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

Knowledge-grounded dialogue generation aims to mitigate the issue of textdegeneration by incorporating external knowledge to supplement the context.However, the model often fails to internalize this information into responsesin a human-like manner. Instead, it simply inserts segments of the providedknowledge into generic responses. As a result, the generated responses tend tobe tedious, incoherent, and in lack of interactivity which means thedegeneration problem is still unsolved. In this work, we first find that suchcopying-style degeneration is primarily due to the weak likelihood objective,which allows the model to "cheat" the objective by merely duplicating knowledgesegments in a superficial pattern matching based on overlap. To overcome thischallenge, we then propose a Multi-level Adaptive Contrastive Learning (MACL)framework that dynamically samples negative examples and subsequently penalizesdegeneration behaviors at both the token-level and sequence-level. Extensiveexperiments on the WoW dataset demonstrate the effectiveness of our approachacross various pre-trained models.

Multi-level Adaptive Contrastive Learning for Knowledge Internalization in Dialogue Generation