Towards Flexible Inference in Sequential Decision Problems via Bidirectional Transformers
This provides a flexible, unified approach for researchers and practitioners in reinforcement learning and decision-making, though it is incremental as it adapts existing masking techniques from language models.
The paper tackles the problem of unifying various sequential decision-making tasks by proposing the FlexiBiT framework, which uses bidirectional transformers to train a single model that achieves performance similar to or better than specialized models across tasks like behavior cloning and offline RL.
Randomly masking and predicting word tokens has been a successful approach in pre-training language models for a variety of downstream tasks. In this work, we observe that the same idea also applies naturally to sequential decision making, where many well-studied tasks like behavior cloning, offline RL, inverse dynamics, and waypoint conditioning correspond to different sequence maskings over a sequence of states, actions, and returns. We introduce the FlexiBiT framework, which provides a unified way to specify models which can be trained on many different sequential decision making tasks. We show that a single FlexiBiT model is simultaneously capable of carrying out many tasks with performance similar to or better than specialized models. Additionally, we show that performance can be further improved by fine-tuning our general model on specific tasks of interest.