MMRefine

该数据集是一个旨在评估多模态大型语言模型（MLLMs）在六种不同错误场景下的错误修正能力的基准测试。它将错误分为六类，并提供了关于MLLMs修正性能的深入见解。该基准测试涵盖了多种MLLMs，包括闭源和开源模型，其任务重点是MLLMs中的错误检测与纠正。

This dataset is a benchmark designed to evaluate the error correction capabilities of multimodal large language models (MLLMs) across six distinct error scenarios. It categorizes errors into six classes and provides in-depth insights into the correction performance of MLLMs. This benchmark covers a variety of MLLMs, including both closed-source and open-source models, with its task focusing on error detection and correction in MLLMs.