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

Style-POPE

Benchmarks

Task NameDataset NameSOTA ResultTrend
Object Hallucination ProbingStyle-POPE Popular
F1 Score80.2
72
Object Hallucination ProbingStyle-POPE (Random)
Accuracy88.2
63
Object Hallucination ProbingStyle-POPE Adversarial
Accuracy75.2
63
Hallucination EvaluationStyle-POPE Overall Average
F1 Score79.1
9
Hallucination EvaluationStyle-POPE Random
F1 Score0.844
9
Hallucination EvaluationStyle-POPE Adversarial
F1 Score75.3
9
Showing 6 of 6 rows