CIBuzzBench - Harmfulness Detection (English Prompt): leaderboard
Metric: Macro F1 (%; binary harmful or non-harmful classification of the documented online sense from the buzzword alone, 14.2% of entries harmful; English instructions; 3,001 Chinese internet buzzwords; zero-shot, temperature 0; the Chinese buzzword is the input under both prompt languages). Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 80.45 |
| 2 | GPT-5.5 | 76.23 |
| 3 | DeepSeek V4 Pro | 75.36 |
| 4 | Qwen 3.7 Max | 74 |
| 5 | GLM-5.2 | 72.94 |
| 6 | Claude Opus 4.8 | 71.09 |
| 7 | Qwen 3 14B | 59.09 |
| 8 | Qwen 3 8B | 58.62 |
| 9 | Qwen 3 4B | 28.77 |
Interactive version: theaggregate.ai/benchmark?slug=cibuzzbench-harmfulness-detection-english-prompt · How It Works · Data refreshed daily, snapshot 2026-09-26.