act_subtask_g2 is an open-source model for imitation learning in robotics, focusing on predicting and executing short action chunks. It is intended for robotics researchers and developers who need to train or evaluate robotic policies using teleoperated data.
Act Subtask G2 is an Other AI product. It focuses on enabling robots to learn and execute action sequences from teleoperated data. It is built as an open-source project for robotics researchers and developers. Act Subtask G2 is open source under the Apache-2.0 license. The product ships for the web and the command line, and it can be self-hosted.
Behind Act Subtask G2 is Haku-2004, and the product first shipped in 2024. Development happens publicly on GitHub with 25.7k stars and 164 commits in the last 90 days. PulseGate's similarity index finds few close equivalents — Act Subtask G2 occupies a relatively distinct niche. Key capabilities include imitation learning, robotics policy, and action chunking.
Latest indexed changes and source events
Haku-2004/act_subtask_g2 discovered by the PulseGate indexer
Other apps tracked under the same category.
