Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
Hal-Evalgithub.com · Low evidenceHal Eval is a universal and fine-grained hallucination evaluation framework for Large Vision Language Models (LVLMs). It introduces a refined taxonomy including Event Hallucination and provides both discriminative and generative evaluation methods to comprehensively assess LVLMs' handling of hallucinations.
Parse Score