Scaling robot data is essential for building generalist Vision-Language-Action (VLA) models, but collecting robot trajectories is costly and sparse compared to web-scale image-text data. This makes representation quality a key bottleneck: under a fixed data budget, continued pre-training must turn limited trajectories into transferable visual-action knowledge rather than merely fitting actions.
To address this, researchers propose VLAct, a VLA-oriented VLM backbone trained on broad, heterogeneous, multi-embodiment robot data before task-specific fine-tuning. VLAct preserves the broad VLM prior and encourages shared action semantics across embodiments through three mechanisms: VLM-prior preservation, multi-head continuous action co-supervision, and a partially unified cross-embodiment action layout, while allowing task-specific action heads during fine-tuning.
Across simulation, real-world, and unseen-embodiment transfer, VLAct consistently improves downstream performance under fixed fine-tuning protocols. On LIBERO-Plus and RoboTwin 2.0, VLAct surpasses industrial VLA systems including ABot-M0 and LingBot-VLA, achieving success rates of 82.6% and 92.5%. On RoboDojo, VLAct ranks sixth among all policies by success rate and outperforms all explicitly designated world-action model (WAM) entries on both metrics.
Most notably, on RoboCasa-GR1, an unseen humanoid embodiment, VLAct using only 20% of downstream trajectories outperforms the full-data GR00T-N1.6 baseline. These results are obtained using fully open-source data and only a 16-GPU training setup, showing that representation-centric continued pre-training can deliver highly competitive performance under a modest compute budget and is an important independent axis of VLA progress beyond data scaling.
VLAct is strong, open, and research-friendly: it releases the data, models, and complete training/fine-tuning pipeline, with full continued pre-training requiring only 16 GPUs.