Bootstrapped Q-learning with Context Relevant Observation Pruning to Generalize in Text-based Games
This addresses generalization issues in text-based games for AI research, but it is incremental as it builds on existing Q-learning and pruning techniques.
The paper tackles the problem of reinforcement learning methods failing to generalize in text-based games, especially with limited data, by proposing a bootstrapped Q-learning approach with observation pruning, resulting in improved generalization using 10x-20x fewer training games than prior methods.
We show that Reinforcement Learning (RL) methods for solving Text-Based Games (TBGs) often fail to generalize on unseen games, especially in small data regimes. To address this issue, we propose Context Relevant Episodic State Truncation (CREST) for irrelevant token removal in observation text for improved generalization. Our method first trains a base model using Q-learning, which typically overfits the training games. The base model's action token distribution is used to perform observation pruning that removes irrelevant tokens. A second bootstrapped model is then retrained on the pruned observation text. Our bootstrapped agent shows improved generalization in solving unseen TextWorld games, using 10x-20x fewer training games compared to previous state-of-the-art methods despite requiring less number of training episodes.