[r/ML] ICML final decisions rant [D]

ICML recently announced its final decisions, accepting roughly 6,500 papers from a total of approximately 24,000 submissions. This acceptance rate, around 27%, means a large number of papers were rejected.

Rejected papers from ICML are anticipated to be resubmitted to other major conferences, such as NeurIPS, which is expected to further inflate their submission numbers. This creates a continuous cycle of high submission volumes and relatively low acceptance rates across top-tier machine learning conferences.

Community discussions also brought attention to issues with the quality of peer reviews. Examples cited included reviews that were perceived as inadequate or overly critical for not including specific benchmarks, even if those benchmarks were not central to the paper's core contribution.

Researchers and the academic community will be observing the submission numbers for upcoming conferences like NeurIPS to see the full impact of this cascade effect. The ongoing discussion around improving the quality and fairness of the peer-review process in high-volume academic publishing remains a critical area for the machine learning community.