Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

NVIDIA's ASPIRE v1.0 frames robot learning as agentic skill discovery

AI summary

Researchers from NVIDIA, the University of Michigan, UIUC, UC Berkeley and Carnegie Mellon released ASPIRE v1.0, an open agentic skill-discovery framework for robot manipulation that treats "training" as iterative skill refinement rather than gradient descent, and a "trained model" as a repository of sensorimotor skills rather than floating weights. The project page says v1.0 can solve a Rubik's Cube and reaches 99% success on 80 LIBERO-Pro Goal, Object, Spatial and Long tasks plus selected Robosuite and BEHAVIOR-1K tasks, all self-reported with no independent reproduction supplied. Simulation workflows ship as agent skills for Claude Code and Codex, NVIDIA-owned material is under Apache 2.0, and the project warns that ASPIRE executes language-model-generated Python and is not a security sandbox.

ASPIRE: Agentic /Skills Discovery for Robotics

link

https://research.nvidia.com/labs/gear/aspire/


ASPIRE: Agentic /Skills Discovery for Robotics

Project Page  |  Paper

#NVIDIA#ASPIRE#LIBERO-Pro#Robosuite#BEHAVIOR-1K#Claude Opus 5.5#GPT-6 Astra#Claude Code#Codex#University of Michigan#UC Berkeley#Carnegie Mellon University

Original · 4 min readRead the original
Group chat

No comments yet. Start the conversation.