Conversational visualization via analytic task reasoning with large language models

Xiangcheng Zhang

doi:10.1016/j.cag.2026.104746

Conversational visualization via analytic task reasoning with large language models

2026-09-08

0

PRE

AI

DOI：10.1016/j.cag.2026.104746

Original

Original request for help

Save

Abstract

En 中文

• Introduces analytic task reasoning for conversational NL2VIS. • Presents TaskDialogData, an extensive dataset with multi-turn dialogues. • Proposes TaskDialogViz with stepwise reasoning and Step-DPO. • Outperforms rule-based, prompt, and chain-of-thought baselines. • Demonstrates improved coherence and intent understanding in user study.

Keywords:

Natural Language to Visualization

Large language models

Analytic task reasoning

Conversational visualization via analytic task reasoning with large language models

Abstract

Journal

Organization

Cited Papers