Classifier run on 42,000+ ICLR 2027 submissions finds over half nearly fully AI-assisted
nsaphra · x · 2026-10-15
A developer built a classifier to detect LLM-assisted writing in ML papers and ran it on all 42,000+ ICLR 2027 submissions, publishing sentence-level results in an integrated PDF viewer. The data shows 23,459 papers (55.3% of scored results) have 95–100% of sentences flagged as AI-assisted, meaning over half the submissions appear nearly fully AI-written. The detector's false positive rate on held-out human sentences is 0.1%.
More from Research
- Stanford's CausalSmith automates causal inference research with Lean-verified proofs — StanfordAILab · 2026-10-15
- Cellular Intelligence wins Best Poster at scverse 2026 for perturbation platform — arjunrajlab · 2026-10-15
- BARRACUDA microfluidic chip traps a million single cells to map gene circuit design space at scale — NikoMcCarty · 2026-10-15
- Stanford's DexAgent turns one human egocentric video into dexterous robot training data — StanfordAILab · 2026-10-15
- Johns Hopkins wins HCOMP 2026 Best Paper for reader-centered summarization evaluation — mdredze · 2026-10-15
- Stanford's new CS 312 'Deep Learning Alchemy' trains intuition via prediction quizzes — stanfordnlp · 2026-10-15