SciLaws-Bench - Real Symbolic Novelty: leaderboard
Metric: Score (%). Source: arxiv.org. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 50.77 |
| 2 | Gemini 3.5 Flash | 46.08 |
| 3 | GLM-5.2 | 45.84 |
| 4 | Claude Opus 4.8 | 44.94 |
| 5 | DeepSeek V4 Pro | 43.96 |
| 6 | Qwen 3.7 Max | 43.8 |
| 7 | GPT-5 Mini | 40.69 |
| 8 | GPT-5.4 Mini | 37.17 |
| 9 | GPT-4o Mini | 20.59 |
Interactive version: theaggregate.ai/benchmark?slug=scilaws-bench-real-symbolic-novelty · How It Works · Data refreshed daily, snapshot 2026-09-21.