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
- FLASH: GPU-native deformable sim learns cloth folding in minutes on one RTX 5090 — ChongZzZhang · 2026-10-26
- Luke Wroblewski spots it: iPhone Duo's corner radius mirrors the original iPhone — LukeW · 2026-10-26
- Humanoids shattered forecasts: half-marathon record in 2026, 100m faster than Bolt — ben_j_todd · 2026-10-26
- Humanoid walking is continuous correction — watch recovery, not locomotion, in demos — Olivier__OG · 2026-10-26
- CircuitHub unveils The Grid, a factory-scale robot making PCBs in the US twice as fast as Shenzhen — garrytan · 2026-10-26
- Shanghai AI Lab-led team releases 100-hour egocentric wearable dataset for humanoid training — ChuckDBrooks · 2026-10-26