# Language-Forcing boosts humanoid robot task success

> The new training method lifts whole‑body robot success from 53% to 71% on real tasks, with no extra cost at run time.

Oossa · 2026-10-20 · https://oossa.com/en/language-forcing-boosts-humanoid-robot-task-success

Researchers led by Yanjun Li introduced LANGUAGE‑FORCING, a way to train robot world‑action models using language descriptions of task progress. The method aligns text embeddings from a frozen vision‑language model with the robot’s predicted future visuals. It only adds overhead during training, so the robot runs as fast as before. On the HumanoidArena benchmark it reached a 79.17% success rate, beating the next best policy by 9.46 points. In five real‑world tasks the success rose from 53.33% to 71.33%.

## The facts

- Paper published 2026-10-20 ("Tue Oct 20 2026").
- Success on HumanoidArena: 79.17%, 9.46 points above the strongest competitor.

## Why it matters

For developers of humanoid robots, the technique offers higher task reliability without slowing down the robot during operation.

## Sources & references

1. [Language-Forcing: Advancing Humanoid Loco-Manipulation via Semantic Alignment of Whole-Body World Action Models](https://arxiv.org/abs/2610.23894) – arXiv cs.RO, 2026-10-20

Last updated: 2026-10-20