Search papers, labs, and topics across Lattice.
4
0
7
7
Systematic misalignments in MLLM-generated captions can be detected with 63.8% accuracy, revealing a critical flaw in image-text pairing that has been largely overlooked.
Current AI models miss critical tumor detections in underrepresented demographics, revealing a hidden bias that could compromise patient outcomes.
Radiologist-in-the-loop AI gets a boost: CheXOne generates CXR reports that are as good as (or better than) those written by residents in over half the cases, thanks to its ability to explicitly reason about visual evidence.
Active data curation during pretraining lets you build a chest X-ray foundation model that rivals full-data models using just 23% of the data and compute.
