LAP: Language-Action Pre-training Enables Zero-Shot Cross-Embodiment Transfer

SOTA VLA: 0% LAP-3B: 52% "How did the robot move between the two frames?""Move right 5cm" "Flip the bottle" "Move up 3cm, move right 2cm, tilt right 45 degrees" "Fold the towel" "Move forward 3cm, tilt forward 30 degrees" Robot Action PredictionMotion Prediction VQA 0.032 0.001 "Fold the towel"Fig. 1: We introduce Language-Action Pre-training (LAP), a general VLA pre-training recipe that represents low-level actions directly in natural language to supervise a vision-language backbone, and instantiate it as LAP-3B, the first VLA to demonstrate strong zero-shot transfer to novel embodiments.Compared to state-of-the-art VLAs, LAP-3B learns more generalizable embodiment representations and exhibits favorable scaling behavior.Abstract-A long-standing goal in robotics is a generalist policy that can be deployed zero-shot on new robot embodiments without per-embodiment adaptation.Despite large-scale multiembodiment pre-training, existing Vision-Language-Action models (VLAs) remain tightly coupled to their training embodiments and typically require costly fine-tuning.We introduce Language-Action Pre-training (LAP), a simple recipe that represents lowlevel robot actions directly in natural language, aligning action supervision with the pre-trained vision-language model's inputoutput distribution.LAP requires no learned tokenizer, no costly annotation, and no embodiment-specific architectural design.Based on LAP, we present LAP-3B, which to the best of our knowledge is the first VLA to achieve substantial zero-shot transfer to previously unseen single-arm manipulators without any embodiment-specific fine-tuning.Across multiple novel robots and manipulation tasks, LAP-3B attains over 50% average zero-shot success, delivering roughly a 2→ improvement over the strongest prior VLAs.We further show that LAP enables efficient adaptation and favorable scaling, while unifying action prediction and VQA in a shared language-action format that yields additional gains through co-training.