Summary of Relic: a Recipe For 64k Steps Of In-context Reinforcement Learning For Embodied Ai, by Ahmad Elawady et al.
ReLIC: A Recipe for 64k Steps of In-Context Reinforcement Learning for Embodied AI
by Ahmad Elawady, Gunjan Chhablani, Ram Ramrakhya, Karmesh Yadav, Dhruv Batra, Zsolt Kira, Andrew Szot
First submitted to arxiv on: 3 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 ReLIC, a new approach for in-context reinforcement learning (RL) for embodied agents, enables rapid adaptation to new tasks by integrating long histories of experience into decision-making. The method outperforms meta-RL baselines in adapting to unseen houses in an embodied multi-object navigation task and demonstrates few-shot imitation learning capabilities despite never being trained with expert demonstrations. ReLIC combines large-scale RL training, a novel policy update scheme called “partial updates,” and a Sink-KV mechanism to utilize long observation histories for embodied agents. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary ReLIC helps robots learn quickly by combining lots of experience with special ways of updating their policies and using memories to help them decide what to do next. This makes the robot better at completing new tasks in new environments, like navigating an unfamiliar house. |
Keywords
* Artificial intelligence * Few shot * Reinforcement learning
