arXiv:2608.14571 Proposes OpenReview Points Credit System to Replace Voluntary ML Review Guidelines

The paper identifies two structural problems: uncontrolled submission growth under reciprocal review policies and the absence of mechanisms to reward high-quality reviews. It evaluates four existing conference tools, finds them insufficient, and advances two designs centered on a currency-like points ledger that reviewers earn through verifiable actions and spend on registration waivers or extra review resources.

Submission volumes at major venues have risen sharply while publication fees remain zero, removing any market signal that could curb low-effort submissions. The authors note that almost all participants report negative review experiences yet lack formal channels to debate system effectiveness, a gap confirmed by OpenReview platform logs and conference post-mortems from 2023-2025.

Existing attempts at reviewer guidelines and reciprocal-review mandates have produced no measurable lift in review depth or consistency, according to internal conference audits. A points system creates fine-grained, auditable incentives that can be spent across multiple conferences, directly addressing free-rider dynamics that voluntary norms have failed to correct.

Operational deployment would require OpenReview to expose point ledgers and conference organizers to accept points for defined perks, with initial pilots likely limited to one or two venues before 2028.