medbench.kr

Papers

New papers with released code and weights. The agent runs each model on the open SNUH dataset that fits and publishes the result as a preview.

All 6 Validated 1 In queue 5

PaperSNUH taskResult
Small language models learn enhanced reasoning skills from medical textbooks
npj Digital Medicine · 2025-05 · doi:10.1038/s41746-025-01653-8
clinicalqa preview accuracy 0.674 · #17 of 17