Before You Read "Graph of Thoughts"

Introduction 👋

You’re here because you’re about to read the Graph of Thoughts (GoT) paper or attend a presentation about it. Either way, jumping straight into GoT may not be ideal without a primer like this one. I hope this article gives you what you need before reading the paper.

This article is your mental calibration. We’ll cover:

What came before GoT
What those methods tried to solve
Why they weren’t enough
The mathematical intuition underlying the methods

By the time graph-based reasoning starts, you’ll already have the tools to understand it.

Let’s Start Simple: Prompting Basics

🟣 Basic Input-Output (IO)

This is the most basic interaction with a language model:

Prompt → Output

There are no intermediate steps, branching logic, or reasoning trail you can inspect.

🧮 Mathematical Thinking

You can think of this as a black-box function:

You provide an input, get an output, but have no visibility into the reasoning in between. The model’s internals are opaque — all we see is how inputs map to outputs.

While the function is stochastic (outputs may vary), it often behaves as if deterministic, producing consistent completions for common prompts due to high-probability token paths.

You provide input x, and the model returns f(x) — typically the most probable sequence continuation given its training.