Polarizing ICML 2026 reviews highlight 'single-reviewer veto' problem

The "single-reviewer veto" is a structural failure that undermines the collaborative intent of peer review and fuels author burnout.

• –Area Chairs (ACs) frequently default to the most negative reviewer to "gatekeep" the conference, regardless of the reviewer's specific technical critiques or alignment with the other reviews.
• –The 2026 ICML cycle is uniquely chaotic due to experimental LLM policies and watermark "traps," which have increased overall reviewer paranoia and friction during the rebuttal phase.
• –Rebuttals often fail when reviewers "double down" on negative scores to justify their initial stance, even when technical errors are pointed out, as seen in this case's 1/5 originality score.
• –The lack of AC accountability for dismissing clearly outlier reviews remains the primary source of frustration for the academic ML community.