SciPaths: leaderboard
Scientific progress depends on sequences of enabling contributions, yet existing AI4Science benchmarks largely focus on citation prediction, literature retrieval, or idea generation rather than the dependencies that make progress possible. In this paper, we introduce discovery pathway forecasting: given a target scientific contribution and the prior literature available at a specified time, the ta
Metric: F1 (self-reported). Source: benchmarklist.com. Saturation forecast: Around May 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 18.9 | #58 |
| 2 | GPT-5.4 | 15.2 | #79 |
| 3 | GPT-4.1 | 11.1 | #248 |
| 4 | GPT-5 Mini | 8.8 | #186 |
| 5 | gemma-4-E4B-it | 6.1 | #566 |
| 6 | Qwen 3 4B Instruct | 6 | #657 |
| 7 | GPT-4o | 5.6 | #343 |
| 8 | Llama 3 8B Instruct | 4.1 | #1129 |
| 9 | Llama 3.1 8B Instruct | 2.9 | #1035 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=scipaths · How It Works · Data refreshed daily, snapshot 2026-10-20.