# When Chain-of-Thought Fails, the Solution Hides in the Hidden States

arXiv: https://arxiv.org/abs/2604.23351 · HTML: https://mlsanity.com/paper/2604.23351/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning.
- Builds on 5 of its 38 references that are in the map.
- First posted on arXiv 25 Apr 2026.

## Placed in

- LLM reasoning / [Prompting and chain-of-thought reasoning](https://mlsanity.com/lr/prompting-and-chain-of-thought-reasoning/)

## Builds on (most cited first, 5 of 5)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Measuring Faithfulness in Chain-of-Thought Reasoning](https://mlsanity.com/paper/2307.13702/) (2023)
- [Can Aha Moments Be Fake? Towards Quantifying Decorative and True Thinking in Chain-of-Thought](https://mlsanity.com/paper/2510.24941/) (2025)

## Built on by (most cited first, 3 of 3)

- [Mechanistic Interpretability of Chain-of-Thought Reasoning via Sequential Activation Patching](https://mlsanity.com/paper/2608.22332/) (2026)
- [Knowing Isn't Always Saying: When Do Spatial Encodings Reach Answers in Vision-Language Models?](https://mlsanity.com/paper/2608.22916/) (2026)
- [Are Stated Reasoning Steps Causally Load-Bearing?](https://mlsanity.com/paper/2609.27038/) (2026)

## Abstract (arXiv)

Whether intermediate reasoning is computationally useful or merely explanatory depends on whether chain-of-thought (CoT) tokens contain task-relevant information. We present a mechanistic causal analysis of CoT on GSM8K using activation patching: transferring token-level hidden states from a CoT generation to a direct-answer run for the same question, then measuring the effect on final-answer accuracy. Across models, generating after patching yields substantially higher accuracy than both direct-answer prompting and the original CoT trace, revealing that individual CoT tokens can encode sufficient information to recover the correct answer, even when the original trace is incorrect. This task-relevant information is more prevalent in correct than incorrect CoT runs and is unevenly distributed across tokens, concentrating in mid-to-late layers and appearing earlier in the reasoning trace. Moreover, patching language tokens such as verbs and entities carry task-solving information that steers generation toward correct reasoning, whereas mathematical tokens encode answer-proximal content that rarely succeeds. Patched outputs are often shorter and yet exceed the accuracy of a full CoT trace, suggesting complete reasoning chains are not always necessary. Together, these findings demonstrate that CoT encodes recoverable, token-level problem-solving information, offering new insight into how reasoning is represented and where it breaks down.