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

Large vision-language models (LVLMs) have recently witnessed rapidadvancements, exhibiting a remarkable capacity for perceiving, understanding,and processing visual information by connecting visual receptor with largelanguage models (LLMs). However, current assessments mainly focus onrecognizing and reasoning abilities, lacking direct evaluation ofconversational skills and neglecting visual storytelling abilities. In thispaper, we propose an evaluation method that uses strong LLMs as judges tocomprehensively evaluate the various abilities of LVLMs. Firstly, we constructa comprehensive visual dialogue dataset TouchStone, consisting of open-worldimages and questions, covering five major categories of abilities and 27subtasks. This dataset not only covers fundamental recognition andcomprehension but also extends to literary creation. Secondly, by integratingdetailed image annotations we effectively transform the multimodal inputcontent into a form understandable by LLMs. This enables us to employ advancedLLMs for directly evaluating the quality of the multimodal dialogue withoutrequiring human intervention. Through validation, we demonstrate that powerfulLVLMs, such as GPT-4, can effectively score dialogue quality by leveragingtheir textual capabilities alone, aligning with human preferences. We hope ourwork can serve as a touchstone for LVLMs' evaluation and pave the way forbuilding stronger LVLMs. The evaluation code is available athttps://github.com/OFA-Sys/TouchStone.

TouchStone: Evaluating Vision-Language Models by Language Models