Report Gallery
Explore visual data reports generated by our agentic system. For each entry, the system processed datasets as provided—handling preparation, analysis, and visualization autonomously. The examples span a variety of domains and data types, illustrating a general, dataset-agnostic approach without manual intervention or hard‑coded configurations.
Before You Dive In
💡 Why Our Reports Are Intentionally Concise
The focused length of these reports reflects a conscious design philosophy rooted in our Human-AI Partnership model as outlined in our paper. Rather than attempting to exhaustively anticipate user needs—which often leads to extensive analysis based on potentially incorrect assumptions—our system produces a curated starting point that serves two complementary purposes.
For Analysts: The report provides a solid foundation for iterative refinement. Human intent is often ambiguous at the outset and evolves as insights emerge. A minor misalignment can render an agent's entire output useless, forcing costly regeneration that erodes user trust and agency. By delivering focused, high-quality insights with full traceability, analysts can efficiently build upon and refine the analysis rather than starting from scratch.
For Readers: Each report enables open-ended exploration beyond the AI's authored narrative. Using interactive data tables linked to visualizations, readers can dynamically filter and cross-reference data to answer ad-hoc questions that arise during their engagement with the content—questions the AI could never have anticipated.
This approach embodies our core belief: successful human-AI partnership stems not from demanding perfection from the AI, but from ensuring its outputs are both deeply auditable & adaptable for experts and flexibly explorable for end-users. Ultimately, we believe the steerability and auditability of the process that generates these reports is more valuable than any individual report instance—enabling not only systematic improvement and adaptation across diverse analytical contexts, but also the trustability that comes from transparent, traceable decision-making of the AI. The result is a more sustainable, collaborative workflow that prioritizes human agency alongside automation.
Challenge Dataset Reports
IEEE VIS Publication Analysis
Our challenge entry uses the VIS publications dataset, processed with the standard agentic pipeline—no special rules or tuning.
Challenge Submission
Challenge submission on the VIS publications dataset, processed with the default pipeline to explore research trends, collaboration patterns, and publication dynamics over three decades.
View Report →Case Study: A Flawed Insight
A transparent look at how the system surfaces and fixes analytical issues: execution traces and modular, rule‑based components make missteps easy to pinpoint and correct with targeted tool updates.
View Case Study →Agent Methodology Variants
Alternative outcomes from separate runs on the same data, without manual tuning or domain‑specific configuration.
Clear Insights
A run focused on selective exploration and clarity. Keeps visualizations simple to support interpretability and highlight core patterns.
View Report →Complex Patterns
A run emphasizing comprehensive multi‑dimensional analysis. Uses broader exploration to reveal nuanced correlations through information‑dense visualizations.
View Report →General Domain Datasets
We applied the system across a range of domains—from agricultural data to biological research, automotive analysis, and gemstone markets—all processed automatically from raw, uncurated datasets.
Barley
Analysis of barley crop yields across different varieties and locations, exploring agricultural patterns and productivity trends.
View Report →Cars
Comprehensive analysis of automotive data, examining vehicle specifications, performance metrics, and market characteristics.
View Report →Cars 93
Historical analysis of 93 car models from the 1993 US market, providing insights into automotive trends of the early 1990s.
View Report →Diamonds
Market analysis of diamond pricing and quality attributes, exploring the relationships between cut, clarity, color, and carat weight.
View Report →Driving
Transportation behavior analysis examining driving patterns, safety trends, and vehicular usage over time.
View Report →Iris
Classic machine learning dataset analysis, exploring the morphological characteristics of three iris flower species.
View Report →Mammal Sleep
Biological research into mammalian sleep patterns, examining evolutionary adaptations and physiological characteristics across species.
View Report →