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

DiaHalu

Benchmarks

Task NameDataset NameSOTA ResultTrend
Binary Hallucination ClassificationDiaHalu (val)
Accuracy0.6224
6
Hallucination ClassificationDiaHalu sampled
Support245
5
Hallucination DetectionDiaHalu sampled (test)
Support (Total Samples)245
5
Multiclass Hallucination ClassificationDiaHalu (val)
Multiclass Accuracy54.68
5
Hallucination DetectionDiaHalu (sampled)
Precision (Class 0)87.7
1
Showing 5 of 5 rows