Real-Time EXPO-FT: async VLA proposals plus RL critics unlock real-time π0.5
philfung · x · 2026-09-22
- Real-Time EXPO-FT makes slow VLAs usable in real time by splitting inference into two paths:
- Async: a frozen VLA generates candidate proposal chunks
- Sync: two small RL-tuned nets (an edit policy and a Q-critic) adjust and select proposals using current observations
- This unlocks π0.5 on dynamic tasks such as balancing a ball on a plate and striking a ball into a goal.
- A practical route to deploying heavyweight VLA policies on real-time robots.
More from Embodied
- Google opens preorders for Googlebook, an Android-based laptop built for Gemini — GeminiApp · 2026-09-22
- Suspicious 'Googlebook' laptop preorder page surfaces at $899, likely parody — ZoubinGhahrama1 · 2026-09-22
- Diag2Diag: AI generates measurements that hardware sensors can't capture — AnneliesGamble · 2026-09-22
- Someone gave a robot dog its own crypto wallet to manage — adamamcbride · 2026-09-22
- Apple's John Ternus defends Vision Pro: it's the early days of a long journey — asymco · 2026-09-22
- 4D Gaussian Splatting streaming demoed on a phone — janusch_patas · 2026-09-22