SlidesGen-Bench: Evaluating Slides Generation via Computational and Quantitative Metrics

TL;DR

SlidesGen-Bench introduces a unified, visual-domain benchmark for automated slide generation, addressing the lack of cross-architecture evaluation by assessing Content, Aesthetics, and Editability. It pairwise grounds outputs as renderings, employs QuizBank-based content fidelity tests, and uses four computational aesthetics metrics (Harmony, Engagement, Usability, Visual Rhythm) along with the PEI taxonomy for editability, all validated against a human-aligned Slides-Align1.5k dataset across nine systems and seven scenarios. The framework demonstrates superior alignment with human judgments compared to prior pipelines and reveals a notable gap in structure-aware, editable slide synthesis, positioning SlidesGen-Bench as a standard for future research and development. The work provides open-source code and data, enabling reproducible, robust comparisons and accelerating progress toward practical, editable, and aesthetically coherent presentation synthesis.

Abstract

The rapid evolution of Large Language Models (LLMs) has fostered diverse paradigms for automated slide generation, ranging from code-driven layouts to image-centric synthesis. However, evaluating these heterogeneous systems remains challenging, as existing protocols often struggle to provide comparable scores across architectures or rely on uncalibrated judgments. In this paper, we introduce SlidesGen-Bench, a benchmark designed to evaluate slide generation through a lens of three core principles: universality, quantification, and reliability. First, to establish a unified evaluation framework, we ground our analysis in the visual domain, treating terminal outputs as renderings to remain agnostic to the underlying generation method. Second, we propose a computational approach that quantitatively assesses slides across three distinct dimensions - Content, Aesthetics, and Editability - offering reproducible metrics where prior works relied on subjective or reference-dependent proxies. Finally, to ensure high correlation with human preference, we construct the Slides-Align1.5k dataset, a human preference aligned dataset covering slides from nine mainstream generation systems across seven scenarios. Our experiments demonstrate that SlidesGen-Bench achieves a higher degree of alignment with human judgment than existing evaluation pipelines. Our code and data are available at https://github.com/YunqiaoYang/SlidesGen-Bench.

Figures (34)

• Figure 1: The main pipeline of SlidesGen-Bench.
• Figure 2: Statistical distribution of content metrics.
• Figure 3: The QuizBank construction pipeline.
• Figure 4: The V-E (Visual-Editability) Matrix.
• Figure 5: The topic distribution of 30k human-generated slides, highlighting prevalent categories such as Education and Technology.