Clinical Cognition Alignment for Gastrointestinal Diagnosis with Multimodal LLMs
Huan Zheng 1, Yucheng Zhou 1, Tianyi Yan 1, Dubing Chen 1, Hongbo Lu 2,3, Wenlong Liao 2, Tao He 2, Pai Peng 2, Jianbing Shen 1,† · 2026
Read Clinical Cognition Alignment for Gastrointestinal Diagnosis with Multimodal LLMs by Huan Zheng 1, Yucheng Zhou 1, Tianyi Yan 1, Dubing Chen 1, Hongbo Lu 2,3, Wenlong Liao 2, Tao He 2, Pai Peng 2, Jianbing Shen 1,† (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-2603-20698.txt.
Description
Zheng and co-authors present CogAlign, a framework that aligns multimodal models with clinical thinking for gut diagnosis. General models often reason off the path doctors use in endoscopy. They may also key on background cues, not the lesion itself. CogAlign first fine tunes on a tiered dataset of expert diagnostic steps. It then uses counterfactual training with masked lesion samples. Rewards push the model to ground calls in true lesion traits. Tests reached state of the art scores on several benchmarks.
Catalog record
| CREATOR | Huan Zheng 1, Yucheng Zhou 1, Tianyi Yan 1, Dubing Chen 1, Hongbo Lu 2,3, Wenlong Liao 2, Tao He 2, Pai Peng 2, Jianbing Shen 1,† |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 42444 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.20698