Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs
Sean Trott, Samuel Taylor, Cameron Jones, James A. Michaelov, Pamela D. Rivière · 2026
Read Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs by Sean Trott, Samuel Taylor, Cameron Jones, James A. Michaelov, Pamela D. Rivière (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-2602-16085.txt.
Description
Sean Trott leads a team of five researchers testing whether language statistics can explain false belief reasoning in language models. Most prior work used a small set of closed models, which limits tests of theories about human social cognition. The team replicates and extends false belief studies across forty-one open-weight models from distinct families, checking sensitivity to implied knowledge states.
Only about a third of the models show such sensitivity, and none fully explains away the effect seen in humans. Larger models show more sensitivity and stronger power to predict human responses. A new test finds that both humans and models lean toward attributing false beliefs when knowledge is cued with a non-factive verb rather than indirectly. The size of that cue effect in humans falls within the model range, suggesting language statistics could account for it, though not for the main knowledge effect.
Catalog record
| CREATOR | Sean Trott, Samuel Taylor, Cameron Jones, James A. Michaelov, Pamela D. Rivière |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-16085