Planning: Chain-of-Thought, Tree-of-Thought & Graph Search
Planning: Chain-of-Thought, Tree-of-Thought & Graph Search
💡 Key Takeaways
- Understand how Chain-of-Thought planning produces step-by-step reasoning to reach a decision.
- Learn how Tree-of-Thought expands reasoning with branching paths and criteria for selecting the best branch.
- Compare Chain-of-Thought and Tree-of-Thought approaches: linear vs branched reasoning in agent planning and when to use each.
- Grasp Graph Search in agent architectures: exploring and remembering states via a graph to prune paths and improve efficiency.
❓ Frequently Asked Questions
What is Chain-of-Thought (CoT) reasoning?
Chain-of-Thought is a prompting approach that asks a model to show its step-by-step reasoning, helping it reach more accurate final answers by making intermediate thoughts explicit.
What is Tree-of-Thought (ToT) search?
Tree-of-Thought extends CoT by exploring multiple reasoning branches as a tree, evaluating and pruning branches to find a reliable solution rather than relying on a single chain of steps.
What is Graph Search in planning?
Graph Search treats possible states as nodes and actions as edges, exploring the state space with search strategies (e.g., BFS, DFS, A*) to find a path to a goal, often using heuristics.
How do CoT, ToT, and Graph Search differ and when should you use them?
CoT yields a single explicit reasoning chain; ToT builds and compares multiple reasoning branches as a tree; Graph Search systematically explores a state-space graph to optimize a path. Use CoT for straightforward tasks, ToT for problems with uncertain steps or multiple strategies, and Graph Search for problems best modeled as state graphs with goals.