CVGEvalKit: leaderboard
CVGEvalKit evaluates lightweight language models on criminal court view generation from case facts, with the same setup also used for charge prediction. The benchmark is text-based, and the pinned ROUGE-1 metric measures overlap between generated court views and reference text. Source: BenchmarkList’s attributed result table.
Metric: ROUGE-1. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 1B (IT) | 79.03 | #1481 |
| 2 | Qwen 3 1.7B | 37.34 | #1175 |
| 3 | Qwen 3 0.6B | 3.36 | #1448 |
| 4 | Llama 3.2 1B Instruct | 3.31 | #1518 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cvgevalkit · How It Works · Data refreshed daily, snapshot 2026-11-05.