A proposal to let ICML, ICLR and NeurIPS area chairs desk-reject roughly half their assigned papers

Maksym Andriushchenko proposes allowing area chairs at ICML, ICLR and NeurIPS to desk-reject roughly half their batches before review. He says a quick scan can identify papers with almost no chance of acceptance. Florian Tramèr says SaTML is testing a similar approach, though probably below 50%, after submissions doubled from last year.