Ai models vary widely in spotting errors in scientific paper reviews

Auditing AI peer reviewers: a benchmark on real scientific papers

Computation and LanguageArtificial IntelligenceDigital Libraries

Summary

Checking scientific papers carefully for mistakes is a big job that some AI tools now help with. The authors created a test with real scientific papers that had made-up errors added, to see how well different AI reviewers could find those mistakes. They found that the choice of AI model makes a big difference—some models find many errors while others miss half. They also made a special AI reviewer called Skepthical that checks claims by looking up citations and verifying math and numbers, which worked especially well for spotting citation problems in physics papers. Most AI reviewers rarely say there’s a mistake when there isn’t one, but math errors are the trickiest for them.

What this means in practice

  • For journal editorial teams: Identify which AI peer reviewing tools best detect specific error types in submitted scientific papers to assist human reviewers.
  • For academic conference organizers: Integrate AI reviewers with citation verification capabilities to improve review quality especially in physics-related submissions.

Authors

Íñigo Zubeldia, Boris Bolliet, Francisco Villaescusa-Navarro, Pablo Villanueva-Domingo, Miles Cranmer

Abstract

LLM-based peer review is now used at production scale, but its evaluation remains limited. We introduce a benchmark that injects objective, localisable errors of four types (general issues, mathematics, numerics, citations) at three severity levels into real papers and measures each reviewer's detection rate by error type and severity, together with its false-positive rate on clean copies. We also present Skepthical, a multi-agent reviewer that verifies citation-backed claims by retrieving the cited papers, mathematics with computer algebra, and numerical claims with executed code. We apply the benchmark to two corpora of 15 recent arXiv preprints each, one in astrophysics and one in machine learning (ML), injecting a total of 357 errors, and evaluate twelve reviewer systems: six single-LLM reviewers, two Claude Code configurations, three publicly available reviewers, and Skepthical. The choice of model has the largest impact on detection rate: given the same prompt, three frontier LLMs differ by about 50 percentage points in total detection on both corpora. Tools and scaffolding help when they contribute what the model cannot supply on its own. In particular, on astrophysics, Claude Code improves on a single Claude call in mathematics and numerics detection, and enforced citation retrieval gives Skepthical the highest citation error detection rate of all systems by a large margin. This citation lead, however, is not observed on ML, where the models appear to know the literature better and Skepthical's citation detection rate is statistically indistinguishable from that of the best single LLM. False-positive rates are below 10% for most systems, but vary with prompt, model, and scaffolding, with mathematics flags the main source of false positives. The evaluation pipeline, the injected errors, and the adjudication logs are released as a benchmark.