ICML final decisions rant [D]

About this article

So, ICML accepted ~6.5K of ~24K; obviously, it doesn't mean that all the rejected papers are "bad," and these rejected papers would cascade to NeurIPS, blowing up NeurIPS' total submission count, and this cycle of massive-influx-small-acceptance would repeat on an endless loop. The reviews themselves can be frustratingly inadequate: - "Only 200 benchmarks included; not included didn't-do-this-benchmark" (exaggerated for dramatic effect, sadly not unrealistic) or - "I don't think this paper, t...