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

In this paper, we study how to improve the zero-shot reasoning ability oflarge language models~(LLMs) over structured data in a unified way. Inspired bythe study on tool augmentation for LLMs, we develop an \emph{IterativeReading-then-Reasoning~(IRR)} approach for solving question answering tasksbased on structured data, called \textbf{StructGPT}. In our approach, weconstruct the specialized function to collect relevant evidence from structureddata (\ie \emph{reading}), and let LLMs concentrate the reasoning task based onthe collected information (\ie \emph{reasoning}). Specially, we propose an\emph{invoking-linearization-generation} procedure to support LLMs in reasoningon the structured data with the help of the external interfaces. By iteratingthis procedures with provided interfaces, our approach can gradually approachthe target answer to a given query. Extensive experiments conducted on threetypes of structured data demonstrate the effectiveness of our approach, whichcan significantly boost the performance of ChatGPT and achieve comparableperformance against the full-data supervised-tuning baselines. Our codes anddata are publicly available at~\url{https://github.com/RUCAIBox/StructGPT}.

StructGPT: A General Framework for Large Language Model to Reason over Structured Data