LGFeb 11, 2025
A Survey of In-Context Reinforcement Learning
arXiv:2502.07978v134 citationsh-index: 67
Originality Synthesis-oriented
AI Analysis
It provides a review for researchers interested in parameter-free RL methods, but it is incremental as it only summarizes existing work.
The paper surveys in-context reinforcement learning, where agents solve new tasks without updating network parameters by conditioning on context like action-observation histories.
Reinforcement learning (RL) agents typically optimize their policies by performing expensive backward passes to update their network parameters. However, some agents can solve new tasks without updating any parameters by simply conditioning on additional context such as their action-observation histories. This paper surveys work on such behavior, known as in-context reinforcement learning.