Large language models can disambiguate opioid slang on social media
Kristy A. Carpenter 1 , Issah A. Samori 2 , Mathew V. Kiang 3 , Keith Humphreys 4,8 , Anna Lembke 4 , Johannes C. Eichstaedt 5 , and Russ B. Altman †,1,2,6,7 · 2026
Read Large language models can disambiguate opioid slang on social media by Kristy A. Carpenter 1 , Issah A. Samori 2 , Mathew V. Kiang 3 , Keith Humphreys 4,8 , Anna Lembke 4 , Johannes C. Eichstaedt 5 , and Russ B. Altman †,1,2,6,7 (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-2603-10313.txt.
Description
Kristy A Carpenter and co-authors test whether large language models can tell real opioid slang from harmless uses of the same words on social media. Words like smack and blues often mean other things, so simple word lists flag far too much. Four leading models were tried with a lexicon, without one, and on simulated new slang. Each model beat the best lexicon strategy by a wide margin, with higher recall as well. The authors say such tools can clean data for work on low-prevalence topics.
Catalog record
| CREATOR | Kristy A. Carpenter 1 , Issah A. Samori 2 , Mathew V. Kiang 3 , Keith Humphreys 4,8 , Anna Lembke 4 , Johannes C. Eichstaedt 5 , and Russ B. Altman †,1,2,6,7 |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43659 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.10313