BARISTA - Relation Extraction: 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 micro F1 in percent for extracting typed relation tuples between scene entities; higher is better.

Metric: Micro F1 (%, times 100) of extracted subject, target, relation and value tuples under exact multiset 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

#ModelSettingScoreOverall rank
1Gemma 4 31B75.1#382
2Qwen 3.5 27B74.5#375
3Gemini 3 FlashLow67.5#175
4Qwen 3.5 9B61.3#828
5GPT-5.4 MiniLow56.5#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-relation-extraction · How It Works · Data refreshed daily, snapshot 2026-11-21.