MT-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MT-Bench
2.14Speedup
16
MT-Bench Arena-derived reference ranking 1.0 (test)
121Coding Score
16
MT-Bench
2.717Acceptance Length (τ)
16
MT-Bench
3.87Speedup
16
MT-Bench
6.55Score
16
MT-Bench
100Coverage
16
MT-Bench (test)
100Accuracy
16
MT-Bench
7.33MTB Score
15
MT-Bench UltraChat
6.21Quality Score
15
MT-Bench ShareGPT
6.12Score
15
MT-Bench Coding
9First Turn Score
15
MT-Bench en
7.67MT-Bench Score (10)
15
MT-Bench OOD
8.64R1 Score
14
MT-Bench
4.76Speedup
14
MT-Bench
7.29MAT Score
14
MT-Bench GPT-4 scored (test)
6.31Score
14
MT-Bench High-Disagreement (Top 20%)
8.72Human Score
13
MT-Bench Overall
8.17Human Score
13
MT Bench
91.4Accuracy
13
Ko-MT-Bench
8.21Overall Score (1-10)
12
MT-Bench
4.55Speedup
12
MT-Bench (test)
8.06MT-Bench Score (1-turn)
12
MT-Bench
0.74Spearman Correlation (ρ)
12
MT-Bench
94.1Response Similarity
12
MT-Bench
4.18Speedup
12