HoloCount introduces a hierarchical taxonomy to evaluate Multimodal Large Language Models on semantic and analytical counting. The benchmark targets numerical hallucinations by testing spatial reasoning and logical composition in complex scenes.
HOW THIS AFFECTS YOU
●
researcherYou can use this to diagnose specific failure modes in MLLM quantitative precision.