2026-10-10T13:54:46Z https://www.scidb.cn/oai

10.57760/sciencedb.j00133.00711 2026-09-20T09:58:06Z

2026-09-20 ICLR 2024 and 2025 Human–AI Parallel Review Dataset doi:10.57760/sciencedb.j00133.00711 en The ICLR 2024 and 2025 Human-AI Parallel Review Dataset is a peer review empirical analysis dataset constructed based on publicly available review data from the OpenReview platform. This dataset covers the ICLR 2024 and ICLR 2025 conferences, with a total of 1186 paper samples, including 586 accepted papers and 600 rejected papers. The dataset integrates 4655 human review comments and generates 18677 AI review comments based on four mainstream language models: Claude Sonnet 4.5, GPT 5.2 Thinking, Gemini 3 Pro review, and DeepSeek V3.2, forming a parallel review corpus for the same paper between human reviewers and multi model AI reviewers. The dataset contains information on paper acceptance results, acceptance levels, review scores, review confidence, human review texts, and AI review texts.  ICLR; Peer Review; Human-AI Collaboration; LLM zhou xiao yu PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank