# Dual-camera video model boosts robot learning across embodiments

> Researchers show a multi‑view latent action model improves robot success rates by up to 20% compared to single‑view approaches.

Oossa · 2026-10-21 · https://oossa.com/en/dual-camera-video-model-boosts-robot-learning-across-embodiments

Oliver Hausdörfer and his team released a new latent action model that uses synchronized views from two cameras. The model infers 6‑degree‑of‑freedom actions more accurately, raising decoding accuracy by 17% over single‑view versions. When plugged into a robot policy, success rates on benchmark tasks rose 13.5%, compared with 6.2% for the single‑view baseline. In real‑world tests with a robot copying human‑hand demonstrations, the dual‑view model added 20% more successful rollouts across 850 attempts.

## The facts

- Published on 2026-10-21
- Dual‑view LAM improves decoding accuracy by 17% and policy success by up to 20%

## Why it matters

The improvement means robots can learn new skills from ordinary videos with fewer demonstrations, helping labs with limited tele‑operation data.

## Sources & references

1. [Multi-View Latent Action Pretraining for Cross-Embodiment Robot Learning from Videos](https://arxiv.org/abs/2610.27794) – arXiv cs.RO, 2026-10-21

Last updated: 2026-10-21