PromptRad: Knowledge-Enhanced Multi-Label Prompt-Tuning for Low-Resource Radiology Report Labeling
Ying-Jia Lin, Tzu-Chin Lo, Ping-Chien Li, Chi-Tung Cheng, Chien-Hung Liao, Hung-Yu Kao · 2026
Read PromptRad: Knowledge-Enhanced Multi-Label Prompt-Tuning for Low-Resource Radiology Report Labeling by Ying-Jia Lin, Tzu-Chin Lo, Ping-Chien Li, Chi-Tung Cheng, Chien-Hung Liao, Hung-Yu Kao (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-2605-20052.txt.
Description
Ying-Jia Lin and five co-workers label radiology reports with little training data using a system called PromptRad. It reframes multi-label classification as masked language modelling and enriches category meanings with synonyms from the UMLS Metathesaurus in a multi-word verbalizer. Fine-tuning the language model without extra classification layers needs far less labelled data than standard fine-tuning. On liver CT reports it beats dictionary and fine-tuning baselines with only thirty-two labelled examples, and it handles complex negation patterns better than existing methods.
Catalog record
| CREATOR | Ying-Jia Lin, Tzu-Chin Lo, Ping-Chien Li, Chi-Tung Cheng, Chien-Hung Liao, Hung-Yu Kao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40717 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.20052