NeurIPS论文假开源，较真AI研究员开锤了

Core Viewpoint - The article highlights the issue of "fake open source" in the AI academic community, where papers claim to be open source but fail to deliver on that promise, leading to a significant number of empty repositories and unfulfilled commitments [3][19]. Group 1: Research Findings - A study of 4,035 papers accepted at NeurIPS 2024 revealed that only 2,404 were genuinely open source, while 1,533 did not provide GitHub links, and 98 papers explicitly stated they were open source but led to empty or non-existent repositories [5][14]. - The research was conducted using an AI system that integrated OpenReview/GitHub API and PDF parsing technology to verify the existence of code linked in the papers [12]. Group 2: Reasons Behind the Issue - The rise of "fake open source" is attributed to the peer review process, where indicating a willingness to open source becomes a de facto requirement for paper acceptance, leading to the prevalence of placeholders like "Coming Soon" [20][21]. - Various factors contribute to the inability to release code, including lengthy compliance processes in industry, high replication costs, and unforeseen circumstances such as team changes or patent issues [24]. Group 3: Community Response - The article notes a growing frustration among researchers and the community regarding the prevalence of empty repositories, with calls for greater accountability in academic commitments [25][28]. - The sentiment expressed by an anonymous researcher emphasizes that lack of time should not be an excuse for failing to fulfill open-source promises, advocating for integrity in academic work [28][30].