- •Basic understanding of tool use / function calling
FAQ
What is ReAct: Synergizing Reasoning and Acting in Language Models?
The foundational ReAct paper (arXiv 2210.03629, ICLR 2023), which shows LLMs generating interleaved reasoning traces and task-specific actions to solve tool-using tasks. Essential reading for anyone building tool-using or agentic systems.
Is ReAct: Synergizing Reasoning and Acting in Language Models free?
ReAct: Synergizing Reasoning and Acting in Language Models is free to access.
What level is ReAct: Synergizing Reasoning and Acting in Language Models for?
ReAct: Synergizing Reasoning and Acting in Language Models is aimed at a advanced audience. Recommended background: Familiarity with LLMs and prompting, Basic understanding of tool use / function calling.
What will I learn from ReAct: Synergizing Reasoning and Acting in Language Models?
You'll learn: The Thought - Action - Observation loop underpinning modern agents; How interleaving reasoning with tool actions improves reliability; How reasoning traces help track plans and handle exceptions; Why grounding actions in external sources reduces hallucination; The interpretability benefits of explicit reasoning traces.