Найти похожее

📊 В топ-50 контрибьюторов NeurIPS 2025 Китай и США идут практически наравне. В США произошёл сдвиг: корпоративные AI-лаборатории - Google DeepMind, Meta, Microsoft, теперь сопоставимы по вкладу с ведущими университетами вроде Stanford, CMU и MIT. Исследования всё заметнее смещаются в индустрию. В Китае лидерами остаются академические центры: Tsinghua, CAS, Peking University, SJTU, HKUST. Сильные позиции держат также Сингапур (NUS, NTU), Корея (KAIST), ОАЭ (MBZUAI) и Канада (Mila). Европа заметно отстаёт: в списке лишь Oxford, EPFL, ETH Zurich и TUM из стран ЕС-27. 👉Источник @ai_machinelearning_big_data #NeurIPS

📖NeurIPS Papers Flagged for AI Hallucinations and Fake Citations GPTZero reports finding 100 AI hallucinations across 51 papers accepted to NeurIPS Conference, including invented authors, fabricated sources, and AI-generated text presented as legitimate scholarship. The discovery follows GPTZero’s earlier identification of 50 hallucinated citations in papers under review at ICLR, suggesting that unreliable generative AI assistance is entering peer-reviewed research pipelines at scale. GPTZero attributes the problem to rapid adoption of generative AI tools amid rising publication pressure. NeurIPS submissions increased more than 220% from 9,467 in 2020 to 21,575 in 2025, expanding reviewer load and increasing risks of oversight failures, expertise mismatches, negligence, and fraud. The same period also saw growth in substantive errors beyond citations, including incorrect formulas, miscalculations, and flawed figures. GPTZero notes that correlation is not causation, but reports a 55.3% increase in NeurIPS paper error rates following the release of ChatGPT. The implication is practical: unchecked AI use can undermine credibility and may invalidate academic work. #AIandLaw#AIHallucinations#NeurIPS#ResponsibleAI