DementiaCare-Bench - 128 Frames: leaderboard
DementiaCare-Bench evaluates understanding of dementia-care videos on 2,023 multiple-choice questions. Each series fixes the frame budget across models; zero frames is the text-only control. Model-specific best-frame selections and their gains are not pooled. This series reports DementiaCare-Bench - 128 Frames from 2609.12929v1, Sx4.T2: Accuracy (%). Higher is better.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around February 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 3 27B (IT) | 85.1 | #530 |
| 2 | Claude Sonnet 4.6 | 83.1 | #97 |
| 3 | Qwen 3 VL 235B A22B | 82 | #253 |
| 4 | Gemini 2.5 Flash | 80.8 | #255 |
| 5 | GPT-4o | 78.6 | #358 |
| 6 | GLM-4.6V | 71.1 | #331 |
| 7 | InternVL3-8B | 65.7 | #623 |
| 8 | MiniCPM-V-2.6 | 58.9 | #831 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=dementiacare-bench-128-frames · How It Works · Data refreshed daily, snapshot 2026-11-07.