Interpretive Canons (BVerfG) - Grammatical Interpretation: leaderboard
Grammatical interpretation from the interpretive-canons benchmark on decisions of the German Federal Constitutional Court (Ringe, ICML 2026 AI4Law Workshop): does an argument rest on the wording of the provision. Positive-class F1 in percent on 120 held-out sentence pairs, with one expert-written prompt for every model; higher is better.
Metric: Positive-class F1 (%; grammatical interpretation (wortlaut): whether the argument rests on the wording of the provision, sentence-level binary classification in German Federal Constitutional Court decisions; expert-written prompt with worked examples (the same for every model), reasoning models at temperature 1.0; decision-disjoint held-out test split (60% of each subtask, 1:3 positives to negatives)). Source: arxiv.org. Saturation forecast: Estimated already saturated. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 86.7 | #181 |
| 2 | MiniMax-M3 | 83.9 | #265 |
| 3 | Gemini 3.5 Flash Lite | 79.4 | #300 |
| 4 | DeepSeek V4 Flash | 78.1 | #254 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=interpretive-canons-bverfg-grammatical-interpretation · How It Works · Data refreshed daily, snapshot 2026-11-12.