Chain of Thought (CoT)

1. Standard Prompting vs. Chain of Thought (CoT)

graph TD
    subgraph "Chain of Thought (CoT)"
        direction TB
        A[Question] --> B{LLM}
        B --> C[Step 1: Reasoning]
        C --> D[Step 2: Reasoning]
        D --> E[...]
        E --> F[Final Answer]
    end

    subgraph Standard Prompting
        direction TB
        G[Question] --> H{LLM}
        H --> I[Final Answer]
    end

2. Zero-Shot vs. Few-Shot CoT

flowchart TD
    subgraph Prompt
        A("User Question: 'Q'")
        A --> B("Combined Prompt")
    end

    subgraph Zero-Shot CoT
        C("Instruction: 'Let's think step by step.'")
        C --> B
    end

    subgraph Few-Shot CoT
        D("Example 1: 
Q: ... 
A: Step 1..., Step 2..., Answer is X")
        E("Example 2: 
Q: ... 
A: Step 1..., Step 2..., Answer is Y")
        D --> E --> F("Add to Prompt")
        F --> B
    end

    B --> G{LLM generates Reasoning & Answer}

3. Self-Consistency with CoT

graph TD
    A[Question] --> B{LLM};
    B -- "Path 1" --> C["Thought Process A
Result: 42"];
    B -- "Path 2" --> D["Thought Process B
Result: 42"];
    B -- "Path 3" --> E["Thought Process C
Result: 57"];

    subgraph Aggregation
        C --> F{Majority Vote};
        D --> F;
        E --> F;
    end

    F --> G[Final Answer: 42];

4. Tree of Thoughts (ToT)

graph TD
    A[Initial Problem] --> B(Thought 1);
    B --> C(Step 2a);
    B --> D(Step 2b - Pruned);
    B --> E(Step 2c);

    style D fill:#f9f,stroke:#333,stroke-width:2px;

    C --> F(Step 3a);
    %% The image shows Step 2a branching to both 3a and 3b
    C --> G(Step 3b);
    E --> H(Step 3c);

    subgraph Evaluation
        I(Solution Path 1)
        J(Solution Path 2)
        K(Solution Path 3)
        L{Select Best Path}

        I --> L;
        J --> L;
        K --> L;
    end

    %% Connect the steps from outside the subgraph to the nodes inside it
    F --> I;
    G --> J;
    H --> K;

    L --> M[Final Answer];