happy8825/MMLongBench_var7_booster_seed20
收藏Hugging Face2025-12-14 更新2025-12-20 收录
下载链接：
https://hf-mirror.com/datasets/happy8825/MMLongBench_var7_booster_seed20
下载链接
链接失效反馈官方服务：
资源简介：
MMLongBench数据集是一个包含多模态长文档问答任务的数据集，其结构复杂，包含多个特征字段，如相关页面、证据页面、分数、文档ID、文档类型、问题、答案、证据来源等。数据集还包含多轮对话的详细记录，如LLM查询输入、检索结果、原始输出、记忆输入输出、VLM提示输入、原始输出等。数据集分为训练集，包含3个样本，总大小为71,795字节。
The MMLongBench dataset is designed for multimodal long-document question-answering tasks, featuring a complex structure with multiple fields such as relevant_pages, evidence_pages, score, doc_id, doc_type, question, answer, evidence_sources, and more. It also includes detailed records of multi-turn dialogues, such as LLM query inputs, retrieval results, raw outputs, memory inputs/outputs, VLM prompt inputs, and raw outputs. The dataset is split into a training set containing 3 samples, with a total size of 71,795 bytes.
提供机构：
happy8825
