Planning & Reasoning

This chapter delves into the planning and reasoning capabilities of agents — from classical planning to chain-of-thought, tree search, reflection mechanisms, and the latest advances in reasoning models.

Contents:

• Planning & Reasoning Overview — From STRIPS to LLM reasoning: the full evolution
• Chain-of-Thought & Reasoning Patterns — CoT and its variants
• ReAct & Tool Reasoning — Thought-Action-Observation loop and tool-augmented reasoning
• Tree Search & Monte Carlo — Tree of Thoughts, Graph of Thoughts, and MCTS
• Reflection & Self-Improvement — Reflexion, Self-Refine, and verbal reinforcement learning
• Planning Execution Frameworks — Plan-and-Execute, LLMCompiler, and hierarchical planning
• Frontiers in Reasoning — o1/o3 reasoning models, DeepSeek-R1, and reasoning scaling laws

---

评论