RewardBench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
RewardBench
96.7Safety Score
284
RewardBench
97.8Accuracy
166
RewardBench 2
89.89Precise IF Score
90
RewardBench v1.0 (test)
0.978Average Score
89
RewardBench Focus 2
90.3Accuracy
82
RewardBench v2
92.1Accuracy
72
RewardBench Precise IF 2
57.5Accuracy
70
RewardBench v2 (test)
86.5Average Score
67
RewardBench Average 2
39.7Accuracy
52
RewardBench Math 2
35.7Accuracy
52
Multimodal RewardBench
60.7Accuracy
50
RewardBench Multimodal
99.6Safety Score
44
VL RewardBench
80.75Accuracy
42
RewardBench Chat
96.4Accuracy
42
RewardBench (full)
99.2Chat Score
41
RewardBench
88.8Accuracy
36
RewardBench 1.0 (test)
0.54Rstd
36
RewardBench
83.25Kappa
36
RewardBench
92.9Accuracy
31
RewardBench 2
89.5Accuracy
30
RewardBench
93.7Accuracy
29
RewardBench v1
95.5Accuracy
28
RewardBench
92.2Accuracy
25
RewardBench (test)
0.933RWBench
25
RewardBench latest (test)
74.9Accuracy
24