EvidFuse: Writing-Time Evidence Learning for Consistent Text-Chart Data Reporting

TL;DR

The paper tackles chart–text inconsistency in data-driven report generation by reframing the task as writing-time evidence construction. It introduces EvidFuse, a training-free, multi-agent framework that couples a Data-Augmented Analysis Agent with a Real-Time Evidence Construction Writer to interleave narrative drafting with on-demand, grounded visual evidence. Across multiple datasets and model configurations, EvidFuse achieves top performance in chart quality, text–chart alignment, and overall usefulness, outperforming staged baselines and showing robustness to backbone choices. The approach enables deeper, decision-oriented analysis by ensuring that narrative claims continually reference live visual evidence, illustrating a practical path toward grounded long-form generation in data-rich domains.

Abstract

Data-driven reports communicate decision-relevant insights by tightly interleaving narrative text with charts grounded in underlying tables. However, current LLM-based systems typically generate narratives and visualizations in staged pipelines, following either a text-first-graph-second or a graph-first-text-second paradigm. These designs often lead to chart-text inconsistency and insight freezing, where the intermediate evidence space becomes fixed and the model can no longer retrieve or construct new visual evidence as the narrative evolves, resulting in shallow and predefined analysis. To address the limitations, we propose \textbf{EvidFuse}, a training-free multi-agent framework that enables writing-time text-chart interleaved generation for data-driven reports. EvidFuse decouples visualization analysis from long-form drafting via two collaborating components: a \textbf{Data-Augmented Analysis Agent}, equipped with Exploratory Data Analysis (EDA)-derived knowledge and access to raw tables, and a \textbf{Real-Time Evidence Construction Writer} that plans an outline and drafts the report while intermittently issuing fine-grained analysis requests. This design allows visual evidence to be constructed and incorporated exactly when the narrative requires it, directly constraining subsequent claims and enabling on-demand expansion of the evidence space. Experiments demonstrate that EvidFuse attains the top rank in both LLM-as-a-judge and human evaluations on chart quality, chart-text alignment, and report-level usefulness.

Figures (14)

• Figure 1: The illustration of different generation paradigms for text-chart interleaved report.
• Figure 2: The Illustration of EvidFuse. Given a user analysis request and multiple data tables, EvidFuse first specializes a Data-Augmented Analysis Agent by EDA-derived dataset overview and raw tables. A Real-Time Evidence Construction Writer then plans an outline and generates the report. During generation, the writer suspends by specific visualization requests and continues when it receives visualization and request-based captions from the analysis agent as context for subsequent generation until <EOS> terminates the generation.
• Figure 3: Prompt for evaluating report chart layout.
• Figure 4: Prompt for evaluating chart readability.
• Figure 5: Prompt for evaluating text-chart consistency.