推理模式
定义
推理模式是引出或组织模型推理的结构化方法: chain-of-thought (step-by-step), tree-of-thoughts (explore branches), ReAct (reason + act), and RDD (检索-决策-设计), among others. Using a clear pattern improves reliability (more consistent 推理) and debuggability (you can inspect steps or actions).
它们是 used in prompt engineering (例如 CoT) and inside agents (例如 ReAct, RDD). Choosing a pattern depends on the task: CoT for math/推理, ReAct for tool use, ToT for search/planning, RDD for spec compliance.
工作原理
You feed input (question, task) into a pattern: the pattern constrains how the model reasons or acts (例如 “think 逐步”, or thought–action–observation loops). The model produces an output (answer, action sequence). Prompts or system 设计 encourage the model to show 推理 (例如 “Think 逐步”) or to interleave thought and action. Patterns can be combined (例如 CoT inside an agent loop). See the linked pages for each pattern’s details.
应用场景
Different patterns suit different needs: CoT for stepwise 推理, ReAct for tool use, ToT for search and planning.
- CoT: math, logic, and multi-step 推理 tasks
- ReAct: tool-using agents that reason before each action
- ToT: search and planning over multiple solution branches
外部文档
- Chain-of-Thought Prompting (Wei et al.) — CoT paper
- ReAct: Synergizing Reasoning and Acting (Yao et al.) — ReAct paper
- Tree of Thoughts (Yao et al.) — ToT paper