BeaverTails
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
BeaverTails Unsafe Animal Abuse
85Safe Rate
4
BeaverTails (Average)
87Safety Score
4
beavertails
40.6Macro Accuracy
4
BeaverTails
60.9Harm Score (HS)
4
BeaverTails 3k (test)
94.7Animal Safety Classification Accuracy
4
BeaverTails V
0.631F1 Score
4
Beavertails-A Audio
0.997Safety Score
4
BeaverTails
10.39ASR
4
BeaverTails
0.639Harm Score Delta (Δ)
3
BeaverTails
96.2F1 Score
2
BeaverTails 5,000 prompts (subsampled)
4.68Relevance
2
BeaverTails V
—Reward Mean
0