In-Context Reinforcement Learning From Suboptimal Historical Data

Computer Science

• Historical Data 100%
• Reinforcement Learning 100%
• Superior Performance 50%
• maximum-likelihood 50%
• Likelihood Estimation 50%
• Process Problem 50%
• Markov Decision Process 50%
• Transformer Model 50%

Social Sciences

• Imitation Learning 100%
• Markov Decision Process 100%
• Maximum Likelihood Method 100%
• Transformer Model 100%

Keyphrases

• Weighted Maximum Likelihood 25%
• Autoregressive Training 25%
• Autoregressive Transformer 25%
• Advantage Function 25%