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

Pre-trained multimodal models have achieved significant success inretrieval-based question answering. However, current multimodal retrievalquestion-answering models face two main challenges. Firstly, utilizingcompressed evidence features as input to the model results in the loss offine-grained information within the evidence. Secondly, a gap exists betweenthe feature extraction of evidence and the question, which hinders the modelfrom effectively extracting critical features from the evidence based on thegiven question. We propose a two-stage framework for evidence retrieval andquestion-answering to alleviate these issues. First and foremost, we propose aprogressive evidence refinement strategy for selecting crucial evidence. Thisstrategy employs an iterative evidence retrieval approach to uncover thelogical sequence among the evidence pieces. It incorporates two rounds offiltering to optimize the solution space, thus further ensuring temporalefficiency. Subsequently, we introduce a semi-supervised contrastive learningtraining strategy based on negative samples to expand the scope of the questiondomain, allowing for a more thorough exploration of latent knowledge withinknown samples. Finally, in order to mitigate the loss of fine-grainedinformation, we devise a multi-turn retrieval and question-answering strategyto handle multimodal inputs. This strategy involves incorporating multimodalevidence directly into the model as part of the historical dialogue andquestion. Meanwhile, we leverage a cross-modal attention mechanism to capturethe underlying connections between the evidence and the question, and theanswer is generated through a decoding generation approach. We validate themodel's effectiveness through extensive experiments, achieving outstandingperformance on WebQA and MultimodelQA benchmark tests.

Progressive Evidence Refinement for Open-domain Multimodal Retrieval Question Answering