Review
Aug 2026
Can You Trust the Confidence? ConfBench for Vision-Language Models on Document Extraction
ConfBench is introduced, the first calibration-specific benchmark for key information extraction (KIE), built by applying 20 controlled degradation pipelines to a diverse document set, yielding 1,346 variants and 70K+ entity-level evaluations spanning the full accuracy spectrum.
P. Roy, Sujitha Martin, Mohammad Rostami et al.
· 0 citations