Skip to content
PidocsOther

Explain how LLMs represent chain-of-thought reasoning

Session from pi-mono-sessions on Hugging Face by thomasmustier · MIT · condensed by Coders Talk

8ssession
0interventions
by library · 14h
GOAL

The developer wanted to understand how different LLMs signal their internal reasoning, specifically whether they use <thinking> tags.

  1. Prompt · +0
    How do LLMs indicate they're thinking? do they use <thinking>?
  2. Agent did · +0m
    Explained that LLMs use Chain-of-Thought reasoning, with some models like Claude, Gemini, and DeepSeek-R1 using tag-based formats such as <thought> or <thinking>, noting visibility depends on the interface.
  3. Outcome · +0m
    Developer received an explanation of tag-based chain-of-thought conventions across different LLMs.
VERDICT

This was just a quick informational question, not a coding task, so there's not much to correct here. If I asked again I'd probably ask for a comparison table of models and their exact tag conventions to save follow-up questions.

I tried this · 0

Discussion · 0

How Builds work