VLMEval/OpenVLMRecords

OpenVLM Records数据集包含了由VLMEvalKit生成的评估记录，这些记录反映了OpenVLM排行榜上的表现。数据集主要用于评估大型多模态模型（VLM）在多模态基准测试（BENCH）上的表现，记录文件包含了模型对每个问题的原始响应。数据集的语言为英文和中文，任务类别为视觉问答，规模在1M到10M之间。

This is a dataset for visual question answering tasks, containing evaluation records of multiple large multi-modal models on different benchmarks. The dataset has a clear file naming system, with record files stored in a specific directory structure, allowing users to browse and analyze these records using the provided tools. The dataset size is between 1M and 10M, supporting both Chinese and English.