UniMASK: Unified Inference in Sequential Decision Problems

arxiv:2211.10869

Original source · Paper JSON

Reinforcement learning · 2022-11-20

Cites (1) · Cited by (0)

Complete cleaned Markdown · Source Markdown including bibliography

Read in parts

1. Part 1
2. Part 2
3. Part 3
4. Part 4

Contents

• Uni​ [MASK] : Unified Inference in Sequential Decision Problems
• Abstract
• 1 Introduction
• 2 Related Work
• 3 The Uni​ [MASK] Framework
• 3.1 Tasks as Masking Schemes
• 3.2 Model Architecture & Training Regimes
• 3.2.1 Training regimes
• 3.2.2 Hypotheses
• 4 A Unified Model for Any Inference Task
• 4.1 One Model to Rule Them All
• 4.2 Future State Predictions
• 4.3 Measuring Single-Task Performance
• 5 Trajectory Generation in a Complex Environment
• 5.1 Environment Setup
• 5.2 Models Trained
• 5.3 Results
• 6 Limitations and Future Work
• 7 Conclusion
• Acknowledgments and Disclosure of Funding
• Appendix A Code
• Appendix B Training regimes
• Appendix C The random masking scheme
• Appendix D Model architecture
• Appendix E Minigrid experiments
• Appendix F Additional Minigrid experiments
• Appendix G Our Decision-GPT model
• Appendix H Maze2D experiments
• Instructions for reporting errors