NeurIPS 2026 AI-Assisted Reviewing Experiment

Summary

NeurIPS 2026 runs an actual randomised experiment inside its own review process: reviewer-paper pairs are assigned by constrained optimisation to unassisted review, open-ended LLM assistance inside OpenReview, or structured task-specific assistance, with blinded area chairs rating review quality and all LLM conversations logged and visible to the experiment chairs. IRB-exempt under UT Austin protocol STUDY00009083. The clearest example of a venue treating AI-in-review as an empirical question rather than a policy edict, and a template for studying AI in the IS review process.

Contribution

Queued — not yet read.

Method

Queued — not yet read.

Relevance to RISE

Surfaced by the 2026-08-29 literature sweep as a high-priority candidate for the RISE knowledge base (angle: AI in peer review). Verified against: NeurIPS 2026 conference page opened 2026-08-29 (design, three conditions, randomisation, permissions, oversight, co-chairs and IRB protocol read)

Critique / open questions

Queued — not yet assessed.

Key quotes

—