BARISTA - Hand-Object Interaction: leaderboard
BARISTA (2026) is an egocentric benchmark of 185 coffee-preparation videos with verified per-frame scene graphs, from which zero-shot language tasks for vision-language models are derived. This board is interaction F1 in percent for recognizing which hand touches which object, with box matching; higher is better.
Metric: Interaction F1 (%, times 100) for recognizing which hand interacts with which object, with IoU at least 0.5 box matching, over 1,000 frozen frames on BARISTA, an egocentric coffee-preparation video dataset with verified per-frame scene graphs, zero-shot with shared task prompts, temperature 0.7 (GPT-5.4 mini at its fixed sampling), low thinking effort for the closed models, mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemma 4 31B | 75.8 | #382 | |
| 2 | Qwen 3.5 27B | 75.3 | #375 | |
| 3 | Gemini 3 Flash | Low | 74.9 | #175 |
| 4 | Qwen 3.5 9B | 64.2 | #828 | |
| 5 | GPT-5.4 Mini | Low | 19.3 | #406 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=barista-hand-object-interaction · How It Works · Data refreshed daily, snapshot 2026-11-21.