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

Reward Modeling Evaluation on RewardBench2 (test)

82.9Accuracy

ACE + Batching

61.89267.34672.878.254Apr 3, 2026
Updated 13d ago

Evaluation Results

MethodLinks
2026.04
82.9
2026.04
82.8
2026.04
82.5
2026.04
82.4
2026.04
81.5
2026.04
80.4
2026.04
79.2
2026.04
79.2
2026.04
77.8
2026.04
77.2
2026.04
76.5
2026.04
75.9
2026.04
71.2
2026.04
69.9
2026.04
69.8
2026.04
67.2
2026.04
65.4
2026.04
63.6
2026.04
63.5
2026.04
62.7