Summary of Seal: Semantic-augmented Imitation Learning Via Language Model, by Chengyang Gu et al.
SEAL: SEmantic-Augmented Imitation Learning via Language Model
by Chengyang Gu, Yuxin Pan, Haotian Bai, Hui Xiong, Yize Chen
First submitted to arxiv on: 3 Oct 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG); Systems and Control (eess.SY)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper presents a novel framework called SEAL, which leverages Large Language Models (LLMs) to tackle challenging decision-making tasks. Specifically, SEAL uses LLMs’ semantic knowledge to specify sub-goal spaces and pre-label states, allowing for more robust sub-goal representations without prior task hierarchy knowledge. The proposed dual-encoder structure combines supervised LLM-guided sub-goal learning with unsupervised Vector Quantization (VQ) for improved representation accuracy. Additionally, SEAL incorporates a transition-augmented low-level planner to enhance adaptation to sub-goal transitions. Experimental results show that SEAL outperforms state-of-the-art Hierarchical Imitation Learning (HIL) methods and LLM-based planning approaches, particularly in settings with limited expert data and complex long-horizon tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Seal is a new way to help machines make decisions when they don’t have all the information. It uses big language models that know a lot about words and meanings to help figure out what sub-goals are important for making good decisions. This framework, called SEAL, is special because it can learn from small amounts of expert data and even adapt to changing situations. The results show that SEAL is better than other methods at solving complex decision-making problems. |
Keywords
» Artificial intelligence » Encoder » Quantization » Supervised » Unsupervised
