Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2402.10450 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917685857615872 |
|---|---|
| author | Zheng, Ruijie Cheng, Ching-An Daumé III, Hal Huang, Furong Kolobov, Andrey |
| author_facet | Zheng, Ruijie Cheng, Ching-An Daumé III, Hal Huang, Furong Kolobov, Andrey |
| contents | Temporal action abstractions, along with belief state representations, are a powerful knowledge sharing mechanism for sequential decision making. In this work, we propose a novel view that treats inducing temporal action abstractions as a sequence compression problem. To do so, we bring a subtle but critical component of LLM training pipelines -- input tokenization via byte pair encoding (BPE) -- to the seemingly distant task of learning skills of variable time span in continuous control domains. We introduce an approach called Primitive Sequence Encoding (PRISE) that combines continuous action quantization with BPE to learn powerful action abstractions. We empirically show that high-level skills discovered by PRISE from a multitask set of robotic manipulation demonstrations significantly boost the performance of both multitask imitation learning as well as few-shot imitation learning on unseen tasks. Our code is released at https://github.com/FrankZheng2022/PRISE. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_10450 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | PRISE: LLM-Style Sequence Compression for Learning Temporal Action Abstractions in Control Zheng, Ruijie Cheng, Ching-An Daumé III, Hal Huang, Furong Kolobov, Andrey Machine Learning Temporal action abstractions, along with belief state representations, are a powerful knowledge sharing mechanism for sequential decision making. In this work, we propose a novel view that treats inducing temporal action abstractions as a sequence compression problem. To do so, we bring a subtle but critical component of LLM training pipelines -- input tokenization via byte pair encoding (BPE) -- to the seemingly distant task of learning skills of variable time span in continuous control domains. We introduce an approach called Primitive Sequence Encoding (PRISE) that combines continuous action quantization with BPE to learn powerful action abstractions. We empirically show that high-level skills discovered by PRISE from a multitask set of robotic manipulation demonstrations significantly boost the performance of both multitask imitation learning as well as few-shot imitation learning on unseen tasks. Our code is released at https://github.com/FrankZheng2022/PRISE. |
| title | PRISE: LLM-Style Sequence Compression for Learning Temporal Action Abstractions in Control |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2402.10450 |