Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failures

Yi Hu, Jiaqi Gu, Ruxin Wang, Zijun Yao, Hao Peng +4 more
1/11/2026
cs.CL

No official code found. Best implementation: AheadOFpotato/Awesome-LRM-Mechanisms (35 stars, MIT, last commit 2026-01-29).

Abstract

Reinforcement learning (RL) has catalyzed the emergence of Large Reasoning Models (LRMs) that have pushed reasoning capabilities to new heights. While their performance has garnered significant excitement, exploring the internal mechanisms driving these behaviors has become an equally critical research frontier. This paper provides a comprehensive survey of the mechanistic understanding of LRMs, organizing recent findings into three core dimensions: 1) training dynamics, 2) reasoning mechanisms, and 3) unintended behaviors. By synthesizing these insights, we aim to bridge the gap between black-box performance and mechanistic transparency. Finally, we discuss under-explored challenges to outline a roadmap for future mechanistic studies, including the need for applied interpretability, improved methodologies, and a unified theoretical framework.

View on arXivView PDF

Code Implementations(1)

AheadOFpotato/Awesome-LRM-MechanismsCommunity74%
MIT

Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failures

354Jan 11, 20268 months agoMIT

Similar Papers

A Survey of Reinforcement Learning for Large Reasoning Models

89% match1 repo2.5k2025

Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models

87% match1 repo32025

Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?

87% match2025

Large Reasoning Models are not thinking straight: on the unreliability of thinking trajectories

86% match2025

ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models

86% match2 repos1.5k2025

On the Interplay of Pre-Training, Mid-Training, and RL on Reasoning Language Models

86% match2025

Cite this paper

@article{hu2026towards,
  title  = {Towards a Mechanistic Understanding of Large Reasoning Models: A Survey of Training, Inference, and Failures},
  author = {Yi Hu and Jiaqi Gu and Ruxin Wang and Zijun Yao and Hao Peng and Xiaobao Wu and Jianhui Chen and Muhan Zhang and Liangming Pan},
  year   = {2026},
  eprint = {2601.19928},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2601.19928v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2601.19928.svg)](https://codeofpaper.com/paper/2601.19928)

Discussion