sebis at CRF Filling 2026: A Two-Stage Local LLM Pipeline for Medical CRF Filling
Researchers have developed a two-stage local LLM pipeline for medical CRF filling, utilizing the MedGemma-27B model. This approach addresses privacy concerns and inference costs associated with deploying LLMs in clinical settings. The pipeline achieved a macro-F1 score of 0.55 on the CL4Health 2026 English test track, securing second place among local, open-source submissions. AI
IMPACT Demonstrates the viability of privacy-preserving, on-premise LLM solutions for clinical NLP tasks.