MT-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MT-Bench (val)
7.925Score
3
MT-Bench-101
8.5Grounding Score
3
MT-Bench Turn-2
1.9Writing Score
3
MT-Bench Factual
56.26Original Ratio
2
MT-Bench Math
55.14Original Ratio
2
MT-Bench Coding
53.88Original Ratio
2
MT-Bench
9.43Score
2
MT-Bench (test)
44.8Win Rate
2
MT-Bench (full set)
9.3Accuracy (%)
2
MT-Bench German
73.9Winrate
2
MT-Bench English
71Winrate
2
MT-Bench
177.54Throughput
2
MT-Bench
82.7Accuracy
2
MT-Bench (test)
82.5Win Rate
2
MT-Bench TH
7.89Overall Score
2
MT-Bench EN
8.69Score
2
MT-Bench Mixture Sampling
77.99Original Success Ratio
1
MT-Bench non-tie
84.34Win Rate
1
MT-Bench German SFT
63.9HATified SFT Score
1
MT-Bench English SFT
63.3HATified SFT Score
1