ResearchBenchmarksEvidence-Grounded Extractive QA on CellularSpecSec-Bench Stage 2Follow0.968Score 2CellSpecSec-ARI0.368960.524480.680.83552Jan 19, 2026Evaluation ResultsMethodMethodLinksScore 2Evidence CorrectnessCellSpecSec-ARI2026.010.96896.4BASE2026.010.3920