Summary of Improving Machine Translation with Human Feedback: An Exploration Of Quality Estimation As a Reward Model, by Zhiwei He et al.
Improving Machine Translation with Human Feedback: An Exploration of Quality Estimation as a Reward Model
by Zhiwei He, Xing Wang, Wenxiang Jiao, Zhuosheng Zhang, Rui Wang, Shuming Shi, Zhaopeng Tu
First submitted to arxiv on: 23 Jan 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
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 approach utilizes quality estimation (QE) models as the reward model to predict human preferences for feedback training in machine translation. The QE model has achieved impressive alignment with human evaluations, but its vulnerability can lead to overoptimization and error propagation when used as a reward model. To address this issue, the authors adopt heuristic rules to detect incorrect translations and assign penalty terms to their reward scores. Experimental results show consistent and significant improvements across various settings, further verified through human preference studies. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine learning researchers have found a way to improve translation quality using human feedback! They’re trying to make computers better at translating languages. To do this, they use something called quality estimation (QE) models. These models are really good at predicting how good a translation is, but sometimes they can get it wrong and make things worse. The researchers figured out that if they add some extra rules to catch mistakes, their system will work way better! They tested it on lots of different translations and it actually got better as time went on. |
Keywords
* Artificial intelligence * Alignment * Machine learning * Translation
