TreeOfThought Decorator¶

Organizes the response as a branching exploration of multiple reasoning paths. This decorator enables the AI to consider several possible approaches or hypotheses simultaneously, exploring the implications of each before reaching conclusions.

Category: Reasoning

Parameters¶

Examples¶

Multi-branch problem solving for a complex question¶

+++TreeOfThought
What might explain the Fermi Paradox?

Explores three different reasoning branches about potential explanations for the Fermi Paradox, developing each path to moderate depth

Deep, focused exploration with pruning¶

+++TreeOfThought(branches=5, depth=5, pruning=true)
How might we solve the climate change crisis?

Starts with five different approaches to climate change, explores each in depth, and eliminates less promising branches to focus on the most viable solutions

Model-Specific Implementations¶

gpt-4o¶

Instruction: Consider {branches} distinct approaches to this problem. For each approach, think step-by-step to a depth of {depth} levels. Label each branch clearly (e.g., 'Approach 1', 'Approach 2') and use subheadings to indicate depth levels. {pruning}

Notes: This model handles complex branching reasoning well, but benefits from clear formatting instructions

gpt-4-turbo¶

Instruction: Explore {branches} different ways to approach this problem. Be very explicit about which approach you're discussing, and break down each approach into {depth} levels of detailed analysis. Start each approach with a clear label and keep them separated.

Notes: This model may need additional structure to manage multiple reasoning branches effectively

Implementation Guidance¶

Multi-branch exploration of a scientific question¶

Original Prompt:

What might explain the Fermi Paradox?

Transformed Prompt:

Please explore multiple reasoning paths simultaneously for this problem, considering different approaches or hypotheses before reaching conclusions. Generate 3 distinct reasoning branches, each exploring a different approach or hypothesis. For each branch, pursue the reasoning to a depth of 3 levels of analysis. Fully explore all branches to their complete depth regardless of their apparent promise.

What might explain the Fermi Paradox?

Deep analysis with pruning for a complex problem¶

Original Prompt:

How might we solve the climate change crisis?

Transformed Prompt:

Please explore multiple reasoning paths simultaneously for this problem, considering different approaches or hypotheses before reaching conclusions. Generate 5 distinct reasoning branches, each exploring a different approach or hypothesis. For each branch, pursue the reasoning to a depth of 5 levels of analysis. Evaluate the promise of each branch as you go, and stop pursuing less promising paths to focus on the most viable ones.

How might we solve the climate change crisis?

Transformation Details¶

Base Instruction: Please explore multiple reasoning paths simultaneously for this problem, considering different approaches or hypotheses before reaching conclusions.

Placement: prepend

Composition Behavior: accumulate

Parameter Effects:

• branches:

• Format: Generate {value} distinct reasoning branches, each exploring a different approach or hypothesis.

• depth:

• Format: For each branch, pursue the reasoning to a depth of {value} levels of analysis.

• pruning:

• When set to true: Evaluate the promise of each branch as you go, and stop pursuing less promising paths to focus on the most viable ones.
• When set to false: Fully explore all branches to their complete depth regardless of their apparent promise.

Compatibility¶

• Requires: None
• Conflicts: None
• Compatible Models: gpt-4o, gpt-4-turbo
• Standard Version: 1.0.0 - 2.0.0

Related Decorators¶

• StepByStep: Enhances TreeOfThought StepByStep can help organize the reasoning within each branch of the tree
• Debate: Enhances TreeOfThought TreeOfThought and Debate work well together to explore multiple perspectives with structured reasoning