NeurIPS Rejects AI-Generated Papers, Citing Integrity Concerns
TL;DR. NeurIPS 2026 has desk-rejected 18.4% of position paper submissions and flagged 12.7% for review based on AI content detection. - The conference established a strict policy for human-written papers, with AI limited to copy-editing functions. - Integration of AI-generated text introduces verifiability costs and attribution questions for peer review. - Researchers partnered with Pangram, an AI detection company, to identify non-compliant submissions. - The decision highlights growing concerns over AI's impact on academic integrity and research ethics.
- NeurIPS 2026 Position Paper Track desk-rejected 18.4% of submissions for substantial AI generation.
- Another 12.7% of papers require evidence of human engagement to avoid rejection.
- The policy mandates papers be 'substantially human-written,' with AI only for peripheral edits.
- AI-generated text poses risks to peer review, externalizing verification costs and raising attribution issues.
- NeurIPS collaborated with AI detection firm Pangram to enforce its human-writing policy.
Sources
- AI-Generated Papers in the NeurIPS 2026 Position Paper Track — blog.neurips.cc