UbuntuGuard - Cross-Lingual: leaderboard
Metric: F1 (%) for PASS/FAIL policy-compliance classification of user-agent dialogues, English policies with dialogues in ten African languages. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Llama 3.3 70B Instruct | 87.62 |
| 2 | gpt-oss-safeguard-20B | 84.85 |
| 3 | DeepSeek V3.1 | 83.14 |
| 4 | GPT-OSS-20B | 81.36 |
| 5 | GPT-OSS-120B | 80.25 |
Interactive version: theaggregate.ai/benchmark?slug=ubuntuguard-cross-lingual · How It Works · Data refreshed daily, snapshot 2026-09-26.