Evolving Reasoning in Large Language Models: From Linear Chain-of-Thought to Graph-Based Diagram-of-Thought Frameworks

Author: Debabrata Pruseth
Publication Date: 2026/04/23
Document Type: Technical Note / Research Article
Language: English

Abstract

Large language models have shown strong capabilities in language generation, reasoning, coding, summarization, and decision support. However, their reasoning quality is highly sensitive to how problems are framed and how intermediate thinking is structured. This article presents a comparative research-style framework for understanding the evolution of reasoning-oriented prompting techniques: Chain-of-Thought, Tree-of-Thought, Logic-ofThought, Iteration-of-Thought, and Diagram-of-Thought. Chain-of-Thought prompting encourages linear step-by-step reasoning; Tree-of-Thought expands reasoning into multiple possible paths; Logic-of-Thought emphasizes rules, facts, and deductive structure; Iterationof-Thought introduces self-review and refinement; and Diagram-of-Thought organizes reasoning as an interconnected graph of ideas, critiques, dependencies, and synthesis. The paper argues that the movement from chains to trees to graphs reflects a broader shift in prompt engineering: from simple answer generation toward structured reasoning orchestration. It concludes that graph-based and hybrid prompting frameworks may provide more robust support for complex decision-making, strategy formulation, and multi-factor reasoning tasks.

Keywords
Large Language Models, Prompt Engineering, Chain-of-Thought, Tree-ofThought, Logic-of-Thought, Iteration-of-Thought, Diagram-of-Thought, AI Reasoning, Generative AI, Graph-Based Reasoning, LLM Reasoning Frameworks

Download Research PDF

Evolving Reasoning in Large Language Models: From Linear Chain-of-Thought to Graph-Based Diagram-of-Thought Frameworks

Suggested Citation
Pruseth, D. (2026). Evolving Reasoning in Large Language Models: From Linear Chain-ofThought to Graph-Based Diagram-of-Thought Frameworks. Debabrata Pruseth AI Blog.

Companion Note
This page provides the abstract and full-text PDF for the research version of the article. A companion blog post explains the same work in a more narrative and implementation-focused style.

Read the companion blog:
The Evolution of AI Thinking: From Chain of Thought to Diagram of Thought