arXiv 2302.04449

Read and Reap the Rewards: Learning to Play Atari with the Help of Instruction Manuals

By Yue Wu, Yewen Fan, et al.

Published 2023-02-09

Citation lineage

Review the prior work and downstream research connected to this paper.

High sample complexity has long been a challenge for RL. On the other hand, humans learn to perform tasks not only from interaction or demonstrations, but also by reading unstructured text documents, e.g., instruction manuals. Instruction manuals and wiki pages are among the most abundant data that could inform agents of valuable features and policies or task-specific environmental dynamics and reward structures. Th…

View the original paper on arXiv