Sequential Dexterity

Stanford

Chains dexterous manipulation policies into longer task sequences, with simulation training and real-system implementation code.

What you can explore

Explore transitions between manipulation skills and failures that arise when policies are chained.

Before you use it

Uses a legacy Isaac Gym Preview environment and separately downloaded checkpoints. The README names MIT but the repository LICENSE is Apache 2.0; resolve that inconsistency before redistribution.

Original contributors

Yuanpei Chen, Chen Wang, Li Fei-Fei and C. Karen Liu

License / access: LICENSE file: Apache 2.0 · README: MIT; inconsistent statements

Original code & documentation

Research overview

Editorial reference · checked 2026-09-22