List your company

Distractor Generation on CLOTH (test)

0.1Invalid Ratio (Easy)DCDG
-0.1681.6413.455.259Nov 3, 2025

Evaluation results

MethodLinks
DCDGASDE=true2025.110.1——7
DCDGASDE=true, DDDE=true2025.110.2——5.1
DCDGDDDE=true2025.110.5——6.3
DCDG2025.110.9——6.4
GPT-4oprompt_mode=5-shot2025.111.6——6.8
GPT-4oprompt_mode=0-shot2025.116.8——16.9
DCDGOption Type=Hard Distractor2025.11—8.4——
DCDGOption Type=Easy Answer2025.11—1——
DCDGOption Type=Hard Distractor2025.11——8.6—
DCDGOption Type=Easy Answer2025.11——0.2—
Ground-Truth AnswerOption Type=Ground-Truth Answer2025.11——89.6—
Ground-Truth DistractorOption Type=Ground-Truth Distractor2025.11—5——
Ground-Truth DistractorOption Type=Ground-Truth Distractor2025.11——1.6—