Anchor-Align lifts xArm7 real-robot success from 28% to 54%
Dwip Dalal · hf · 2026-07-23
What it changes
This paper proposes Anchor-Align, a finetuning recipe for vision-language-action (VLA) policies that tries to preserve pretrained generalization while learning robot control.
- Standard behavior cloning overwrites useful VLM representations.
- The method adds two objectives:
- Vision-Language Anchoring: distills layer-wise features from a frozen VLM copy to reduce representation drift.
- Language-Action Alignment: maps action targets to discrete motion-direction labels and trains language and action prediction on the same observation.
- On a physical xArm7 robot, it raises success rates from 28% → 54% and 37% → 60% across two VLA architectures.
- In simulation, it improves robustness on OOD perturbations, perceptual shifts, and long-horizon control across LIBERO-PRO, LIBERO-Plus, and CALVIN.
More from Embodied
- A robot finger actuation demo answers the question with a simple yes — IanPritchard · 2026-07-23
- Michelle Sun says software engineers should pivot into hardware for the next decade — Fowe · 2026-07-23
- Kepler v0.1 trains robot vision, touch, and pose into one shared world model — freelerobot · 2026-07-23
- Humanoid robots move from lab demos to real household appliance handling — CurieuxExplorer · 2026-07-23
- iFlytek-backed robot startup bets on world models, body awareness and less real-robot data — 量子位 · 2026-07-23
- A phone-only agent played Space Invaders and solved mini crossword in 16 ms loops — AI Engineer · 2026-07-23