3rd New England Mechanistic Interpretability (NEMI) Workshop

The New England Mechanistic Interpretability (NEMI) Workshop holds its third edition on 14 August 2026 at Boston University's George Sherman Union on Commonwealth Avenue. It is a single-day gathering of academic and industry researchers working on mechanistic interpretability — the effort to understand the internal computations of machine-learning models rather than treating them as black boxes.

The programme covers interpretability of neural circuits, probe-based analysis, feature attribution, model simplification and "interpretability for science", combining invited keynotes with oral presentations and a poster session. Confirmed speakers include Tamar Rott Shaham (MIT), Lee Sharkey (Goodfire), Ekdeep Singh Lubana and Andrew Lee (Harvard), Michael Hanna (University of Amsterdam) and Aaron Mueller (Boston University), with sponsorship from Goodfire and the NSF's NDIF. Registration and submissions were open at the time of writing; as a focused regional workshop it is a good fit for researchers and graduate students who want a concentrated day inside one of the fastest-moving corners of AI safety and interpretability.