ACT
Summary
Action Chunking with Transformers is an imitation-learning method that predicts future continuous action chunks and temporally ensembles overlapping predictions.
Role In The Wiki
ACT anchors action chunking before the later diffusion/flow robotics wave. It is useful for separating “chunked continuous control” from “diffusion or flow denoising.”
Evidence
Relation To Foundation TSFM Agenda
Use the source-level agenda mapping in act-2023 rather than duplicating verdict rows here.
At the entity level, ACT anchors action chunking before the later diffusion/flow robotics wave. It is useful for separating “chunked continuous control” from “diffusion or flow denoising.” This page should stay as the object card; source pages carry slot-level verdicts, evidence, and missing pieces.