Mistral-7B
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Mistral-7B model tree (test)
1Rank
21
Mistral-7B Generation Evaluation Set
97.01CA
15
Mistral-7B Generation (Evaluation Set)
96.38Control Accuracy (CA)
15
Mistral-7B black-box setting v0.3
98.4True Positive Rate (TPR)
10
Mistral-7B Long-context (8k window)
4.568Perplexity
8
Mistral-7B Long-context (4k window)
5.241Perplexity
8
Mistral-7B Jailbreak Evaluation
0GCG Attack Success Rate
6
Mistral-7B Permute and Flip v0.1 S1
91.6IoU
5
Mistral-7B (successful attacks)
3,021Unique Queries
3
Mistral-7B v0.3 (test)
34.2S-BLEU
3
Mistral-7B activations
95.3KL Score
2
Mistral-7B
5.36MSE
2