One policy controlling 4 different robot hands for in-hand manipulation rejected by CoRL

YuXiang_IRVL · x · 2026-09-05

The author laments that one of his group's most exciting works was rejected by CoRL reviewers and the AC, receiving 3 borderline scores after rebuttal mainly due to limited experiments. The work's action representation lets a single policy control 4 different robot hands for in-hand manipulation, and generalizes beyond cubes to other objects. The author asks if anyone has seen similar work.

Original post →

More from Embodied

Embodied channel →