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

Reward Modeling Evaluation on Reward Bench Safety 2

72.3Pairwise Accuracy

Distribution-Calibrated Aggregation

61.3864.21567.0569.885Dec 2, 2025
Updated 1mo ago

Evaluation Results

MethodLinks
2025.12
72.3
2025.12
69.1
2025.12
65
2025.12
63.5
2025.12
63.3
2025.12
63
2025.12
62.9
2025.12
62.6
2025.12
62.5
2025.12
62.3
2025.12
61.9
2025.12
61.8