Robotics paper says dense patch features beat bigger vision-language models
stepjamUK · x · 2026-07-22
A robotics paper argues that precise manipulation does not need a billion-parameter vision-language model if the policy can preserve dense spatial features.
What Patch Policy changes
- It avoids pooling observations into a single CLS/global vector.
- It also avoids training a new visual backbone from scratch.
- Instead, it feeds frozen dense patch features from a pretrained ViT directly into a transformer policy head.
Why it helps
The method keeps spatial attention within each frame bidirectional, while using causal masking across frames so temporal order stays intact.
Reported results
- Across four simulated and three real-world suites, dense patch features beat global-pooled baselines by 40% relative.
- Compared with a fine-tuned OpenVLA-OFT, Patch Policy is 18% better with about 0.7% of the parameters and 11 ms inference.
- Replacing patch features with pooled or convolutionally compressed ones hurts performance across the board.
The author’s takeaway: for manipulation, precision comes from preserving patch-level structure, not from making the backbone bigger.
Related event: Patch Policy Outperforms 7B VLA with 0.7% Parameters(5 posts)→
More from Embodied
- Amazon and Google sold 600M+ smart speakers, so why no AGI-era successor? — julianlehr · 2026-09-11
- ECCV26 Oral: Flow Matching Enables Single-Stage Multi-View Point Cloud Registration — ducha_aiki · 2026-09-11
- Polish developers build iPhone app that detects nearby Meta smart glasses — Low-Honeydew6483 · 2026-09-11
- Ant's Afu health AI hits 150M users, unveils AI+hardware health alliance at Bund Summit — APPSO · 2026-09-11
- Johns Hopkins Launches Full-Stack Hands-on Robot Learning Class with SO-101 Arm Kits — _krishna_murthy · 2026-09-11
- SyncWorld: In-Context Robot World Model Simulates Unseen Views and Embodiments Zero-Shot — ChongZzZhang · 2026-09-11