Describes the series’ typical recent cycle; last reviewed 2026-08-07. The official call is the authority on this year’s dates.
About ICML
ICML is machine learning’s summer flagship, with a reputation tilted slightly toward methodological and theoretical depth relative to its siblings. Its scope spans learning theory, optimization, deep learning, probabilistic methods, and applications.
The cycle runs roughly half a year ahead of the July meeting: full papers due around the end of January, reviews and rebuttals through spring, decisions around May. Together with NeurIPS (May deadline) and ICLR (September deadline), it forms the field’s three-beat annual rhythm — miss one and the next is months, not a year, away.
Typically late January to early February for full papers, with the conference in July. Exact dates per cycle are on icml.cc.
ICML or NeurIPS — which should I target?
Prestige is comparable; the practical difference is timing (January vs May deadlines) and modest cultural tilt (ICML somewhat more methods/theory-flavored). Most ML groups target whichever deadline their result is ready for.