ECAC: leaderboard
Teaching toy recognition on ECAC (2026), an early childhood education captioning benchmark of kindergarten daily-activity photos: the model captions each test image zero-shot, and the Teaching Toy Recognition Score is the share of expert-reviewed teaching-toy names the caption mentions, matched at three naming-precision levels. Percent; higher is better.
Metric: Teaching Toy Recognition Score (TTS, %): share of the teaching-toy names annotated for each test image (GPT-4o draft names at low, medium and high precision, reviewed by early childhood education experts) that the generated caption names, by string matching against each name's accepted descriptions; zero-shot image captioning of kindergarten daily-activity photos; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 4 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2.5 VL 7B Instruct | 45.25 | #1902 |
| 2 | Qwen 2.5 VL 3B Instruct | 38.07 | #2713 |
| 3 | GLM-4V-9B | 29.36 | #3142 |
| 4 | Gemma 3 27B | 10.74 | #1663 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=ecac · How It Works · Data refreshed daily, snapshot 2026-11-21.