Dataset for Deep Learning-Based Attribution of OpenAI GPT Models
收藏Zenodo2025-09-09 更新2026-05-26 收录
下载链接：
https://zenodo.org/doi/10.5281/zenodo.17086635
下载链接
链接失效反馈官方服务：
资源简介：
This dataset accompanies the paper "Deep Learning-Based Attribution of OpenAI GPT Models” and contains 100,000 text responses generated across four OpenAI large language models: GPT-3.5-Turbo, GPT-4-Turbo, GPT-4o, and GPT-4o-mini.
Contents
25,000 unique prompts spanning 50 subject domains:
Art & Culture, History, Nature, Science, and Technology
100,000 responses:
25k per model
Each record includes structured fields:
Category, Subcategory, Prompt variation, Prompt text, Prompt token count, Model name, Response text, Response token count, and Generation time
Intended UseThe dataset was created to support research in authorship attribution, synthetic text analysis, and classifier benchmarking, and may be reused in broader studies of large language model behavior.
提供机构：
Zenodo创建时间：
2025-09-09
