FASTer: Toward Efficient Autoregressive Vision Language Action Modeling via neural Action Tokenization

2025-12-04

AI Summary

Autoregressive vision-language-action (VLA) models face a fundamental trade-off between reconstruction fidelity and inference efficiency in action tokenization. To address this, we propose FASTer—a novel end-to-end action modeling framework that encodes continuous action sequences as single-channel action images. FASTer integrates a learnable tokenizer, vector quantization (VQ), and block-wise autoregressive decoding, augmented by a lightweight action expert network. This design enables high-fidelity action reconstruction at aggressive compression ratios while significantly improving cross-task and cross-morphology generalization. Evaluated on both simulation and real-robot benchmarks, FASTer achieves higher task success rates with substantially faster inference speed, consistently outperforming existing state-of-the-art methods.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Autoregressive vision-language-action (VLA) models have recently demonstrated strong capabilities in robotic manipulation. However, their core process of action tokenization often involves a trade-off between reconstruction fidelity and inference efficiency. We introduce FASTer, a unified framework for efficient and generalizable robot learning that integrates a learnable tokenizer with an autoregressive policy built upon it. FASTerVQ encodes action chunks as single-channel images, capturing global spatio-temporal dependencies while maintaining a high compression ratio. FASTerVLA builds on this tokenizer with block-wise autoregressive decoding and a lightweight action expert, achieving both faster inference and higher task performance. Extensive experiments across simulated and real-world benchmarks show that FASTerVQ delivers superior reconstruction quality, high token utilization, and strong cross-task and cross-embodiment generalization, while FASTerVLA further improves overall capability, surpassing previous state-of-the-art VLA models in both inference speed and task performance.

Problem

• Improves action tokenization efficiency in VLA models
• Enhances robot task performance with faster inference
• Achieves superior cross-task generalization in robotics

Innovation

• Encodes action chunks as single-channel images
• Uses block-wise autoregressive decoding for efficiency
• Integrates lightweight action expert for performance