What is it?
Reasoning (in language models)
In this context, reasoning means working through a problem in multiple connected steps to reach a conclusion, rather than recalling a fact or pattern in one shot. A model that can take "all men are mortal" and "Socrates is a man" and conclude "Socrates is mortal" is reasoning, chaining pieces of information toward a new result.
The whole of folder 02 is about strengthening this ability: giving models room to reason (Chapter 1), grounding it with tools (Chapter 2), checking it step by step (Chapter 3), and growing it through practice (Chapter 4).
From the Planning and Reasoning glossary
Where Reasoning is explained
- Model-Provider Abstraction and Multi-Model StrategyArchitecture and System Design · Real code: one interface every provider implements
- DeepSeek-R1, Learning to Reason Through Reinforcement LearningPlanning and Reasoning · Why this paper mattered
- ReAct, Reasoning and Acting TogetherPlanning and Reasoning · Two failure modes ReAct fixes
- Let's Verify Step by Step, Rewarding Good ReasoningPlanning and Reasoning
- Chain of Thought, Teaching Models to Think Out LoudPlanning and Reasoning · Why this paper mattered