Are Your Reasoning Models Reasoning or Guessing? A Mechanistic Analysis of Hierarchical Reasoning Models
Zirui Ren, Ziming Liu · 2026
Read Are Your Reasoning Models Reasoning or Guessing? A Mechanistic Analysis of Hierarchical Reasoning Models by Zirui Ren, Ziming Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-10679.txt.
Description
Zirui Ren and Ziming Liu ask whether a new kind of reasoning model truly reasons. The hierarchical reasoning model, or HRM, scores far above large language model reasoners on many tasks. Such a gap calls for a close look inside. The authors run a mechanistic study of how HRM works through problems.
Their study turns up three surprising facts. One is that HRM can fail on extremely simple puzzles. That mix of high scores and basic lapses hints at shortcut behavior in places. The paper gives researchers a map of where HRM is strong and where it may only guess. It warns against trusting benchmark scores alone.
Catalog record
| CREATOR | Zirui Ren, Ziming Liu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-10679