Unitree Go2 robot asks clarifying questions via Gemini and Bayesian active learning
DynamicWebPaige · x · 2026-11-10
DynamicWebPaige (Google) spotlights a conference project giving robots the sense of when to ask a clarifying question: Gemini Pro and Flash-Lite with Bayesian active learning disambiguate human intent in real time on a Unitree Go2 quadruped. The paper's framework: an LLM proposes candidate task specs with priors, a Grammar VAE embeds them in latent space, a Gaussian Process models utility, and the acquisition function picks the most informative contrast to phrase as a clarification question; the converged spec is decoded for a formal planner with a safety filter.
She also highlights SimFoundry from Fei-Fei Li's team: Gemini handles scene layout and object articulation while NanoBanana generates assets, automatically producing interactive 3D simulation scenes for policy learning.
More from Embodied
- Amazon Ring unveils $250 autonomous home security drone that patrols and streams live video — Polymarket · 2026-11-10
- NVIDIA-led study: unblocking VLM gradients in VLA training lifts manipulation progress from 10.3% to 32.9% — stepjamUK · 2026-11-10
- Google DeepMind launches Gemini Robotics 2.5 with 96% accuracy sorting Mercedes-Benz parts — GoogleDeepMind · 2026-11-10
- SlashVibe is a physical desk controller for managing multiple running AI coding agents — Scobleizer · 2026-11-10
- Path Robotics raises $300M Series E at $2.45B valuation to expand physical AI — Rewkang · 2026-11-10
- SimFoundry auto-builds interactive 3D sim scenes from video for robot policy training — DynamicWebPaige · 2026-11-10