PaperAudit-Bench (Fast Mode) - GPT-5 Errors: leaderboard

Metric: Macro-F1 (%; the per-paper harmonic mean of error coverage and finding precision, averaged over the 46 NeurIPS 2025 oral papers of the benchmark's NeurIPS branch as corrupted by gpt-5-2025-08-07 with 10-20 injected errors of eight types; Fast mode, one global review pass over the whole paper; findings matched to the injected errors by a GPT-5.1 judge; the three text-only detectors read the papers with figures removed). Source: arxiv.org. Saturation forecast: Around December 2026. 10 models tracked.

Top models

#ModelScore
1GPT-562.5
2Gemini 2.5 Pro60.5
3Grok 448.1
4Qwen 3 235B A22B 2507 Instruct45.7
5O4 Mini45.3
6DeepSeek V3.1 (Non-reasoning)43.6
7Claude Sonnet 4.543.1
8GLM-4.6 (Non-reasoning)35.5
9Kimi K2 090521.6

Interactive version: theaggregate.ai/benchmark?slug=paperaudit-bench-fast-mode-gpt-5-errors · How It Works · Data refreshed daily, snapshot 2026-09-29.