Task: summarizing discharge reports from mimic iv with llms and pretrained lms
original data from https://github.com/AI-in-Health/ClinicBench
- zero shot with custom prompt comparison to clinicbench's zero shot results
- finetuning, zero/few shot experiments for custom split of the original data (train:92 dev: 40 test: 250)
- scoring script from discarge-me shared task: https://github.com/Stanford-AIMI/discharge-me/tree/main/scoring