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/

Zendy Plus

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

Zendy Tools

Zendy Open

LLMs (Large Language Models) usually interact with users in the form ofdialogue and generate responses following their instructions, which naturallyrequire dialogue comprehension abilities. However, dialogue comprehension is ageneral language ability which is hard to be evaluated directly. In this work,we propose to perform the evaluation focusing on the factual consistency issuewith the help of the dialogue summarization task. Besides evaluating andanalyzing the dialogue summarization performance (DIAC-Sum) of different LLMs,we also derive factual questions from the generated summaries and use them as amore flexible measurement of dialogue comprehension (DIAC-QA). Our evaluationshows that, on average, 26.8% of the summaries generated by LLMs containfactual inconsistency. Even ChatGPT, the strongest model evaluated, has sucherrors in 16% of its summaries. For answering the factual questions, which ismore challenging, the average error rate of all evaluated LLMs is 36.1%. Bothresults indicate serious deficiencies. Detailed analysis shows that theunderstanding of subject/object of the conversation is still challenging forLLMs. Furthermore, to stimulate and enhance the dialogue comprehension abilityof LLMs, we propose a fine-tuning paradigm with auto-constructed multi-taskdata, which achieved a relative error rate reduction of 11% on DIAC-QA.

Exploring the Factual Consistency in Dialogue Comprehension of Large Language Models