Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
StarVLA is introduced as an open-source codebase designed to unify and streamline Vision-Language-Action (VLA) research by providing a modular backbone-action-head architecture that supports both Vision-Language Model (VLM) and world-model backbones. It offers reusable training strategies like cross-embodiment learning and multimodal co-training, and integrates major benchmarks through a unified evaluation interface. Experiments show that simple training recipes within StarVLA match or exceed the performance of prior methods on multiple benchmarks, demonstrating its effectiveness and potential for lowering barriers to VLA research.
Stop reinventing the wheel for VLA research: StarVLA offers a modular, open-source framework that already matches or beats existing methods with minimal data engineering.
Building generalist embodied agents requires integrating perception, language understanding, and action, which are core capabilities addressed by Vision-Language-Action (VLA) approaches based on multimodal foundation models, including recent advances in vision-language models and world models. Despite rapid progress, VLA methods remain fragmented across incompatible architectures, codebases, and evaluation protocols, hindering principled comparison and reproducibility. We present StarVLA, an open-source codebase for VLA research. StarVLA addresses these challenges in three aspects. First, it provides a modular backbone--action-head architecture that supports both VLM backbones (e.g., Qwen-VL) and world-model backbones (e.g., Cosmos) alongside representative action-decoding paradigms, all under a shared abstraction in which backbone and action head can each be swapped independently. Second, it provides reusable training strategies, including cross-embodiment learning and multimodal co-training, that apply consistently across supported paradigms. Third, it integrates major benchmarks, including LIBERO, SimplerEnv, RoboTwin~2.0, RoboCasa-GR1, and BEHAVIOR-1K, through a unified evaluation interface that supports both simulation and real-robot deployment. StarVLA also ships simple, fully reproducible single-benchmark training recipes that, despite minimal data engineering, already match or surpass prior methods on multiple benchmarks with both VLM and world-model backbones. To our best knowledge, StarVLA is one of the most comprehensive open-source VLA frameworks available, and we expect it to lower the barrier for reproducing existing methods and prototyping new ones. StarVLA is being actively maintained and expanded; we will update this report as the project evolves. The code and documentation are available at https://github.com/starVLA/starVLA.