A new research paper published on arXiv details significant BibTeX citation errors generated by large language models, even when equipped with web search capabilities. The study found that models like GPT-5, Claude Sonnet-4.6, and Gemini-3 Flash struggle with accuracy, particularly for recent or less-cited papers, often substituting entire entries or making isolated field errors. To address this, the researchers developed 'clibib,' an open-source tool that, when integrated into a two-stage process, significantly improves citation accuracy and reduces regression rates compared to single-stage tool loops. AI
IMPACT Highlights the need for improved citation accuracy in AI-powered scientific workflows and introduces a potential solution.
RANK_REASON The cluster contains a research paper detailing an evaluation and mitigation of errors in LLM-generated BibTeX citations. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
