RINoBench: leaderboard
Metric: Macro-averaged F1 (%) of the predicted 1-5 novelty score against the reviewers' gold score on the RINoBench test split (one fifth of the 1,381 ICLR 2022-2023 research ideas, gold scores from human reviewers), zero-shot with the five-point novelty rubric and the idea's related works; higher is better. Source: arxiv.org. Saturation forecast: Around 2035. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 17.2 | #91 |
| 2 | O3 | 16.2 | #121 |
| 3 | GPT-OSS-120B | 14.6 | #330 |
| 4 | Llama 3.1 8B | 14.6 | #1139 |
| 5 | Llama 4 Scout | 13 | #646 |
| 6 | DeepSeek R1 | 12.3 | #245 |
| 7 | Llama 3.3 70B Instruct | 9.5 | #520 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=rinobench · How It Works · Data refreshed daily, snapshot 2026-10-11.