Real-Time EXPO-FT: Fast, Reliable RL for Real-Time VLA Policies
Loading video
Loading videoReal-Time EXPO-FT is a fast and reliable reinforcement learning fine-tuning method for real-time VLA policies. It unlocks pi0.5 on challenging dynamic tasks such as balancing a ball on a plate and striking a ball into a goal, pairing slow VLA inference with high-frequency action generation.
Category: research
Author: @perryadong
Date: 2026-09-21T00:00:00
Duration: 59.881s
