Actionpiece improves how robots learn and execute physical actions
ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models
RoboticsArtificial IntelligenceComputation and LanguageComputer Vision and Pattern Recognition
Summary
Current models that teach robots to perform actions often break those actions into tokens to predict and recreate motions. However, small errors in these tokens might not capture how accurately the robot can adjust actions between different situations. The authors introduce a new way to measure and improve how well these tokens preserve the physical relationships between actions, leading to more faithful reproduction. Their method, called ActionPiece, trains models to keep the right order and distance between actions, which helps robots succeed more often in performing tasks.
What this means in practice
- •For robotics developers: Enable robots to learn and perform physical tasks more reliably by improving action token representation and execution accuracy.
- •For automation engineers: Create better control systems that preserve nuanced physical action differences for complex automated workflows.
Authors
Shijie Lian, Bin Yu, Zhaolong Shen, Xiaopeng Lin, Yichao Du, Zhirui Zhang, Laurence T. Yang, Kai Chen
Abstract
Action tokenizers play a central role in autoregressive vision-language-action (VLA) models, determining both the targets for policy training and the executable commands recovered from predicted tokens. Their fidelity is commonly evaluated using pointwise reconstruction metrics such as mean squared error (MSE), yet small individual errors do not fully characterize how faithfully action adjustments across demonstrations are preserved. After compression, similar actions may still cluster around a representative motion, while the adjustments needed for different contexts are diminished, distorted, or even reversed. We introduce physical rank consistency (PRC) to measure how well tokenization preserves local physical distance rankings after reconstruction. Evaluating decoded actions provides a common reference across token vocabularies and decoder architectures, complementing pointwise accuracy with a measure of relational fidelity. We further present ActionPiece, which preserves physical action relationships through joint supervision of representation learning and quantization. Physical rank preservation supervises near-far ordering in encoder and quantized feature distances, while quantization regularization applies the same ordering to codeword assignment distributions. Both objectives augment reconstruction, producing discrete action tokens for standard autoregressive policy learning and execution through a frozen decoder. Under the same Qwen3-VL-4B policy training setup, ActionPiece achieves 94.8% on LIBERO and 68.8% on unseen LIBERO-Plus, with additional evaluations reaching 71.9% on SimplerEnv and 51.5% across VLA-Arena L0-L2. Component ablations show that the two objectives jointly improve PRC and policy success, demonstrating the value of physical relationship supervision for action tokenization.