Misconception Acquisition Dynamics in Large Language Models
Naiming Liu ∗, Xinghe Chen ∗, Richard Baraniuk, Mrinmaya Sachan, Shashank Sonkar · 2026
Read Misconception Acquisition Dynamics in Large Language Models by Naiming Liu ∗, Xinghe Chen ∗, Richard Baraniuk, Mrinmaya Sachan, Shashank Sonkar (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-2604-00818.txt.
Description
Liu and colleagues study how large language models pick up misconceptions, comparing a student model trained on one misconception with a tutor model trained on several. Using their MalAlgoLib library of algebra problems with correct and misconception-laden solution traces, they find the student model overapplies its single error. The tutor model juggles several without losing correct-solving accuracy. A key bottleneck is intermediate reasoning: final-answer supervision alone teaches neither model where errors enter.
Catalog record
| CREATOR | Naiming Liu ∗, Xinghe Chen ∗, Richard Baraniuk, Mrinmaya Sachan, Shashank Sonkar |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35584 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.00818