The Decoder says the 2027 ICLR round already holds roughly 50,000 abstracts while the cutoff remains a week out, versus about 19,500 valid papers last cycle. Some authors will likely pull work after NeurIPS decisions, so the final tally should drop yet remain well above 2026. Likely drivers include publication-linked pay and faster machine drafting; a NeurIPS study found heavy assistant use in writeups. The 2026 meeting already saw trust slip after fabricated-citation papers.
The Decoder reported on 19 September that ICLR 2027 has already pulled in about 50,000 abstracts, with the cutoff still the following week. ICLR 2026 saw around 19,500 valid submissions. The outlet says some abstracts likely come from authors hedging, waiting on NeurIPS results and planning to withdraw if accepted there, so the final count will be lower but still far above last year.[1]
The Decoder lists broader AI hype and corporate research spending, where pay is sometimes tied to publication records, among likely drivers of the flood. It says the biggest factor is probably that AI makes it much faster to crank out papers. It cites a NeurIPS analysis that found authors used AI heavily to write their submissions. Those are the outlet's attributions; this card does not treat the NeurIPS analysis as a second news publisher of the 50,000-abstract count.[1]
The Decoder says ICLR 2026 already struggled with low-quality machine-written submissions and reviews that eroded trust in peer review. Authors submitted generated papers packed with fabricated citations, and reviewers turned to assistants to keep up with the volume. With even more papers flooding in this year, it expects those complaints to get louder. The trail the page names is a forum thread; the conference's own tally table is not printed there.[1]
