Summary of Binary Reward Labeling: Bridging Offline Preference and Reward-based Reinforcement Learning, by Yinglun Xu et al.
Binary Reward Labeling: Bridging Offline Preference and Reward-Based Reinforcement Learning
by Yinglun Xu, David Zhu, Rohan Gumaste, Gagandeep Singh
First submitted to arxiv on: 14 Jun 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 The proposed framework bridges the gap between offline reinforcement learning (RL) and preference-based offline RL by transforming preference feedback into scalar rewards via binary reward labeling. This enables the application of existing reward-based offline RL algorithms to datasets with preference feedback. The framework minimizes information loss during the signal transition, and its connection to recent PBRL techniques is theoretically shown. By combining reward labeling with different algorithms, new and potentially more efficient offline PBRL algorithms can be developed. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Offline reinforcement learning has a practical setting where machines learn from data without interacting with the environment. A challenge arises when trying to apply this learning to situations that use “preference” feedback instead of traditional rewards. This paper proposes a way to fix this problem by changing preference feedback into reward signals, allowing existing algorithms to work on new types of data. |
Keywords
* Artificial intelligence * Reinforcement learning
