VisJudge-Bench: Aesthetics and Quality Assessment of Visualizations

TL;DR

VisJudge-Bench introduces a comprehensive benchmark for evaluating multimodal LLMs on visualization aesthetics and quality using Fidelity, Expressiveness, and Aesthetics. It compiles 3,090 expert-annotated samples across single, multi-view, and dashboard visualizations spanning 32 chart types, and employs an adaptive question-generation framework with rigorous expert annotation. The authors also present VisJudge, a domain-tuned evaluator based on Qwen2.5-VL-7B-Instruct with LoRA and reinforcement learning, which substantially narrows the gap to human judgment. Results show that leading baselines lag behind humans in visualization evaluation, while VisJudge achieves lower MAE and higher correlation, demonstrating the effectiveness of domain-specific training for visualization assessment. The VisJudge-Bench dataset and VisJudge are released to the community to foster improved automated visualization evaluation and future research.

Abstract

Visualization, a domain-specific yet widely used form of imagery, is an effective way to turn complex datasets into intuitive insights, and its value depends on whether data are faithfully represented, clearly communicated, and aesthetically designed. However, evaluating visualization quality is challenging: unlike natural images, it requires simultaneous judgment across data encoding accuracy, information expressiveness, and visual aesthetics. Although multimodal large language models (MLLMs) have shown promising performance in aesthetic assessment of natural images, no systematic benchmark exists for measuring their capabilities in evaluating visualizations. To address this, we propose VisJudge-Bench, the first comprehensive benchmark for evaluating MLLMs' performance in assessing visualization aesthetics and quality. It contains 3,090 expert-annotated samples from real-world scenarios, covering single visualizations, multiple visualizations, and dashboards across 32 chart types. Systematic testing on this benchmark reveals that even the most advanced MLLMs (such as GPT-5) still exhibit significant gaps compared to human experts in judgment, with a Mean Absolute Error (MAE) of 0.551 and a correlation with human ratings of only 0.429. To address this issue, we propose VisJudge, a model specifically designed for visualization aesthetics and quality assessment. Experimental results demonstrate that VisJudge significantly narrows the gap with human judgment, reducing the MAE to 0.442 (a 19.8% reduction) and increasing the consistency with human experts to 0.681 (a 58.7% improvement) compared to GPT-5. The benchmark is available at https://github.com/HKUSTDial/VisJudgeBench.

Figures (26)

• Figure 1: The "Fidelity, Expressiveness, and Aesthetics" evaluation framework.
• Figure 2: From natural images to visualization: the need for specialized visualization assessment. Green and red denote positive and negative assessments, respectively, highlighting the contrast between MLLMs' capabilities in general aesthetics versus visualization-specific evaluation.
• Figure 3: VisJudge-Bench construction framework.
• Figure 4: Distribution and bias analysis of MLLM scores. Score distribution density curves showing the rating patterns of different models compared to human experts on the 1-5 scale.
• Figure 5: Model–Human rating correlation across visualization types.