ICML 2027 Program Chair Publicly Seeks Ideas to Fix AI Slop and Review Crisis

On September 21, Andrew Gordon Wilson, Program Chair of ICML 2027, posted on X publicly soliciting creative ideas for running the conference, directly confronting systemic problems facing top ML venues. The discussion remains at the idea-collection stage with no proposal officially adopted, but it exposes deep structural stress in the conference ecosystem.

Confirmed

• Wilson's solicitation covered: managing the flood of AI-slop submissions and explosive paper volume, improving peer review, incentivizing high-quality original and creative work, and reducing reviewer burden and bureaucratic overhead.
• RL researcher MarlosCMachado suggested staggering ICLR notification dates.
• Researcher tallinzen offered a concrete proposal (post truncated).
• Mirco Musolesi proposed five reforms, one explicit one being capping per-author submissions at NeurIPS/ICML/ICLR (others incomplete in the material); roydanroy relayed replies jokingly echoing a per-author cap.
• Zeeshan Zia proposed using SOTA LLMs to pre-screen and desk-reject roughly 90% of submissions, noting his team already uses a "prospectus desk" mechanism (details truncated).
• In a repost thread, @galyo proposed a radical track with AI fully in charge of reviewing (details truncated).
• Per a repost, furongh announced co-chairing ICML 2027 with Weijie Su, Andrew Wilson, Julien Mairal, and Courtney Paquette; discussion noted ICLR abstract submissions have exceeded 60,000.
• Other replies suggested borrowing practices from ACL/NAACL/EMNLP/CVPR.

Unconfirmed

• One repost attributed the solicitation to Andrew Ng, but multiple first-hand accounts identify Andrew Gordon Wilson as the poster—likely a misattribution; another post misspelled the name as "Andrew Glove."
• Details of Musolesi's five proposals, tallinzen's and @galyo's plans are truncated and cannot be fully verified.

Why it matters

• Submission volumes and AI-generated junk are surging in tandem—ICLR abstracts topping 60,000 is evidence—and reviewer load is near its limit; the organizers' open call signals urgency.
• Radical ideas like LLM pre-screening with 90% desk rejection and fully AI-reviewed tracks reflect some researchers' appetite for automated quality control, potentially shaping future review processes.
• Proposals such as staggered notification dates and per-author caps point to ecosystem-wide scheduling and capacity constraints rather than a single conference's internal reform.

2026-09-21 ~ 2026-09-22 · 14 related posts

• [source] ICML 2027 Program Chair Asks Publicly: How to Handle AI Slop and Submission Explosion — andrewgwils · 2026-09-21
• [source] ICML 2027 chair crowdsources fixes for AI slop; researcher proposes LLM desk-reject of 90% papers — ZeeshanZiaML · 2026-09-21
• [source] ICML 2027 chair crowdsources reform ideas; researcher proposes deadlines ahead of ICLR notices — MarlosCMachado · 2026-09-22
• ICML 2027 program chair on AI slop: writing help shouldn't be punished as ICLR hits 60K+ abstracts — thegautamkamath · 2026-09-22
• ICML 2027 chair crowdsources fixes for submission explosion; proposal for AI-only review track — thegautamkamath · 2026-09-22
• ICML 2027 chair asks publicly how to handle AI slop and submission explosion — roydanroy · 2026-09-22
• ICML 2027 chair asks for ideas; scholar proposes 5 fixes for submission explosion — mircomusolesi · 2026-09-22
• Andrew Ng, ICML 2027 Program Chair, asks for ideas to fight AI slop and submission explosion — xuanalogue · 2026-09-22

6 near-duplicate retellings: tallinzen · ZeeshanZiaML · wgrathwohl · mircomusolesi · mircomusolesi · xwang_lk