English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Insufficient modeling of human preferences within the reward model is a majorobstacle for leveraging human feedback to improve translation quality.Fortunately, quality estimation (QE), which predicts the quality of a giventranslation without reference, has achieved impressive alignment with humanevaluations in the last two years. In this work, we investigate the potentialof employing the QE model as the reward model (the QE-based reward model) topredict human preferences for feedback training. We first identify theoveroptimization problem during QE-based feedback training, manifested as anincrease in reward while translation quality declines. We examine the problemand argue that the vulnerability of the QE model might lead to high rewards forincorrect translations, resulting in overoptimization and error propagation. Toaddress the problem, we adopt a simple yet effective method that uses heuristicrules to detect the incorrect translations and assigns a penalty term to theQE-based rewards for the detected incorrect translations. Experimental resultsshow that the proposed QE-based feedback training achieves consistent andsignificant improvements across various settings, further verified throughhuman preference studies. Our subsequent analysis demonstrates the high dataefficiency of the proposed QE-based feedback training: the proposed approachusing a small amount of monolingual data can outperform systems using largerparallel corpora.

Improving Machine Translation with Human Feedback: An Exploration of Quality Estimation as a Reward Model