SeaSlides: Semantic Abstraction Layer for Agentic Slide Generation

2026-08-04

AI Summary

Existing intelligent slide generation systems suffer from inadequate fidelity, visual consistency, and reliable rendering of technical content such as mathematical formulas, code snippets, and charts. To address these limitations, this work proposes SeaSlides, a novel framework that introduces a semantic abstraction layer to decouple content creation from layout rendering. In SeaSlides, structured components and capability modules generate semantic content, while templates handle styling and rendering—supporting both HTML and Typst backends. The capability modules route specialized content to dedicated renderers and integrate a three-stage feedback mechanism to identify build errors, constraint violations, and visual defects. Experimental results demonstrate that SeaSlides significantly improves rich-content generation quality on both the UltraPresent benchmark and the newly introduced SeaSlidesBench-Rich, achieving the best macro-average scores across three out of four evaluated models while maintaining strong overall performance.

Abstract

Agentic presentation generation must preserve source content, maintain coherent visual design, render specialized objects, and produce usable artifacts. Existing systems meet only part of this requirement: templates preserve regularity but restrict adaptation, whereas free-form HTML or SVG gives models flexibility at the cost of low-level rendering decisions. This mismatch makes long technical decks brittle, especially when slides contain formulas, code, or data graphics. We present SeaSlides, an agentic slide-generation framework built around a semantic abstraction layer. Rather than authoring coordinates, inline styles, or raw SVG geometry, the model writes structured slide content through reusable components and capability modules, while templates own layout, style, and rendering. We instantiate this principle separately in HTML and Typst: SeaSlides-HTML uses template-defined DOM components, whereas SeaSlides-Typst uses template functions and package-backed modules. Capability modules route equations, code, and charts to dedicated renderers, and three feedback stages localize build errors, project-constraint violations, and visual defects before export. The two systems retain backend-specific syntax and contracts while sharing the same authoring boundary. For evaluation, we combine the 128-task UltraPresent validation setting with SeaSlidesBench-Rich, a new 32-task benchmark stressing mathematics, code, pseudocode, tables, charts, and diagrams. Across four generation models, both SeaSlides backends produce more readable, content-oriented source than SVG-heavy generation. A SeaSlides backend attains the highest rich-content macro-average under three of the four models while maintaining competitive overall qualitative performance. These results support semantic abstraction as a practical authoring principle across presentation backends.

Problem

• agentic slide generation
• semantic abstraction
• technical content rendering
• presentation generation
• visual design coherence

Innovation

• semantic abstraction
• agentic slide generation
• capability modules
• template-component decoupling
• multi-backend presentation