Who Gets to Submit? Regulating Submission Volume Through Token Mechanisms

Scientific venues want to attract as many high-quality submissions as possible---and as few low-quality ones. Reviewing capacity is limited, so venues must ration submission opportunities; flat per-author caps do so bluntly, ignoring authors' differing capacities, while adapting future submission rights to past decisions runs into a basic obstacle: peer review is noisy, so outcome-based penalties risk punishing reviewer error rather than low quality---hurting exactly the authors a venue most wants to attract. Generative AI, by collapsing the cost of producing plausible manuscripts, is making this rationing problem increasingly salient. To address it, we introduce and analyze \emph{token-based submission mechanisms}: each author holds a renewable budget of submission tokens that function as credits---each submission consumes one token, and the venue's update rule replenishes the balance from the current balance and the observed accept/reject outcomes. In a stylized model with one venue, solo-authored manuscripts, two quality levels, and noisy binary decisions, we show that a natural rule---granting an extra token only when every submitted manuscript is accepted---works under perfect review but provably fails under noise. We then propose a noise-aware \emph{credibility} rule that does not penalize rejection counts consistent with the review error rate. We prove that it never excludes an author, leaves inactivity unpunished, costs only a small expected amount during reduced-output periods, and lets budgets grow as long as an author has additional high-quality work to submit---while driving down the budgets of authors whose submissions contain a large share of low-quality manuscripts. More broadly, this work defines submission-right allocation under noisy evaluation as a new mechanism-design problem, and establishes a baseline and desiderata for richer models of submission systems.