2026-10-14T14:09:11Z https://www.scidb.cn/oai

10.57760/sciencedb.33718 2026-08-10T11:21:15Z

2026-08-10 A Multi-level Gradient Rewriting Dataset of Chinese Academic Paper Abstracts for AIGC Detection MGRD doi:10.57760/sciencedb.33718 en  To support AIGC detection, text source identification, and academic integrity research in Chinese academic writing, we constructed the MGRD (A Multi-level Gradient Rewriting Dataset of Chinese Academic Paper Abstracts for AIGC Detection) dataset. MGRD sources human-written abstracts from Chinese core journal papers in CNKI and Wanfang Data (2011–2022), covering computer technology, architectural theory, and Chinese drama. Based on these original abstracts, five LLMs (glm-4.5-air, glm-4.6v, qwen3-14b, deepseek-R1, and gpt-4o-mini) were used to generate AIGC samples at three levels: light polishing, medium rewriting, and heavy rewriting. Notably, the heavy samples were independently generated using only paper titles and keywords. After rule filtering, hierarchical constraint, anomaly removal, semantic consistency verification, perplexity analysis, and expert blind review, we finalized four core data files (light_paired.csv, medium_paired.csv, heavy_paired.csv, all_paired.csv) and one comprehensive file (complete_paired.csv). The core files consist of 5,822 light, 5,748 medium, and 6,117 heavy pairs, totaling 17,687 pairs (35,374 texts). The complete_paired.csv file is used to extract complete three-level gradient subsets and is excluded from the total sample count to avoid duplication. Retaining core fields (text, label) and auxiliary fields (title, keywords, rewriting intensity, and rewriting rate), the dataset supports mixed-scenario model training, cross-dataset generalization evaluation, and text source analysis. Evaluations confirm that MGRD serves as a foundational data resource for AIGC detection research in Chinese academic abstracts. AIGC Detection; Abstracts Of Chinese Academic Papers; Multi-Level Gradient Rewriting; Text Source Detection; Datasets NAN Zhe LI Shuo WANG Yanfeng PUBLIC https://creativecommons.org/licenses/by-nc/4.0/ dataset Science Data Bank