Explainable Iterative Data Visualisation Refinement via an LLM Agent

2026-04-20 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, extended

Summary

A novel agentic AI pipeline automates and explains the iterative refinement of high-dimensional data visualizations, addressing challenges in hyperparameter tuning for dimensionality reduction (DR) algorithms like t-SNE, UMAP, and PaCMAP. The system integrates quantitative metrics (e.g., Trustworthiness, Stress-1, Silhouette score), hierarchical dendrograms, 2D embedding coordinates, and visualization plots as multi-modal input for a Large Language Model (LLM) agent. This LLM acts as an expert diagnostician, generating structured JSON reports with quality scores, structural/visual inspections, outlier forensics, and actionable hyperparameter recommendations. Experiments on complex single-cell RNA sequencing datasets, including Healthy Human Kidney and Mature Human Kidney, demonstrate that frontier LLMs (GPT-5.2, Gemini-3-Pro-Preview, Claude-Opus-4.5) can rapidly navigate hyperparameter spaces, achieving high-quality convergence typically within a few iterations. The framework balances local neighborhood preservation with global geometric fidelity, producing biologically faithful and interpretable plots with clear rationales for each refinement step.

Key takeaway

For research scientists working with high-dimensional data visualization, this agentic AI pipeline offers a significant advancement in automating and explaining the complex process of dimensionality reduction hyperparameter tuning. You should consider integrating such LLM-guided frameworks to accelerate pattern discovery and ensure biologically faithful, interpretable plots, reducing the manual effort and subjectivity traditionally associated with visualization refinement. This approach allows you to focus on scientific interpretation rather than tedious parameter optimization.

Key insights

An LLM-driven agent automates and explains data visualization refinement by iteratively optimizing DR hyperparameters.

Principles

• Combine quantitative metrics with qualitative visual assessment.
• LLMs can bridge rigorous assessment and human insight.
• Iterative refinement improves visualization quality.

Method

An LLM agent iteratively refines DR hyperparameters by evaluating multi-modal input (metrics, dendrograms, embeddings, plots) and generating structured JSON recommendations for subsequent iterations until visualization quality stabilizes.

In practice

• Use t-SNE, UMAP, or PaCMAP for high-dimensional data visualization.
• Employ LLMs for automated hyperparameter tuning.
• Track Trustworthiness, Stress-1, and Silhouette scores.

Topics

• LLM Agent
• Data Visualization Refinement
• Dimensionality Reduction
• Hyperparameter Optimization
• Explainable AI

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Data Scientist

Related on AIssential

• Winning a Kaggle Competition with Generative AI–Assisted Coding — LLM agents, paired with GPU acceleration, dramatically speed up ML experimentation by automating code generation and execution.
• Accelerating pharmaceutical research with a multi-agent AI system — Multi-agent AI systems can transform fragmented data into actionable insights, significantly accelerating complex research processes.
• [P] Visualizing token-level activity in a transformer — A developer is experimenting with a 3D visualization tool designed to make Large Language Model (LLM) inference more intuitive by animating activation paths across a network of components like attenti…
• Agentic Authoring of Interactive Multiview Visualizations in Genomics — Agentic LLM iteration significantly enhances genomics visualization quality, but complex agent architectures offer no further benefit.
• Probe-and-Refine Tuning of Repository Guidance for Coding Agents — Iterative, synthetic bug-fix probing significantly refines LLM coding agent guidance, boosting resolve rates by improving coverage.
• Can Claude Code Learn To Draw In MS PAINT? — Automated iterative refinement against test criteria enables AI agents to achieve complex tasks like drawing.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.