NVIDIA's VoLo lands at CoRL 2026: a VLM agent that orchestrates robots through long-horizon tasks
erwincoumans · x · 2026-09-08
VoLo, a CoRL 2026 paper from NVIDIA and University of Michigan, frames open-vocabulary long-horizon manipulation as "physical orchestration": a VLM agent with memory plans, monitors, detects failures, recovers, and steers heterogeneous capabilities (VLA/WAMs, SAM3, grasp/place primitives) as interruptible tools—crucially without pausing the physical world.
- Introduces RoboVoLo, a high-fidelity simulation benchmark with 126 tasks spanning common sense, memory/state tracking, complex references, and world knowledge, with failure-mode diagnostics.
- Work led by Siyi Chen during an NVIDIA internship; project page, arXiv paper, and code are public.
More from Embodied
- OpenAI Offers Up to $423,000 Plus Equity for Robot Actuator Engineers — VraserX · 2026-09-08
- TrainBud: Open-Source MCP Server Turns Your Fitness Data Into an AI You Can Ask From Your Wrist — Cr1tsh0t · 2026-09-08
- MiTaS fuses slow and fast tactile sensing, lifts robot manipulation success to 80% vs 31% — Jan_R_Peters · 2026-09-08
- Schmidhuber slams Chamath's 'AGI has arrived': no AGI without mastering the physical world — SchmidhuberAI · 2026-09-08
- Robot spider goes musical as "Maestro Webber", playing Hotel California notes — creatoroff · 2026-09-08
- Meta Remotely Disables Recording on Thousands of Tampered Smart Glasses Hiding LED Indicators — Polymarket · 2026-09-08