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

HalLoc

Benchmarks

Task NameDataset NameSOTA ResultTrend
Probability CalibrationHalLoc-Caption (test)
Object ECE0.04
12
Token-level hallucination detectionHalLoc Caption
Object Precision66
7
Token-level hallucination detectionHalLoc Instruct
Object Precision94
7
Token-level hallucination detectionHalLoc VQA
Object Precision61
7
Hallucination LocalizationHalLoc (out-of-domain)
Object Score82
2
Showing 5 of 5 rows