Interactive Inverse Reinforcement Learning of Interaction Scenarios via Bi-level Optimization
Authors: Yue Mao, Shicheng Liu, Siyuan Xu et al.
Summary
The authors extend inverse reinforcement learning (IRL) to interactive settings where the learner actively interacts with an expert rather than passively observing demonstrations. Traditional IRL just watches expert behavior and infers their reward function; interactive IRL (IIRL) has the learner trying to learn both the expert's reward function and a good policy for interacting with them simultaneously. They formulate this as a bi-level optimization problem and develop an algorithm (BISIRL) with formal convergence guarantees.
Main takeaways:
- Traditional IRL is passive (observe expert demonstrations); interactive IRL has the learner actively interact with the expert
- Formulated as bi-level optimization: lower level learns reward function, upper level learns interaction policy
- BISIRL algorithm solves this with inner loop (reward learning) and outer loop (policy learning)
- Formal convergence guarantees provided
- Validated through experiments on interactive scenarios
Relevance
Not directly related to my persona work—this is about reinforcement learning in interactive settings. Only tangentially relevant if I ever frame persona installation as inferring a "reward function" that explains assistant behavior.
Abstract
arXiv:2605.08131v1 Announce Type: new Abstract: Inverse reinforcement learning (IRL) learns a reward function and a corresponding policy that best fit the demonstration data of an expert. However, in the current IRL setting, the learner is isolated from the expert and can only passively observe the expert demonstrations. This limits the applicability of IRL to interactive settings, where the learner actively interacts with the expert and needs to infer the expert's reward function from the interactions. To bridge the gap, this paper studies interactive IRL (IIRL) where a learner aims to learn the reward function of an expert and a policy to interact with the expert during its interactions with the expert. We formulate IIRL as a stochastic bi-level optimization problem where the lower level learns a reward function to explain the behaviors of the expert, and the upper level learns a policy to interact with the expert. We develop a double-loop algorithm, Bi-level Interactive Scenarios Inverse Reinforcement Learning (BISIRL), which solves the lower-level problem in the inner loop and the upper-level problem in the outer loop. We formally guarantee that BISIRL converges and validate our algorithm through extensive experiments.