Share your thoughts, 1 month free Claude Pro on usSee more

Factual mistake detection on DocENT (PoSh)

64.8Accuracy

Soft-MSD

Updated 26d ago

Evaluation Results

Method	Links
Soft-MSD 2026.05		64.8	0.259	0.195
Soft-MSD 2026.05		64.4	0.219	0.165
CLIPScore 2026.05		53.5	0.181	0.136
CLIPScore 2026.05		45.3	0.145	0.108
FLEUR 2026.05		41.2	0.04	0.031
FLEUR 2026.05		35.2	0.053	0.04