Same Name, Moving Target: Model Version Drift in Syntheses of Large Language Models for Mental Health
Bing Zhang Sr · 2026
Read Same Name, Moving Target: Model Version Drift in Syntheses of Large Language Models for Mental Health by Bing Zhang Sr (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/pmc-13671681.txt.
Description
Chatbots built on large language models are being tested for mental health care, and reviews have begun pooling their results. In a 2026 letter, Bing Zhang Sr raises a problem those reviews may have missed: the models change under the same name, so studies may not be testing the same tool.
Zhang notes that across one review's search window, GPT-4 shipped in several versions whose behavior differed in clear ways, including a sharp drop on a number task between two 2023 releases. Many studies recorded only the product name, not the version. Pooling such results, the letter argues, mixes unlike interventions and blurs what the combined numbers mean.
Catalog record
| CREATOR | Bing Zhang Sr |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 3 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |