CSSBench - Illegal Activities and Compliance: leaderboard

Metric: Attack success rate (%; share of obfuscated malicious illegal activities and compliance queries answered unsafely, Qwen3Guard-Gen-8B judge; lower is better). Source: arxiv.org. Saturation forecast: Estimated already saturated. 10 models tracked.

Top models

#ModelScore
1Qwen 3 1.7B33.33
2Qwen 3 8B36
3Qwen 3 0.6B42.5

Interactive version: theaggregate.ai/benchmark?slug=cssbench-illegal-activities-and-compliance · How It Works · Data refreshed daily, snapshot 2026-09-26.