---
title: "Study finds distinct internal signatures for each reasoning step"
url: https://www.parallelquant.com/posts/study-finds-distinct-internal-signatures-for-each-reasoning-step-21bc46
source_name: "The Decoder"
source_url: https://the-decoder.com/ai-models-written-reasoning-steps-correspond-to-distinct-internal-patterns-a-new-study-finds/
published: 2026-09-12T13:39:43.000Z
topics: ["research", "security"]
publisher: "Parallel Quant"
---

# Study finds distinct internal signatures for each reasoning step

*2026-09-12 · Source: [The Decoder](https://the-decoder.com/ai-models-written-reasoning-steps-correspond-to-distinct-internal-patterns-a-new-study-finds/)*

Researchers found that when AI models perform reasoning steps like calculation, formula retrieval, or deduction, each corresponds to a distinct, separable pattern in the model's internal states, especially in the middle layers.

**Why it matters:** This adds to evidence that a model's internal processing doesn't fully match what it writes in its visible chain-of-thought, reinforcing safety researchers' concerns that chain-of-thought monitoring alone may not reliably reveal what a model is actually doing.

**Topics:** research, security

---
Read the original: https://the-decoder.com/ai-models-written-reasoning-steps-correspond-to-distinct-internal-patterns-a-new-study-finds/
Canonical: https://www.parallelquant.com/posts/study-finds-distinct-internal-signatures-for-each-reasoning-step-21bc46
Published by Parallel Quant — https://www.parallelquant.com