CoCoReviewBench - Correctness: leaderboard
CoCoReviewBench (2026) evaluates AI peer reviewers on ICLR and NeurIPS papers against category-specific human review opinions that were filtered for correctness using reviewer, author and meta-review discussions. This board measures how well each review comment agrees with the most reliable human opinion on a 1 to 5 judge scale, shown as the difference from the leave-one-out human reviewers, so positive values beat the humans; higher is better.
Metric: Category-level correctness score (agreement of each review comment with the single most reliable human opinion per cluster) on the 1-5 judge scale, printed as the difference from the human reference reviews (3.55, leave-one-out: one human review scored against the others), so a positive value is better than the human reviewers and the range is -2.55 to +1.45; AI reviews of 1,300 ICLR and NeurIPS papers sampled from CoCoReviewBench (100 per year, original score distribution kept), each split into atomic opinions and compared category by category with the filtered human opinions by a GPT-5-Mini judge at temperature 0, general models prompted with the AI Scientist review prompt at temperature 0.6; higher is better. Source: arxiv.org. Saturation forecast: Around October 2027. 17 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | GPT-5.2 | 0.36 | #185 | |
| 2 | GPT-5 Mini | 0.29 | #364 | |
| 3 | Gemini 3 Pro | 0.14 | #145 | |
| 4 | Gemini 3 Flash | 0.09 | #175 | |
| 5 | Qwen 3 32B | Thinking | 0.01 | #972 |
| 6 | QwQ-32B | -0.01 | #907 | |
| 7 | nemotron-3-nano-30B-a3B | Reasoning | -0.01 | #1946 |
| 8 | OpenReviewer-8B | -0.08 | ||
| 9 | Qwen 2.5 7B | -0.13 | #2929 | |
| 10 | CycleReviewer-70B | -0.15 | #1880 | |
| 11 | CycleReviewer-8B | -0.16 | ||
| 12 | DeepReviewer-14B | -0.17 | #844 | |
| 13 | DeepReviewer-7B | -0.22 | ||
| 14 | Llama 3.3 70B | -0.25 | #1514 | |
| 15 | Qwen 3 8B | Non-reasoning | -0.28 | #1873 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=cocoreviewbench-correctness · How It Works · Data refreshed daily, snapshot 2026-11-21.