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
- Physical world agent swarm demo shared on X — pduan · 2026-10-26
- Looking for a wearable mic with best battery life that feeds recordings to personal agents via MCP — altryne · 2026-10-26
- 'Lamp Swarm': a desk of lamp-shaped robots shows what physical AI work looks like — dee_hw · 2026-10-26
- Tencent's HiSTEVE 1.0: hierarchical VLA agent plays Minecraft with native keyboard and mouse, hits 85% SOTA — _akhaliq · 2026-10-26
- SJTU's VaM lets robots learn online from mistakes, 95%+ success over 4-hour autonomous runs — chris_j_paxton · 2026-10-26
- DIY humanoid PhoneGirl day 4: collision tests, servo rework, 160° knee bend — _Stocko_ · 2026-10-26