Task Tokens: A Flexible Approach to Adapting Behavior Foundation Models
This addresses the challenge of adapting behavior foundation models to specific control tasks for robotics/AI researchers, representing an incremental improvement over existing prompting methods.
The paper tackles the problem of behavior foundation models requiring meticulous prompt engineering for specific tasks by introducing Task Tokens, a method that learns a task-specific encoder through reinforcement learning while keeping the original model frozen, demonstrating efficacy across various tasks including out-of-distribution scenarios.
Recent advancements in imitation learning have led to transformer-based behavior foundation models (BFMs) that enable multi-modal, human-like control for humanoid agents. While excelling at zero-shot generation of robust behaviors, BFMs often require meticulous prompt engineering for specific tasks, potentially yielding suboptimal results. We introduce "Task Tokens", a method to effectively tailor BFMs to specific tasks while preserving their flexibility. Our approach leverages the transformer architecture of BFMs to learn a new task-specific encoder through reinforcement learning, keeping the original BFM frozen. This allows incorporation of user-defined priors, balancing reward design and prompt engineering. By training a task encoder to map observations to tokens, used as additional BFM inputs, we guide performance improvement while maintaining the model's diverse control characteristics. We demonstrate Task Tokens' efficacy across various tasks, including out-of-distribution scenarios, and show their compatibility with other prompting modalities. Our results suggest that Task Tokens offer a promising approach for adapting BFMs to specific control tasks while retaining their generalization capabilities.