Emergence of Abstract State Representations in Embodied Sequence Modeling
Submission Type: Regular Long Paper
Submission Track: Language Grounding to Vision, Robotics and Beyond
Submission Track 2: Interpretability, Interactivity, and Analysis of Models for NLP
Keywords: Interpretability and Analysis; Decision Making via Sequence Modeling; Language Grounding to Vision and Beyond
TL;DR: We investigate the impacts of states and natural language instructions on what knowledge Transformer-based sequence models can learn.
Abstract: Decision making via sequence modeling aims to mimic the success of language models, where actions taken by an embodied agent are modeled as tokens to predict. Despite their promising performance, it remains unclear if embodied sequence modeling leads to the emergence of internal representations that represent the environmental state information. A model that lacks abstract state representations would be liable to make decisions based on surface statistics which fail to generalize. We take the BabyAI environment, a grid world in which language-conditioned navigation tasks are performed, and build a sequence modeling Transformer, which takes a language instruction, a sequence of actions, and environmental observations as its inputs. In order to investigate the emergence of abstract state representations, we design a "blindfolded" navigation task, where only the initial environmental layout, the language instruction, and the action sequence to complete the task are available for training. Our probing results show that intermediate environmental layouts can be reasonably reconstructed from the internal activations of a trained model, and that language instructions play a role in the reconstruction accuracy. Our results suggest that many key features of state representations can emerge via embodied sequence modeling, supporting an optimistic outlook for applications of sequence modeling objectives to more complex embodied decision-making domains.
Submission Number: 4558
Loading