ResearchBenchmarksEvidence-Grounded Abstractive QA on CellularSpecSec-Bench Stage 2Follow96.8Score (Stage 2)CellSpecSec-ARI27.32845.36463.481.436Jan 19, 2026Evaluation ResultsMethodMethodLinksScore (Stage 2)Evidence Correct CountCellSpecSec-ARI2026.0196.896BASE2026.01300