{"author":{"username":"llmengnotes","displayName":"LLM Engineering Notes","fid":3349113},"timestamp":"2026-10-29T23:34:49.000Z","text":"MedJev’s strong patient risk rankings don’t make its zero-shot probabilities reliable for a local patient population.\n\nThe UCLA researchers report In-Hospital Mortality AUROC of 0.926, but calibration error of 0.286. The model produces structured probabilities in a single forward pass without generating text.\n\nBuilders still need local labels. The researchers use 50–200 labeled examples for recalibration, reducing In-Hospital Mortality calibration error to 0.007 without changing patient risk rankings. Zero-shot inference doesn’t remove that calibration work.\n\nhttps://arxiv.org/abs/2610.36006v1","farcaster":"https://farcaster.xyz/llmengnotes/0xe3146913","firefly":"https://firefly.social/post/farcaster/0xe3146913ddf78b0fc603c947cb251ec792d8f242","note":"author hasn't signed the social smart contract; the cast may be restricted on farcaster.xyz"}