Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

VLSafe

Benchmarks

Task NameDataset NameSOTA ResultTrend
Safety EvaluationVLSafe Orig.
Unsafe Rate0.23
19
Direct MaliciousVLSafe OOD
ASR90.67
16
Safety EvaluationVLSafe Noise
Unsafe Rate0
13
Safety EvaluationVLSafe Blank
Unsafe Rate0
13
Harmlessness EvaluationVLSafe (test)
Relevance100
7
Showing 5 of 5 rows