Generating Reports or Repeating Templates? Measuring and Mitigating Template Collapse in 3D CT Report Generation
Tom Maye-Lasserre, Yitong Li, Bailiang Jian, Morteza Ghahremani, Benedikt Wiestler, Christian Wachinger · 2026
Read Generating Reports or Repeating Templates? Measuring and Mitigating Template Collapse in 3D CT Report Generation by Tom Maye-Lasserre, Yitong Li, Bailiang Jian, Morteza Ghahremani, Benedikt Wiestler, Christian Wachinger (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-2605-30984.txt.
Description
Maye-Lasserre and colleagues diagnose and fix template collapse in 3D CT reports. Modern vision-language models write fluent text yet detect few problems, falling to forms that under-report rare findings. This failure stems from limited data, label skew, and weak signals from volume inputs. Their CLarGen splits what to say from how to say it. It joins a query model for problem spotting, recall of matched cases, and a medical language model to write the report. CLarGen lifts medical scores while keeping reports fluent.
Catalog record
| CREATOR | Tom Maye-Lasserre, Yitong Li, Bailiang Jian, Morteza Ghahremani, Benedikt Wiestler, Christian Wachinger |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 55900 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.30984