TF-RefusalBench - Summarization Disclaimer: leaderboard
Metric: Disclaimer rate (%; share of 2,800 summarization prompts over Swiss Federal Supreme Court criminal-law passages in which the model performs the task but appends a warning or caveat about the sensitive content; 3 samples per prompt at temperature 0.7, 2-of-3 majority of three LLM judges). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-OSS-20B | 0 |
| 2 | Qwen 3 32B | 0 |
| 3 | Llama 3.3 70B Instruct | 3.7 |
| 4 | Gemma 4 31B (IT) | 8.2 |
Interactive version: theaggregate.ai/benchmark?slug=tf-refusalbench-summarization-disclaimer · How It Works · Data refreshed daily, snapshot 2026-09-26.