ReAct
Updated 2026-08-09
INTRODUCTION
English translation pending.
CORE DEFINITION
ReAct, short for Reasoning and Acting, is a prompting and agent pattern in which a large language model interleaves thought and action. The model first reasons about the current state and the next step, then takes an action such as searching, calculating, or calling an API, and finally receives an observation that becomes input for the next cycle. Because external information corrects reasoning and reasoning chooses the next action, the loop grounds the model in real feedback rather than a single unchecked generation.
SCAFFOLDING EFFECT
Reduce cognitive load
- Use agent loop: alternate a thought step with exactly one concrete tool action. - Use observation grounding: feed every tool result back into the model before deciding again. - Use stopping rule: define an exit condition so the cycle cannot spin forever.
Anchor fast decisions
A single-pass generation drifts and cannot consult the outside world. ReAct has the model alternate thought, which plans and reasons, with action, which calls tools or queries sources, then feeds the observation back as new input. Action supplies real information that corrects reasoning, and reasoning selects the next action, so the two reinforce each other across cycles.
MINIMUM ACTION
In progress 0/1Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- en.wikipedia.orghttps://en.wikipedia.org/wiki/Prompt_engineeringverified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS