RM-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
RM-Bench
96Accuracy
137
RM-Bench
85.3Chat Score
99
RM-Bench (test)
84.11Chat Score
72
RM Bench Code
0.154EF
52
RM-Bench Chat
78.5Accuracy
42
RM-Bench Chat Hard
83.3Accuracy
34
RM-Bench Hard
0.709Average Accuracy
32
RM-Bench v1.0 (test)
74.3Overall Score
29
RM Bench Math
-0.077EF
26
RM Bench Safety-accept
0.698EF
26
RM Bench Chat
0.313EF
26
RM-Bench Chat subset Normal
86Accuracy
16
RM-Bench (full)
83Chat Score
11
RM-Bench
87.8Accuracy
10
RM-Bench Normal
80Accuracy
10
RM-Bench Easy
92.2Accuracy
10
RM-Bench 1k
73.5Positional Consistency
8