LO
CapabilityPhysical automation2025-03-06
A humanoid research robot achieved an 81% success rate in a human-robot bartending scenario after training on a new dataset of glasses, researchers reported
Bartender / baristaoccupation page →
Event date / reported
2025-03-06
Evidence stage
CapabilityA demo, benchmark or paper shows the task can be done. Updates what the technology can do — not what employers will do.
Tasks this bears on
Making coffee and drinks
Pulling espresso, steaming milk, mixing cocktails and pouring beer and wine to a consistent standard.
Automating≈ Platform inference
Where this applies
A research paper noting that object detectors often fail to distinguish glasses because they are transparent and reflective. The authors collected a dataset of 7,850 images of glasses on a humanoid robot platform and, using a model trained on it, the robot achieved an 81% success rate in a human-robot bartending scenario. A laboratory result, not a bar.
What this means
Even recognising a glass is still hard for robots; in the lab one gets bartending right about four times in five.
What it does not yet show
A lab study on one robot; it does not show robots working in bars.
What you can check
Open arXiv 2503.04308 and find "success rate of 81% in a human-robot bartending scenario".
Does it change the assessment?
No — and this stage does not move it either. A "Capability" record is real evidence, but it does not upgrade a task judgement on its own. The 1 linked judgement above stand where they were.
Source
Gajdošech, Ali, Habekost, Madaras, Kerzel, Wermter — Shaken, Not Stirred: A Novel Dataset for Visual Understanding of Glasses in Human-Robot Bartending Tasks, arXiv 2503.04308 (6 Mar 2025) · verified 2026-09-30 · Claude (VOLO agent) · interpreted 2026-09-30 · Claude (VOLO agent)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.