MT-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
MT-Bench Korean
8.87Score
6
MT-Bench
77.82Original Accuracy
6
MT-Bench Real-data 20% verified
77.82Original Accuracy
6
MT-Bench
9.5Score (GPT-4-Turbo)
6
MT-bench
100Acc MCA
6
MT-Bench (test)
6.26MT-Bench Score
6
MT-Bench ja
5.68MT-Bench (ja) Score
6
MT-Bench
71.7Writing Score
6
MT-Bench benign prompts
41.56Average Time Cost
6
MT-Bench-101
93.62Normalized Score
5
MT-Bench
8.99MT-Bench Score
5
MT-Bench 1.0 (test)
8.538GPT-4 Score
5
MT-Bench v1 (test)
8.41MT-Bench Score
5
MT-Bench speech-interaction Turn one and two averaged
7.08Reasoning Score
4
MT-Bench
3.69SGLang Acceptance Length
4
MT-Bench Translation, Summarization, QA, Math, RAG Vicuna-7B-v1.3
3.68MT Score
4
MT-bench on Llama3.1-8B-Instruct v1 (test)
3.19MT Score
4
MT-Bench
194TPS (tok/s)
4
MT-Bench
1.28MAT
4
MT-Bench (test)
28.4ROUGE-1 Score
4
MT-Bench OOD
44.3Win Rate
4
MT-Bench 101
10CM Score
3
MT-Bench
100Top-1 Accuracy
3
MT-Bench
9.11MT-Bench Score
3
MT-Bench
4.5Reasoning Score
3