ICML
One of the three biggest academic conferences in machine learning, where much of the field's research is first published.
ICML, the International Conference on Machine Learning, is one of the venues where new AI research is formally published. Along with NeurIPS and ICLR, it is where a paper has to land for the field to take it seriously. Getting accepted means a panel of anonymous reviewers read your submission and decided it was sound, which is why “presented at ICML” carries weight in press releases.
The conference has also become a good illustration of a scale problem. ICML 2026 accepted 6,341 papers, roughly double the previous year, which is far more than any volunteer review process can check carefully. That is the backdrop to Hugging Face’s Open Reproductions project in August 2026, where 1,221 volunteers used coding agents to re-run experiments from more than 2,200 of those papers and found that about 23 percent had at least one claim contested or falsified.