LIBERO-PRO

'LIBERO-Pro' is a benchmark dataset used to evaluate one-step action generation in vision-language-action models, containing various tasks that assess the performance of these models in generating actions based on rich observations and language inputs.

Papers using LIBERO-PRO (3)

Playful Agentic Robot Learning2026ASPIRE: Agentic /Skills Discovery for Robotics2026Learning More from Less: Reinforcement Learning from Hindsight2026