MedHarm - Refusal Adequacy: leaderboard

Metric: Refusal adequacy (%): mean judge score for refusals that block the requested harm and explain why (a bare refusal scores half); MedHarm: 1,100 high-risk medical queries in ten categories of 110 (covert poisoning, lethal dosing, lethal drug interactions, euthanasia, illegal organ harvesting and live anesthesia, drug abuse, medicalized chemical and biological weapons, date-rape drugs, fetal harm and more), all requests a model should decline or redirect; greedy decoding, 512 new tokens; GPT-5-mini judge (91% agreement with human labels on unsafe responses); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 15 models tracked.

Top models

#ModelScore
1GPT-5.3 Instant99.9
2Llama 3.1 8B Instruct98.1
3GPT-5.595.7
4Grok 4.392.4
5DeepSeek V4 Pro92.1
6Qwen 2.5 7B Instruct82.1

Interactive version: theaggregate.ai/benchmark?slug=medharm-refusal-adequacy · How It Works · Data refreshed daily, snapshot 2026-09-29.