Implementation:Iamhankai Forest of Thought ToT Task Run

Knowledge Sources	Forest-of-Thought Tree of Thoughts
Domains	Search_Algorithms, Reasoning
Last Updated	2026-02-14 03:00 GMT

Overview

Concrete tool for executing Tree-of-Thought search over reasoning steps provided by the Forest-of-Thought repository.

Description

The ToT_Task class extends SearchTask to implement Tree-of-Thought reasoning. It manages tree construction via DFS or BFS, step generation via LLM prompting, value estimation via LLM scoring, and final answer extraction via summary generation. The class supports configurable branching factor, maximum depth, value threshold gating, and both greedy and sampling-based selection.

Usage

Instantiated by Monte_Carlo_Forest.tot_run() with problem data and configuration. The run() method executes the full search and returns a result dictionary with the solution and correctness information.

Code Reference

Source Location

• Repository: Forest-of-Thought
• File: methods/tot/task.py
• Lines: L11-263

Signature

class ToT_Task(SearchTask):
    def __init__(
        self, data, propose_method='glm', value_method='glm',
        algorithm='dfs', branch=1, select_branch=1,
        max_depth=1, end_gate=0.9, select_method='greedy',
        temperature=0.7, max_tokens=2048, seed=170,
        max_length=2048, truncation=True, do_sample=True,
        max_new_tokens=256, use_case_prompt=False,
        low=0, high=1, evaluate='', multiply_value=False,
        lang='en', answer=None, verify_method='string'
    ):
        """
        Args:
            data: Input problem text.
            algorithm (str): Search method, 'dfs' or 'bfs'.
            branch (int): Branching factor per step.
            select_branch (int): Branches to pursue after evaluation.
            max_depth (int): Maximum tree depth.
            end_gate (float): Value threshold for stopping.
            lang (str): Language ('en' or 'zh').
            answer: Ground truth for verification.
        """

    def run(self) -> tuple:
        """
        Execute ToT search.

        Returns:
            tuple: (
                result_dict: Dict with 'content', 'solution', 'summary',
                    'accurate', 'real_answer', 'multiply_value',
                root: Root Node of the search tree
            )
        """

Import

from methods.tot.task import ToT_Task

I/O Contract

Inputs

Name	Type	Required	Description
data	str	Yes	Problem text to solve
algorithm	str	No	Search algorithm: dfs or bfs (default: dfs)
branch	int	No	Branching factor per step (default: 1)
max_depth	int	No	Maximum search depth (default: 1)
end_gate	float	No	Value threshold for stopping (default: 0.9)
answer	str	No	Ground truth for verification

Outputs

Name	Type	Description
result_dict	dict	Keys: content, solution, summary, accurate, real_answer
root	Node	Root node of the constructed search tree

Usage Examples

from methods.tot.task import ToT_Task

task = ToT_Task(
    data="Solve: What is 15 * 23?",
    algorithm='dfs',
    branch=3,
    max_depth=4,
    end_gate=0.9,
    lang='en',
    answer="345"
)

result, root = task.run()
print(f"Solution: {result['summary']}")
print(f"Correct: {result['accurate']}")

Related Pages

Implements Principle

• Principle:Iamhankai_Forest_of_Thought_Tree_of_Thought_Search

Requires Environment

• Environment:Iamhankai_Forest_of_Thought_Python_CUDA_Runtime