ICLR abstracts using the word "Novel" plummet from 37% to 5%, possibly AI-writing related
hugo_larochelle · x · 2026-11-07
Lacuna Research found the share of ICLR abstracts containing the word "Novel" dropped from 37% in 2025 to 5% this year. They suspect AI writing, but Pangram AI-detection scores don't correlate—possibly a shift in newer model styles, or the ML community collectively recoiling from AI-sounding prose. Hugo Larochelle amplified the finding.
More from Research
- ICLR Reviewer Asks Authors to Explain 'Standard Deviation', sparking peer-review mockery — deliprao · 2026-11-08
- AI-quantified coronary inflammation explains why two major cardio trials failed, Topol says — EricTopol · 2026-11-08
- Researcher flags scaling fit that hard-codes capability jump to zero — JacquesThibs · 2026-11-08
- PIVOT, a robot policy improvement method from deployment experience, accepted to CoRL 2026 — ZeYanjie · 2026-11-08
- Two Minute Papers Explains the OpenAI Astra Weight-Space Programming Hack — Two Minute Papers · 2026-11-08
- SAIGE-QTL, a mixed-model tool for single-cell eQTL mapping, published in Nature Genetics — anshulkundaje · 2026-11-08