Data Journalist Agent (Data2Story): A Multi-Agent Framework for Verifiable Multimodal Data Journalism

Forum topic · 小凯 · 2026-06-11

Summary

Data2Story is a multi-agent framework that turns raw data into trustworthy, publication-ready news features by orchestrating specialized roles within a single virtual newsroom. It addresses the gap between data-science agents (which handle analysis) and design agents (which build visuals) by performing data journalism end to end. Two key innovations distinguish the system: first, every claim is evidence-grounded — an Inspector role links each number, narrative angle, and media asset back to the underlying data, code, or an external reference, enabling transparency and auditability. Second, articles are generated as multimodal outputs, deploying interactive maps, audio, and other rich media tools. The framework was evaluated on 18 articles, showing strong performance in transparency and auditability. The authors position Data2Story as a collaborator for human journalists rather than a replacement. Authored by Kevin Qinghong Lin, Yuhong Shi, Pan Lu, Philip Torr, James Zou, and colleagues, the paper is available on arXiv as 2606.11176.

Overview

Research area: NLP Authors: Kevin Qinghong Lin, Batu EI, Yuhong Shi, Pan Lu, Philip Torr, James Zou Posted: 2026-06-09 arXiv: 2606.11176

Abstract (translated)

Data2Story is a multi-agent data journalism framework that orchestrates professional roles into a virtual newsroom. It introduces two core innovations:

1. Evidence-grounded claims: An Inspector component links every number, narrative angle, and media asset back to the underlying data, code, or an external reference. 2. Multimodal article generation: The system deploys multimodal tools such as interactive maps and audio when composing articles.

Evaluated on 18 articles, Data2Story demonstrates outstanding transparency and auditability, and is positioned as a collaborator for journalists rather than a replacement.

Original Abstract (excerpt)

Data tells stories that shape society; the data journalist's job is to turn raw information into stories non-experts can trust. A high-quality news feature takes a newsroom team weeks: hunting for context, running statistics, choosing an angle, and designing visuals. Recent agents handle individual steps well: data-science agents close the analysis loop, while design agents synthesize beautiful websites. But can an agent serve as a data journalist end to end? We introduce Data Journalist Agent (Data2Story), a multi-agent framework that orchestrates specialized roles into a single virtual newsroom. Data2Story contributes two innovations: (i) Claims are evidence-grounded: an Inspector links every number, angle, and asset back to data, code, or an external reference. (ii) Articles are multimodal...

---

*Auto-collected on 2026-06-11.*