Charts as Epistemic Actions for Experience-Driven Agentic Data Analysis
Xinzhe Wang, Xudong Weng, Zang Li (Tencent) and Shaolei Zhang (Renmin University) test whether data agents can use charts they draw as reasoning steps, and propose a training-free method that distills experience about when to plot.
Ask this paper
Probe suite. 200 controlled tasks under four matched interface conditions show that agents score as well from self-generated charts as from numeric output, yet having plotting access does not make them use it consistently.
Where charts help. Charts lead on temporal, pattern and anomaly questions; text leads on precise reading and conditional verification.
Method. Two families of experience are distilled from conditioned rollouts, one about selecting and verifying analytical operations and one about when to use text, charts or both. A trajectory judge labels steps and a validation gate admits only experience that improves held-out performance.
Result. On three held-out benchmarks with four backbones, accuracy and visual evidence quality rise over matched baselines, and backbones that never drew charts start using them, with no parameter updates.
Abstract
Data agents built on ReAct-style loops now execute complex analytical workflows over structured data, yet their intermediate reasoning still runs almost entirely through text. Human analysts instead move between numbers, tables, and charts, reading visual evidence to expose trends and anomalies that text summaries drop. We ask whether charts can serve as internal epistemic actions inside a data agent's reasoning loop, and how to teach an agent to use them. We first build \probe{}, a suite of 200 controlled tasks under four matched interface conditions, and find a persistent gap between what agents can read and what they choose to use. Agents score as well from self-generated charts as from numeric output, yet plotting access alone does not translate into consistent use. Task slices show that charts lead on temporal, pattern, and anomaly reasoning, while text leads on precise reading and conditional verification. We then propose \method{}, which distils two families of experience from conditioned trajectory rollouts. One family records how to select and verify analytical operations, and the other records when to invoke text, charts, or both. A trajectory judge scores outcomes and labels intermediate steps, and a validation gate admits only experience that improves held-out performance. On three held-out benchmarks with four LLM backbones, \method{} raises analytical accuracy and visual evidence quality over matched baselines, and backbones that never drew charts begin using them, all without touching model parameters.