Search papers, labs, and topics across Lattice.
This paper introduces WIZARD, a weight-space meta-learning framework that enables robotic models to adapt to new tasks without the need for task-specific demonstrations or action annotations. By generating task-specific LoRA parameters for a frozen Vision-Language-Action (VLA) policy based solely on language instructions and a short demonstration video, WIZARD achieves significant performance improvements on unseen tasks and datasets. Experimental results demonstrate that WIZARD can enhance performance by up to 14x on new tasks, showcasing its effectiveness in real-world applications beyond simulation environments.
WIZARD achieves up to 14x better performance on unseen tasks by generating task-specific adaptation weights in a single forward pass, eliminating the need for costly fine-tuning.
Vision-Language-Action (VLA) models are emerging as a promising paradigm for robotic manipulation, enabling general-purpose policies trained from large corpora of demonstrations and action labels. However, adapting these models to new tasks still typically requires task-specific demonstrations, action annotations, and additional fine-tuning, making deployment costly and difficult to scale. We propose WIZARD, a weight-space meta-learning framework that sidesteps task-specific fine-tuning by generating task-specific LoRA parameters for a frozen VLA policy. Given only a language instruction and a short demonstration video, WIZARD predicts the corresponding adaptation weights in a single forward pass, without target-task action labels or test-time optimization. During meta-training, WIZARD learns to map task evidence directly to expert LoRA updates, capturing relationships between tasks in weight space. Experiments on LIBERO show that WIZARD improves performance by up to ~2x on unseen dataset collections and up to ~14x on unseen tasks. On a Franka Emika Panda, WIZARD consistently improves over a real-domain adapted baseline, showing that generated adapters provide task-level specialization beyond simulation.
