Do LLMs Know What They Know? Measuring Metacognitive Efficiency with Signal Detection Theory
Jon-Paul Cacioli · 2026
Read Do LLMs Know What They Know? Measuring Metacognitive Efficiency with Signal Detection Theory by Jon-Paul Cacioli (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-2603-25112.txt.
Description
Cacioli asks whether language models know what they know, using signal detection tools. Standard scores mix two skills: being right, and sensing when one is right. He treats token log-probability as a graded confidence signal. A new measure, normalized metacognitive information, rates how well that signal tracks truth. Across four models and 224,000 quiz trials, the measure varies twofold and does not follow accuracy. It also forecasts gains from refusing unsure questions. This version fixes a length bias in the answer checker, checked against human judges.
Catalog record
| CREATOR | Jon-Paul Cacioli |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 32686 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.25112