List your company

Hallucination detection on BEGIN (test)

47.3Macro F1 ScoreHaluGen
31.735.7539.843.85Jul 7, 2024

Evaluation results

MethodLinks
HaluGenEvaluation protocol=Finetuned, Backbone=T5-base, Training data=Synthetic Data2024.0747.3—
BEGIN-Adv.Evaluation protocol=Finetuned, Backbone=T5-base, Training data=Synthetic Data (Dziri et al., 2022a)2024.0745.9—
AugWowEvaluation protocol=Finetuned, Backbone=T5-base, Training data=Synthetic Data (Gupta et al., 2021)2024.0737.8—
GPT-4 (Internal)Evaluation protocol=Zero-shot Detection2024.0732.3—