AIR-Bench
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Air-Bench
90Score
18
AIR-Bench 1.0 (test)
65.1Score
13
Air-Bench 2024
91.7Safety Score
12
AIR-Bench
0.66Average Score
12
AIR-Bench Speech (test)
71.45Emotion Acc
11
AIR-Bench
7.54Score (Speech Domain)
11
AIR-Bench
29.9SER
10
AIR-Bench English 24.04
65.5Wiki Score
10
AIR-Bench Foundation
63.2Speech Performance
9
AIR-Bench Music
6.16Score
8
AIR-Bench Speech
47.16Emotion Acc (MELD)
8
AIR-Bench Sound
7.01Score
8
AIR-Bench Foundation
36.8Accuracy
8
AIR-Bench Text + Image (test)
83Precision
8
AIR-Bench Image Only (test)
94Precision
8
AIR-Bench Text Only (test)
94Precision
8
AIR-Bench Music 1.0 (test)
65.8Inst. Classification Acc
7
AIR-Bench Speech Foundation
5,920Speech Grounding
7
AIR-Bench 1.0 (test)
7.18Overall Score
7
AIR-Bench 64
3.12Harm Score
6
AIR-Bench
66.25Overall Score
6
Air-Bench
0.905Accuracy
6
AIR-Bench
0.71Spearman's Rho
5
AIR-Bench 64 harmful prompts
1.13Harm
4
AIR-Bench foundation
64.85Accuracy
4