Skip to content

40. Collaborative Tree Search (CoTS)

Mini-Project: Collaborative Tree Search

A multi-agent tree search where specialized agents explore different branches in parallel, cross-examine each other's scores, then all collaborate on a deep-dive of the winning branch before synthesizing a final solution.

View on GitHub


Description

Collaborative Tree Search (CoTS) uses multiple agents to explore different branches of a search tree simultaneously. Each agent specializes in a different type of reasoning (e.g., one explores creative solutions, another explores conservative ones). Agents share discoveries through a shared tree structure, enabling faster and more diverse exploration.

How It Works

Multiple agents explore different branches of the same search tree in parallel, sharing their findings through a common tree data structure.

Diagram

flowchart TD
    A[Search Tree Root] --> B[Branch 1]
    A --> C[Branch 2]
    A --> D[Branch 3]
    B --> E[Agent A: Creative Explorer]
    C --> F[Agent B: Analytical Explorer]
    D --> G[Agent C: Conservative Explorer]
    E -->|Share findings| H[Shared Tree]
    F -->|Share findings| H
    G -->|Share findings| H
    H --> I[Best Combined Solution]

    style E fill:#9C27B0,color:#fff
    style F fill:#2196F3,color:#fff
    style G fill:#00BCD4,color:#fff
    style H fill:#FF5722,color:#fff