NeurIPS Mech Interp Workshop is getting more AI slop

An analysis of submissions to the NeurIPS Mechanistic Interpretability Workshop reveals a significant increase in AI-generated content, or 'slop', between 2024 and 2025, according to researcher @andyarditi. The study leverages three years of workshop data to quantify the rise of LLM-generated submissions and assess the effectiveness of the review process in filtering them.

An unexpected benefit of having run 3 mech interp workshops: we have a great dataset for analysing the rise of LLM slop in submissions @andyarditi investigated how much AI slop we let in, how things have changed since 2024, and more Our review process isn't entirely noise