CLIMB: Data Foundations for Large Scale Multimodal Clinical Foundation Models

Wei Dai, Peilin Chen, Malinda Lu, Daniel Li, Haowen Wei, Hejie Cui, Paul Pu Liang · 2025

Read CLIMB: Data Foundations for Large Scale Multimodal Clinical Foundation Models by Wei Dai, Peilin Chen, Malinda Lu, Daniel Li, Haowen Wei, Hejie Cui, Paul Pu Liang (2025) 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-2503-07667.txt.

Description

Wei Dai and co-authors introduce CLIMB in 2025, a large benchmark for clinical AI that spans many data types. Progress in medical AI has clustered on a few kinds of data, which blocks models that judge a patient's whole state. CLIMB unifies imaging, language, time series, and graph data into one test bed for training and evaluation.

The benchmark holds 4.51 million patient samples totaling nineteen terabytes, across flat images, 3D video, signals, graphs, and mixed data. Broad multitask pre-training lifted results on neglected areas by up to twenty-nine percent for ultrasound and twenty-three percent for ECG tasks over single-task learning. Pre-training also transferred well to new tasks, and strong single-type encoders paired well with the right fusion methods.

Catalog record

CREATORWei Dai, Peilin Chen, Malinda Lu, Daniel Li, Haowen Wei, Hejie Cui, Paul Pu Liang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE140 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-07667

Browse the full library · About the Tower of Babel