Provable Interactive Learning with Hindsight Instruction Feedback
This work addresses the problem of reducing supervision costs in interactive learning for AI agents, though it is incremental as it builds on existing theoretical frameworks with a specialized assumption.
The paper tackles interactive learning where an agent generates responses to instructions, proposing a method that uses hindsight instruction feedback instead of expert supervision, and shows that their LORIL algorithm achieves regret scaling as √T independent of response space size under a low-rank assumption, with experiments demonstrating outperformance over baselines.
We study interactive learning in a setting where the agent has to generate a response (e.g., an action or trajectory) given a context and an instruction. In contrast, to typical approaches that train the system using reward or expert supervision on response, we study learning with hindsight instruction where a teacher provides an instruction that is most suitable for the agent's generated response. This hindsight labeling of instruction is often easier to provide than providing expert supervision of the optimal response which may require expert knowledge or can be impractical to elicit. We initiate the theoretical analysis of interactive learning with hindsight labeling. We first provide a lower bound showing that in general, the regret of any algorithm must scale with the size of the agent's response space. We then study a specialized setting where the underlying instruction-response distribution can be decomposed as a low-rank matrix. We introduce an algorithm called LORIL for this setting and show that its regret scales as $\sqrt{T}$ where $T$ is the number of rounds and depends on the intrinsic rank but does not depend on the size of the agent's response space. We provide experiments in two domains showing that LORIL outperforms baselines even when the low-rank assumption is violated.