Reasoning Models Struggle to Control their Chains of Thought
Chen Yueh-Han NYU, MATS Robert McCarthy UCL, MATS Bruce W. Lee UPenn, MATS He He NYU Ian Kivlichan OpenAI Bowen Baker OpenAI Micah Carroll OpenAI Tomek Korbak OpenAI · 2026
Read Reasoning Models Struggle to Control their Chains of Thought by Chen Yueh-Han NYU, MATS Robert McCarthy UCL, MATS Bruce W. Lee UPenn, MATS He He NYU Ian Kivlichan OpenAI Bowen Baker OpenAI Micah Carroll OpenAI Tomek Korbak OpenAI (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-2603-05706.txt.
Description
Chen and team measure chain-of-thought control in models. Chain-of-thought watching offers spotting, but steerable wording could hurt it. Their CoT-Control suite limits orders, such as gene thinking without the word chromosome. Thinking models show far lower chain-of-thought than output control: Claude Sonnet controls its chain of thought only 2.7% of the time, versus over three-fifths for output. Control rises with size but falls with training, compute, and difficulty. The authors remain warily hopeful about watching and urge tracking control.
Catalog record
| CREATOR | Chen Yueh-Han NYU, MATS Robert McCarthy UCL, MATS Bruce W. Lee UPenn, MATS He He NYU Ian Kivlichan OpenAI Bowen Baker OpenAI Micah Carroll OpenAI Tomek Korbak OpenAI |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 89920 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.05706