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

Step-level reasoning verification on Calendar

69.8PR-AUC

ReProbe, Hidden States, GPT-OSS-anno

29.34439.84750.3560.853Nov 9, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.11
69.8
2025.11
67.6
2025.11
63.7
2025.11
59.4
59.2
2025.11
56.8
2025.11
55.5
2025.11
50.8
2025.11
49.7
2025.11
48.9
2025.11
43.4
2025.11
38.6
2025.11
38
2025.11
36.2
2025.11
34.3
2025.11
32.7
2025.11
32.4
2025.11
32.3
2025.11
30.9