RubRIX - Dementia Caregiver Queries: leaderboard
Risk in LLM answers to 10,321 real questions from dementia caregivers on ALZConnected, audited with RubRIX (Goel et al., 2026): 29 yes-or-no checks for inattention, bias and stigma, inaccurate information, uncritical affirmation and epistemic arrogance, flagged by a GPT-5-nano judge. Percentage of checks flagged per response, averaged over responses; lower is better.
Metric: Risk rate (% of the 29 RubRIX audit questions flagged per response by a GPT-5-nano judge, mean over responses; initial responses before any rubric-guided refinement; lower is better; 10,321 posts from ALZConnected, an online community for caregivers of people with Alzheimer's disease and related dementias). Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o Mini | 4 | #1627 |
| 2 | Claude Sonnet 4 | 5 | #397 |
| 3 | Qwen 3 4B | 6 | #2297 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=rubrix-dementia-caregiver-queries · How It Works · Data refreshed daily, snapshot 2026-11-12.