{"id":"arxiv:2211.10869","title":"UniMASK: Unified Inference in Sequential Decision Problems","date":"2022-11-20","section":"Reinforcement learning","quality":null,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/","source_url":"https://arxiv.org/abs/2211.10869","text_available":true,"authors":["Micah Carroll","Orr Paradise","Jessy Lin","Raluca Georgescu"],"outgoing":[{"source":"arxiv:2211.10869","target":"arxiv:2005.01643","evidence":"id","id":"arxiv:2005.01643","title":"Offline Reinforcement Learning: Tutorial, Review, and Perspectives on Open Problems","url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/2ae228e6f0a418b4cd46f033b57c7452ba84a9c8cb985b198224186333c17d97/"}],"incoming":[],"text_parts":[{"start":0,"end":15540,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html"},{"start":15540,"end":31494,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html"},{"start":31494,"end":47020,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-3.html"},{"start":47020,"end":50391,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-4.html"}],"contents":[{"title":"Uni​ [MASK] : Unified Inference in Sequential Decision Problems","offset":0,"level":1,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"Abstract","offset":67,"level":6,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"1 Introduction","offset":1001,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"2 Related Work","offset":5440,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"3 The Uni​ [MASK] Framework","offset":8587,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"3.1 Tasks as Masking Schemes","offset":10241,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"3.2 Model Architecture & Training Regimes","offset":11116,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"3.2.1 Training regimes","offset":12373,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"3.2.2 Hypotheses","offset":13967,"level":4,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"4 A Unified Model for Any Inference Task","offset":14918,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-1.html#text"},{"title":"4.1 One Model to Rule Them All","offset":16128,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"4.2 Future State Predictions","offset":17430,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"4.3 Measuring Single-Task Performance","offset":18574,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"5 Trajectory Generation in a Complex Environment","offset":20668,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"5.1 Environment Setup","offset":20896,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"5.2 Models Trained","offset":22123,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"5.3 Results","offset":23643,"level":3,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"6 Limitations and Future Work","offset":26281,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"7 Conclusion","offset":29440,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"Acknowledgments and Disclosure of Funding","offset":30299,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"Appendix A Code","offset":30773,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"Appendix B Training regimes","offset":30978,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-2.html#text"},{"title":"Appendix C The random masking scheme","offset":33524,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-3.html#text"},{"title":"Appendix D Model architecture","offset":34884,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-3.html#text"},{"title":"Appendix E Minigrid experiments","offset":37887,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-3.html#text"},{"title":"Appendix F Additional Minigrid experiments","offset":42627,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-3.html#text"},{"title":"Appendix G Our Decision-GPT model","offset":44841,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-3.html#text"},{"title":"Appendix H Maze2D experiments","offset":45215,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-3.html#text"},{"title":"Instructions for reporting errors","offset":49427,"level":2,"url":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/part-4.html#text"}],"text_code_points":50391,"source_markdown":"https://vinnik-dmitry07.github.io/favorite-papers/agents/papers/919bff55efff4ed3549bd106b843afa3d140b32938f5283c77821d7fd569452e/source.md"}