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

Over-refusal evaluation on Poisoned GPT4-LLM subset

35.7Informative Refusal Rate

QCB

-1.0128.51918.0527.581Jun 28, 2026
Updated 25d ago

Evaluation Results

MethodLinks
2026.06
35.7
2026.06
34.2
2026.06
31.3
2026.06
30.2
2026.06
26.1
2026.06
23.8
2026.06
12.6
2026.06
12.3
2026.06
11
2026.06
1.07
2026.06
1
2026.06
1
2026.06
1
2026.06
0.86
2026.06
0.86
2026.06
0.8
2026.06
0.8
2026.06
0.8
2026.06
0.73
2026.06
0.73
2026.06
0.67
2026.06
0.67
2026.06
0.67
2026.06
0.6
2026.06
0.4
2026.06
0.4
2026.06
0.4