NVIDIA's ASPIRE v1.0 Solves Rubik's Cube, Hits 99% on 80 Robot Tasks
Summary
NVIDIA researchers release ASPIRE v1.0, an Apache-2.0 framework that treats robot learning as skill refinement rather than gradient descent. NVlabs reports 99% success on 80 LIBERO-Pro tasks and Rubik's Cube solving, with Claude Opus 5.5 reaching 97 to 100% across three benchmarks. It warns generated code runs without a security sandbox.
Key Points
- NVlabs packages the simulation workflow as agent skills for Claude Code and Codex.
- The real-robot recipes are distilled from demo sessions, not a new real-robot validation, NVlabs says.
- Six contributors from NVIDIA, Michigan, UC Berkeley and Carnegie Mellon develop ASPIRE.