Four anchors, twelve months
Ask for machine learning conferences and the honest answer is a calendar. In ConfJournal's directory, the four A* anchors of the ML year carry these tentative deadline months: ICML in January, NeurIPS in May, ICLR in September, CVPR in November. Public deadline trackers' indicative dates for the 2027 cycle land in the same months (mid-January for ICML, late May for NeurIPS, mid-November for CVPR), which is what tentative means: the month is stable, the day firms up when the call prints. Around those anchors sit the rest of the directory's 141 conferences, and most ML labs plan backwards from two of the four.
Why the anchors pace everything else
A rejected paper at one anchor becomes a submission at the next; the four deadlines form a relay. January's ICML decision feeds May's NeurIPS deadline, whose outcome shapes September's ICLR run, and November's CVPR closes the year. Workshops, journals and the field's second tier schedule themselves in the gaps. This is also why a single missed week in January costs more than a missed week in March: the relay has no spare legs.
Tentative months, honestly labeled
Our directory records a tentative deadline month for each venue because conferences move their days, occasionally their months, and a tracker that prints a false precise day is worse than an honest month. When a call for papers prints, the day replaces the estimate; this blog's CVPR 2027 post did exactly that when the November dates fixed. Treat every month here as a planning horizon, and confirm the day on the official call before you book the all-nighter.
Put the anchors on your own calendar
The venue directory lists all 141 conferences with their tiers and months, and the Odds Checker helps decide which anchors deserve the paper. Pick two anchors a year, protect the month before each, and the calendar stops being a surprise.