Summary of Hindsight Planner: a Closed-loop Few-shot Planner For Embodied Instruction Following, by Yuxiao Yang et al.
Hindsight Planner: A Closed-Loop Few-Shot Planner for Embodied Instruction Following
by Yuxiao Yang, Shenao Zhang, Zhihan Liu, Huaxiu Yao, Zhaoran Wang
First submitted to arxiv on: 27 Dec 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Robotics (cs.RO)
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 research proposes a task planner for Embodied Instruction Following (EIF) using Large Language Models (LLMs). Unlike previous works, which treat planning as a supervised task, this approach frames it as a Partially Observable Markov Decision Process (POMDP) and aims to develop a robust planner under a few-shot assumption. The proposed closed-loop planner includes an adaptation module and a novel hindsight method to utilize available information. Experimental results on the ALFRED dataset show that the planner achieves competitive performance under the few-shot assumption, even surpassing the full-shot supervised agent in some cases. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study creates a task planner for teaching robots using big language models. Normally, planners are trained like students, following what experts do. But this plan makes it more robust by treating planning as a puzzle that needs solving. It proposes a special type of planner with two new parts: an adaptation module and hindsight method. These help the planner make better choices even when things don’t go exactly as planned. The researchers tested their planner on a specific dataset and found it works just as well, if not better, than previous methods. |
Keywords
» Artificial intelligence » Few shot » Supervised
