List your company

Question Answering on HotpotQA (Leakage Error, Answerability Accuracy)

1.4Leakage ErrorSeedRG
-0.54412.57825.738.822May 9, 2026

Evaluation results

MethodLinks
SeedRGModel=GPT-52026.051.441.8
SeedRGModel=Gemini 2.5 Flash2026.054.147.3
SeedRGModel=Claude Sonnet 4.52026.0513.559.5
Direct Generation (DG)Model=Gemini 2.5 Flash2026.051846
Direct Generation (DG)Model=GPT-52026.051950
Direct Generation (DG)Model=Claude Sonnet 4.52026.053147
Original BenchmarkModel=Gemini 2.5 Flash2026.0531.150
Original BenchmarkModel=Claude Sonnet 4.52026.0545.947.3
Original BenchmarkModel=GPT-52026.055033.8