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

Safety Boundary Over-Refusal on ORBench

7.2ORBench Over-Refusal Rate

Mistral-7B-v2

4.1824.56544.9565.335May 23, 2026
Updated 8d ago

Evaluation Results

MethodLinks
2026.05
7.2
2026.05
9.6
2026.05
14.3
2026.05
14.5
2026.05
17.4
2026.05
19.3
2026.05
32.3
2026.05
35
2026.05
37.9
2026.05
39.5
2026.05
42.8
2026.05
45.4
2026.05
45.5
2026.05
45.7
2026.05
46.1
2026.05
48.4
2026.05
48.8
2026.05
52
2026.05
57.2
2026.05
82.7