RyanWW/XModBench
收藏官方服务：
资源简介：
XModBench是一个用于评估全语言模型在跨模态能力和一致性方面的综合基准测试。它包括文本、视觉和音频模态的全面测试，涵盖感知、空间、时间、语言和知识五个任务维度，评估了13个最先进的模型，并提供了性能稳定性的一致性分析以及人类性能基线。
XModBench is a comprehensive benchmark designed to evaluate the cross-modal capabilities and consistency of omni-language models. It includes comprehensive testing across text, vision, and audio modalities, covering five task dimensions: perception, spatial, temporal, linguistic, and knowledge. It evaluates 13 state-of-the-art models and provides consistency analysis for performance stability, as well as human performance baselines.
提供机构：
RyanWW
