Humanoid Policy ∼ \sim Human Policy

Ri-Zhao Qiu, Shiqi Yang, Xuxin Cheng, Chaitanya Chawla, Jialong Li, Tairan He, Ge Yan, David J. Yoon, Ryan Hoque, Lars Paulsen, Ge Yang, Jian Zhang, Sha Yi, Guanya Shi, Xiaolong Wang · 2025

Read Humanoid Policy ∼ \sim Human Policy by Ri-Zhao Qiu, Shiqi Yang, Xuxin Cheng, Chaitanya Chawla, Jialong Li, Tairan He, Ge Yan, David J. Yoon, Ryan Hoque, Lars Paulsen, Ge Yang, Jian Zhang, Sha Yi, Guanya Shi, Xiaolong Wang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-13441.txt.

Description

Ri-Zhao Qiu and a large team teach humanoid robots from human video in a 2025 paper. Training robots on robot demos is slow and costly, since each demo needs a human operator steering the machine. First-person human video is far easier to gather at scale, but bodies differ, and that gap can poison the training.

The team collected a first-person data set of task demos aligned with matched humanoid demos. They then shrank the body gap on both sides: cleaning the data and shaping the model. A single policy learns from both human and robot examples, treating human motion as a close cousin of the robot’s own. Early results suggest human video can carry much of the training load. That would cut the cost of teaching new skills to humanoids.

Catalog record

CREATORRi-Zhao Qiu, Shiqi Yang, Xuxin Cheng, Chaitanya Chawla, Jialong Li, Tairan He, Ge Yan, David J. Yoon, Ryan Hoque, Lars Paulsen, Ge Yang, Jian Zhang, Sha Yi, Guanya Shi, Xiaolong Wang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-13441

Browse the full library · About the Tower of Babel