← all releases
Harvard University and Stanford University·Feb 4, 2026·v1.0·open source
OAT: Ordered Action Tokenization
OAT is a learned action tokenizer designed to enhance autoregressive policies in robot learning by discretizing continuous action sequences into ordered tokens. It addresses challenges in action tokenization by providing reasonable compression, universal decodability, and a left-to-right causally ordered token space.
RoboticsRL
overview
OAT is a learned action tokenizer designed to enhance autoregressive policies in robot learning by discretizing continuous action sequences into ordered tokens. It addresses challenges in action tokenization by providing reasonable compression, universal decodability, and a left-to-right causally ordered token space.
key features
- 01Transformer-based register tokens
- 02Finite scalar quantization
- 03Ordering-inducing training mechanisms
related productsbrowse all →