ML//agent//ReAct

- Reasoning + Acting: the model alternates between thinking (reasoning traces) and acting (tool calls)

Reasoning + Acting: the model alternates between thinking (reasoning traces) and acting (tool calls)

Loop: "Thought → Action → Observation → Thought → ..." until the task is solved.

The paper that started the tool-using LLM paradigm. Led to function calling, Gorilla, BFCL leaderboard.

Simple but effective. Most modern agent frameworks implement variations of ReAct.If I could say anything, I would say seomthing.