OmniJARVIS: Unified Vision-Language-Action Tokenization Enables Open-World Instruction Following Agents

This paper presents \method, a novel Vision-Language-Action (VLA) model for open-world instruction-following agents in Minecraft. Compared to prior works that either emit textual goals to separate controllers or produce the control command directly,\method seeks a different path to ensure both strong reasoning and efficient decision-making capabilities via \textit{unified} tokenization of \textbf{multimodal interaction data}. First, we introduce a \textit{self-supervised} approach to learn a behavior encoder that produces discretized tokens for behavior trajectories $\tau = \{o_0, a_0, \dots\}$ and an imitation learning (IL) policy decoder conditioned on these tokens. These additional \textit{behavior tokens} will be augmented to the vocabulary of pretrained Multimodal Language Models (MLMs). With this encoder, we then pack long-term multimodal interactions involving task instructions, memories, thoughts, observations, textual responses, behavior trajectories, \etc into unified token sequences and model them with autoregressive transformers. Thanks to the semantically meaningful behavior tokens, the resulting VLA model, \method, can reason (by producing chain-of-thoughts), plan, answer questions, and act (by producing behavior tokens for the IL policy decoder). \method demonstrates excellent performances on a comprehensive collection of atomic, programmatic, and open-ended tasks in open-world Minecraft. Our analysis further unveils the crucial design principles in interaction data formation, unified tokenization, and its scaling potentials.The dataset, models, and code will be released at \url{https://craftjarvis.org/OmniJarvis/}.