RuleWeaver - Cross-Source - Rule Precision: leaderboard
Metric: Correctly applied share of cited rules (%). Source: arxiv.org. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 (Non-reasoning) | 78.64 |
| 2 | Claude Opus 4.6 (Non-reasoning) | 73.74 |
| 3 | Gemini 3.1 Pro (Preview) (Low) | 72.95 |
| 4 | GPT-5.5 (Non-reasoning) | 69.48 |
| 5 | GLM-5 (Non-reasoning) | 69.21 |
| 6 | DeepSeek V4 Pro (Non-reasoning) | 60.1 |
| 7 | Kimi K2.6 (Non-reasoning) | 56.53 |
| 8 | Claude Sonnet 4.6 (Non-reasoning) | 54.74 |
| 9 | MiniMax-M2.7 | 43.06 |
Interactive version: theaggregate.ai/benchmark?slug=ruleweaver-cross-source-rule-precision · How It Works · Data refreshed daily, snapshot 2026-09-19.