Skepthical Study - Astrophysics Numerics: leaderboard

Skepthical tests models and review pipelines on deliberately flawed research papers. Astrophysics and machine-learning results use separate 15-paper cohorts and expert error annotations. Free-form, structured and concise review prompts remain distinct. False-positive rates use all raised flags on the clean-paper controls. This board reports error detection rate (%). Higher is better.

Metric: Error detection rate (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 7 models tracked. Updated 14 Nov 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Opus 4.7Evaluation run: Skepthical Study Claude Code Structured91.1#78
2GPT-5.5Evaluation run: Skepthical Study Concise Structured88.9#52
3SkepthicalEvaluation run: multi-provider pipeline frozen 2026-04-2180
4Gemini 3.1 Pro PreviewEvaluation run: Skepthical Study Structured57.8
5Stanford Agentic ReviewerEvaluation run: hosted reviewer 2026-05-0153.3
6PaperAudit_Qwen3_14B_sft_rlEvaluation run: Skepthical Study Deep6.7
7Llama-OpenReviewer-8BEvaluation run: Skepthical Study ICLR Template0

Interactive version: theaggregate.ai/benchmark?slug=skepthical-study-astrophysics-numerics · How It Works · Data refreshed daily, snapshot 2026-11-16.