E-commerce
收藏资源简介:
这是一个商业领域的电子商务数据集,发布用于生成引擎优化(GEO)研究。数据集包含多个配置:主数据集用于GEO训练和评估(约1.6k训练样本/400测试样本),规则候选数据集用于内容偏好规则提取(约1.6k样本),冷启动数据集用于AutoGEO Mini的监督微调(约500样本),推理数据集用于推理(约400样本),GRPO输入数据集用于GRPO训练(约1.6k样本),GRPO评估数据集用于评估GRPO训练模型(约1.6k样本)。数据集与AutoGEO项目相关,支持文本生成和生成式搜索等任务。
This is an e-commerce dataset in the business domain, released for Generative Engine Optimization (GEO) research. The dataset includes multiple subsets: the main subset is used for GEO training and evaluation, with approximately 1.6k training samples and 400 test samples; the rule candidate subset is for content preference rule extraction, containing about 1.6k samples; the cold-start subset is for supervised fine-tuning of AutoGEO Mini, with roughly 500 samples; the inference subset is for inference tasks, with around 400 samples; the GRPO input subset is for GRPO training, containing approximately 1.6k samples; and the GRPO evaluation subset is for evaluating GRPO-trained models, with about 1.6k samples. This dataset is associated with the AutoGEO project and supports tasks such as text generation and generative search.
E-commerce Dataset (AutoGEO) 概述
数据集简介
这是一个用于生成引擎优化(GEO)研究的商业领域数据集,随AutoGEO项目发布。
核心信息
- 许可证: mit
- 任务类别: 文本生成
- 标签: 生成式搜索、geo、autogeo、电子商务、llm、强化学习
数据集配置
数据集包含六个不同的配置,每个配置对应特定的用途和样本量:
- main: 用于GEO训练和评估的主要训练/测试数据(约1.6k训练样本 / 约400测试样本)
- rule_candidate: 用于内容偏好规则提取的数据(约1.6k样本)
- cold_start: 用于AutoGEO Mini监督微调的数据(约500样本)
- inference: 仅用于推理的数据(约400样本)
- grpo_input: 用于GRPO训练的输入数据(约1.6k样本)
- grpo_eval: 用于评估GRPO训练模型的数据(约1.6k样本)
相关资源
- 论文: "What Generative Search Engines Like and How to Optimize Web Content Cooperatively"
- 作者: Yujiang Wu*, Shanshan Zhong*, Yubin Kim, Chenyan Xiong (*Equal contribution)
- 代码: AutoGEO on GitHub
引用格式
bibtex @article{wu2025generative, title={What Generative Search Engines Like and How to Optimize Web Content Cooperatively}, author={Wu, Yujiang and Zhong, Shanshan and Kim, Yubin and Xiong, Chenyan}, journal={arXiv preprint arXiv:2510.11438}, year={2025} }




