List your company

Long-form Question Answering refinement on ASQA (test)

16.63Error Rate (%)EIR
15.894420.859725.82530.7903Jul 16, 2024

Evaluation results

MethodLinks
EIRFeedback=Fine-grained (EIR)2024.0716.630.510.730.820.77
GenericFeedback=Coarse-grained (GENERIC)2024.0718.670.610.720.750.74
ImproveFeedback=Coarse-grained (IMPROVE)2024.0720.850.680.70.710.7
BaselineType=Original dataset instances2024.0734.811.2———
Zero-shotModel=LLaMA2-13B-chat, Prompting=Zero-shot2024.0735.021.080.50.620.55