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

Preference Modeling on Helpfulness & Harmlessness

72.2Accuracy

BTPO

54.20858.87963.5568.221Oct 17, 2025
Updated 2mo ago

Evaluation Results

MethodLinks
2025.10
72.2
2025.10
70.5
2025.10
70.4
2025.10
69.2
2025.10
68.7
2025.10
68.7
2025.10
68.5
2025.10
67.9
2025.10
67.4
2025.10
67.4
2025.10
67.1
2025.10
67.1
2025.10
65.3
2025.10
64.7
2025.10
63.2
2025.10
61
2025.10
59.3
2025.10
58.7
2025.10
55.8
2025.10
54.9