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

LLM-as-a-Judge Performance on MT-Bench (test)

2.24StdDev

DI

1.27087.812914.35520.8971Oct 20, 2024
Updated 1mo ago

Evaluation Results

MethodLinks
2024.10
2.2482.26
2024.10
2.8982.29
2024.10
3.2382.27
2024.10
3.5781.93
2024.10
3.8482.1
2024.10
4.7281.05
2024.10
4.8481.23
2024.10
4.9981.25
2024.10
5.272.98
2024.10
5.272.98
2024.10
5.2270.63
2024.10
5.2270.63
2024.10
5.2481.23
2024.10
5.7681.32
2024.10
5.7980.82
2024.10
6.0172.86
2024.10
6.0172.86
2024.10
6.2668.64
2024.10
6.2668.64
2024.10
6.6767.86
2024.10
6.7565.09
2024.10
6.7680.52
2024.10
7.0169.84
2024.10
7.6672.67
2024.10
7.6672.67
2024.10
8.2368.48
2024.10
8.769.09
2024.10
9.4266.79
2024.10
9.7267.74
2024.10
11.6470.63
2024.10
11.6470.63
2024.10
12.6367.56
2024.10
13.7267.47
2024.10
14.7367.58
2024.10
16.3165.29
2024.10
16.3165.29
2024.10
16.3170.49
2024.10
16.4267.08
2024.10
16.4267.08
2024.10
17.2465.61
2024.10
17.2465.61
2024.10
19.0962.18
2024.10
20.2368.67
2024.10
20.8664.6
2024.10
26.4758.21