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

Misalignment Detection on WorkBench (n=326)

52Average IR

ProvePrompt

2.91215.65628.441.144May 1, 2026
Updated 22d ago

Evaluation Results

MethodLinks
2026.05
520.1061
2026.05
49.50.0441
2026.05
45.2--
2026.05
44.40.0711
2026.05
43.10.0371
2026.05
41.4--
2026.05
40.70.0250.978
2026.05
39.4--
2026.05
38.3--
2026.05
37.50.1331
2026.05
37.3--
2026.05
29.40.0961
2026.05
27.50.0070.659
2026.05
26.90.1161
2026.05
26.7--
2026.05
24.2--
2026.05
19.8--
2026.05
18.1-0.2020.001
2026.05
17.2-0.2420.001
2026.05
16.9--
2026.05
15.3--
2026.05
15.2-0.0020.479
2026.05
14.2-0.2310.001
2026.05
13.9-0.0310.006
2026.05
13.2-0.320.001
2026.05
12.5-0.0730.011
2026.05
11.9-0.1480.001
2026.05
11-0.1320.001
2026.05
10-0.0690.005
2026.05
4.8-0.3460.001