Reasoning & CoT

15 frameworks — Chain-of-thought methods that guide AI through transparent, step-by-step logical reasoning.

Overview

Chain-of-Thought and related reasoning techniques guide AI through transparent, step-by-step logical processes. By making intermediate thinking visible, these methods dramatically improve accuracy on math, logic, and multi-step problems.

Use reasoning techniques when tasks require more than a simple lookup — when the AI needs to think through a problem, weigh evidence, or build an argument step by step. They range from the simplicity of “Let’s think step by step” to advanced approaches like memory pre-thinking and backward verification.

Reasoning & CoT 15

Chain-of-Thought

2022

Encourage step-by-step reasoning through intermediate thinking steps.

Still active technique

Zero-Shot CoT

2022

Add “Let’s think step by step” without examples for reasoning tasks.

Still active technique

Auto-CoT

2022

Automatically generate diverse reasoning demonstrations with clustering.

Still active technique

Contrastive CoT

2023

Provide both correct and incorrect reasoning examples for contrast.

Still active technique

Structured CoT

2023

Triple-phase reasoning: analysis, structured thinking, summarization.

Still active technique

Faithful CoT

2023

Symbolic reasoning combined with natural language for verifiable answers.

Still active technique

Complexity-Based

2023

Select the most complex reasoning paths for harder problems.

Still active technique

Tab-CoT

2023

Tabular format to organize step-by-step reasoning in columns.

Still active technique

Analogical Reasoning

2023

Solve problems by generating and applying relevant analogies.

Still active technique

Step-Back Prompting

2023

Abstract to higher-level concepts before solving specific problems.

Still active technique

Thread of Thought

2023

Segment and reason through lengthy context systematically.

Still active technique

Active Prompting

2023

Identify uncertain examples and annotate them for better demonstrations.

Still active technique

Memory of Thought

2023

Pre-think across unlabeled data before tackling target problems.

Still active technique

Reversing CoT

2023

Work backward from conclusions to verify reasoning validity.

Still active technique

Uncertainty-Routed CoT

2023

Apply chain-of-thought selectively based on model confidence for cost-optimal reasoning.

Still active technique

Algorithm of Thoughts

2024

Embed algorithmic search patterns like DFS and BFS directly into prompts for single-call solution-space exploration.

Still active technique

Buffer of Thoughts

2024

Maintain reusable thought-templates from past problem-solving for template-guided reasoning.

Still active technique

Maieutic Prompting

2022

Build recursive explanation trees and use logical consistency to verify commonsense reasoning.

Still active technique

Scratchpad Prompting

2021

Show intermediate computation steps in a dedicated workspace before producing final answers.

Still active technique

Selection-Inference

2023

Alternate between selecting relevant premises and drawing conclusions for structured logical reasoning.

Still active technique

Thought Propagation

2023

Solve analogous problems first and propagate insights to refine the target problem’s solution.

Still active technique

Chain-of-Symbol

2023

Replace natural language with compact symbolic representations for spatial and structural reasoning.

Still active technique

Symbolic CoT

2023

Integrate formal logic rules and symbolic expressions with chain-of-thought for verifiable reasoning.

Still active technique

Dual Process Prompting

2024

Route between fast intuitive responses and slow deliberate reasoning based on task complexity.

Still active technique

Chain of Knowledge

2023

Generate structured evidence triples with explanation hints for transparent, verifiable reasoning.

Still active technique

Chain of Abstraction

2024

Reason with abstract placeholders first, then fill in concrete values using domain tools.

Still active technique

Everything of Thoughts

2024

Combine reinforcement learning and Monte Carlo Tree Search for flexible, efficient reasoning.

Still active technique

Technique Comparison

Side-by-side comparison of all 15 frameworks in this category.

Technique Year Best For Key Strength Complexity
Chain-of-Thought 2022 Complex reasoning Transparent logic Low
Zero-Shot CoT 2022 Quick reasoning boost No examples needed Very Low
Auto-CoT 2022 Automated demos Diversity clustering Medium
Contrastive CoT 2023 Error reduction Contrast learning Medium
Structured CoT 2023 Formal reasoning Three-phase approach Medium
Faithful CoT 2023 Verifiable answers Symbolic + NL High
Complexity-Based 2023 Hard problems Path selection Medium
Tab-CoT 2023 Organized reasoning Tabular format Low
Analogical Reasoning 2023 Novel problems Analogy generation Medium
Step-Back Prompting 2023 Abstraction tasks Concept elevation Low
Thread of Thought 2023 Long contexts Sequential processing Medium
Active Prompting 2023 Uncertain examples Active learning High
Memory of Thought 2023 Pre-thinking Unlabeled data High
Reversing CoT 2023 Verification Backward reasoning Medium
Uncertainty-Routed CoT 2023 Cost optimization Selective reasoning Medium