Narrative-Driven Paper-to-Slide Generation via ArcDeck
Summary: arXiv:2604.11969v1 Announce Type: new
Abstract
We introduce ArcDeck, a multi-agent framework that formulates paper-to-slide generation as a structured narrative reconstruction task. Unlike existing methods that directly summarize raw text into slides, ArcDeck explicitly models the source paper’s logical flow. It first parses the input to construct a discourse tree and establish a global commitment document, ensuring the high-level intent is preserved. These structural priors then guide an iterative multi-agent refinement process, where specialized agents iteratively critique and revise the presentation outline before rendering the final visual layouts and designs. To evaluate our approach, we also introduce ArcBench, a newly curated benchmark of academic paper-slide pairs. Experimental results demonstrate that explicit discourse modeling, combined with role-specific agent coordination, significantly improves the narrative flow and logical coherence of the generated presentations.
Introduction
The transition from academic papers to presentation slides is often a challenging task. Many researchers struggle to distill complex ideas into a cohesive and engaging format. Traditional summarization techniques frequently fall short of preserving the logical flow of the original text, leading to presentations that lack clarity and coherence. ArcDeck addresses this challenge by employing a multi-agent framework that ensures a structured approach to narrative reconstruction.
Key Features of ArcDeck
- Discourse Tree Construction: ArcDeck begins by analyzing the source paper to create a discourse tree. This visual representation captures the relationships between various sections of the paper, providing a clear roadmap for the presentation.
- Global Commitment Document: The framework establishes a global commitment document that articulates the high-level intent of the paper. This foundational element serves as a guide throughout the generation process, ensuring that the core message remains intact.
- Iterative Refinement Process: Specialized agents engage in an iterative critique and revision process. Each agent focuses on specific aspects of the presentation, such as content accuracy, visual design, and audience engagement, leading to a more polished final product.
- ArcBench Benchmark: To assess the effectiveness of ArcDeck, the authors introduce ArcBench, a curated benchmark of academic paper-slide pairs. This resource provides a basis for evaluating the quality of generated presentations against established standards.
Experimental Results
In their experiments, the authors found that presentations generated using ArcDeck exhibited a marked improvement in both narrative flow and logical coherence compared to those produced by existing methods. The explicit modeling of discourse, along with the collaborative efforts of role-specific agents, allowed for a more nuanced understanding of the source material, ultimately leading to presentations that better resonated with audiences.
Conclusion
ArcDeck represents a significant advancement in the field of automated presentation generation. By focusing on the narrative structure and employing a multi-agent approach, it not only enhances the quality of generated slides but also supports researchers in effectively communicating their ideas. The introduction of ArcBench further establishes a foundation for ongoing research and development in this area, paving the way for future innovations in academic communication.