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

Safety Alignment on SORRY-Bench

10.22ASR

LED-Merging

7.012828.661450.3171.9586Feb 24, 2025
Updated 22d ago

Evaluation Results

MethodLinks
2025.02
10.22-
2025.02
10.44-
2025.02
11.33-
2025.02
11.78-
2025.02
12.44-
2025.02
12.67-
2025.02
15.11-
2025.02
17.11-
2025.02
17.78-
2025.02
18.67-
2025.02
24.22-
2025.02
24.67-
2025.02
28.89-
2025.02
31.11-
2025.02
32.44-
2025.02
34.67-
2025.02
35.56-
2025.02
35.78-
2025.02
36.89-
2025.02
37.11-
2025.02
37.78-
2025.02
38.22-
2025.02
38.44-
2025.02
39.55-
2025.02
40-
2025.02
40.22-
2025.02
40.44-
2025.02
40.44-
2025.02
40.44-
2025.02
43.56-
2025.02
46.44-
2025.02
46.44-
2025.02
50.6-
2025.02
53.11-
2025.02
56.67-
2025.02
69.55-
2025.02
69.56-
2025.02
70.89-
2025.02
79.78-
2025.02
90.4-
2026.05
-3.18
2026.05
-3.18
2026.05
-3.48
2026.05
-3.26
2026.05
-2.74
2026.05
-3.04
2026.05
-2.96
2026.05
-2.22
2026.05
-2.74
2026.05
-1.71
2026.05
-10.45