PaperAudit-Bench (Fast Mode) - GLM-4.6 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 glm-4.6 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 May 2028. 10 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro30.3
2GPT-529.1
3Claude Sonnet 4.527.2
4GLM-4.6 (Non-reasoning)25.3
5Qwen 3 235B A22B 2507 Instruct25.1
6Grok 423.2
7O4 Mini23.2
8DeepSeek V3.1 (Non-reasoning)19
9Kimi K2 090515.9

Interactive version: theaggregate.ai/benchmark?slug=paperaudit-bench-fast-mode-glm-4-6-errors · How It Works · Data refreshed daily, snapshot 2026-09-29.