Beyond Mimicry: Toward Lifelong Adaptability in Imitation Learning
This addresses the problem of imitation learning agents failing in shifting contexts for researchers and practitioners, though it is incremental as it outlines a research agenda rather than presenting a fully developed solution.
The paper argues that imitation learning has been optimized for perfect replay rather than adaptability, and proposes a research agenda to shift toward compositional adaptability by learning behavioral primitives that can be recombined in novel contexts without retraining.
Imitation learning stands at a crossroads: despite decades of progress, current imitation learning agents remain sophisticated memorisation machines, excelling at replay but failing when contexts shift or goals evolve. This paper argues that this failure is not technical but foundational: imitation learning has been optimised for the wrong objective. We propose a research agenda that redefines success from perfect replay to compositional adaptability. Such adaptability hinges on learning behavioural primitives once and recombining them through novel contexts without retraining. We establish metrics for compositional generalisation, propose hybrid architectures, and outline interdisciplinary research directions drawing on cognitive science and cultural evolution. Agents that embed adaptability at the core of imitation learning thus have an essential capability for operating in an open-ended world.