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

Safety Evaluation on SorryBench (FFR Metrics)

54.5Reasoning Success Rate (FFR)

No Defense

-2.1812.53527.2541.965Aug 6, 2025
Updated 27d ago

Evaluation Results

MethodLinks
2025.08
54.575-
2025.08
43.286.4-
2025.08
38.659.1-
2025.08
34.152.3-
2025.08
34.159.1-
2025.08
2545.5-
2025.08
2545.5-
2025.08
19.740.9-
2025.08
18.940.1-
2025.08
18.227.3-
2025.08
18.227.6-
2025.08
15.938.6-
2025.08
15.236.4-
2025.08
15.124.2-
2025.08
13.634.1-
2025.08
13.638.6-
2025.08
13.66.8-
2025.08
13.629.5-
2025.08
13.636.4-
2025.08
11.434.1-
2025.08
9.127.3-
2025.08
9.118.2-
2025.08
9.128-
2025.08
6.811.4-
2025.08
6.823.5-
2025.08
4.527.2-
2025.08
4.531.8-
2025.08
2.315.9-
2025.08
2.329.5-
2025.08
2.327.3-
2025.08
013.6-
2025.08
022.7-
2025.08
--72.7
2025.08
--13.6
2025.08
--20.5
2025.08
--77.3
2025.08
--38.6
2025.08
--70.5
2025.08
--34.1
2025.08
--47.7
2025.08
--22.7
2025.08
--77.3
2025.08
--9.1
2025.08
--79.5
2025.08
--40.9
2025.08
--79.5
2025.08
--47.7
2025.08
--40.9
2025.08
--20.5