Presenting a Paper is an Art: Self-Improvement Aesthetic Agents for Academic Presentations

2025-10-07

AI Summary

Existing automated academic presentation systems suffer from narrative discontinuity, suboptimal visual aesthetics, and absence of self-improvement capabilities. This paper introduces PresAesth—the first multi-task reinforcement learning framework for aesthetic modeling in academic presentations—capable of aesthetic scoring, defect correction, and iterative self-refinement driven by comparative feedback, even under data scarcity. We propose EvoPresent Benchmark, the first evaluation benchmark jointly quantifying content quality and aesthetic perception. Furthermore, we design an end-to-end slide generation pipeline integrating narrative generation, virtual presenter embodiment, and closed-loop aesthetic optimization. Experiments demonstrate that high-quality comparative feedback significantly enhances presentation quality, uncovering fundamental trade-offs between content fidelity and visual design. Multi-task RL exhibits superior generalizability in aesthetic modeling compared to single-task alternatives.

Technology Category

• Cognitive Modeling & Cognitive Systems: Computational Creativity
• Natural Language Processing: Generation
• Humans and AI: Learning Human Values and Preferences

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

The promotion of academic papers has become an important means of enhancing research visibility. However, existing automated methods struggle limited storytelling, insufficient aesthetic quality, and constrained self-adjustment, making it difficult to achieve efficient and engaging dissemination. At the heart of those challenges is a simple principle: emph{there is no way to improve it when you cannot evaluate it right}. To address this, we introduce extbf{EvoPresent}, a self-improvement agent framework that unifies coherent narratives, aesthetic-aware designs, and realistic presentation delivery via virtual characters. Central to EvoPresent is extbf{PresAesth}, a multi-task reinforcement learning (RL) aesthetic model that provides reliable aesthetic scoring, defect adjustment, and comparative feedback, enabling iterative self-improvement even under limited aesthetic training data. To systematically evaluate the methods, we introduce extbf{EvoPresent Benchmark}, a comprehensive benchmark comprising: extit{Presentation Generation Quality}, built on 650 top-tier AI conference papers with multimodal resources (slides, videos and scripts) to assess both content and design; and extit{Aesthetic Awareness}, consisting of 2,000 slide pairs with varying aesthetic levels, supporting joint training and evaluation on scoring, defect adjustment, and comparison. Our findings highlight that (i) High-quality feedback is essential for agent self-improvement, while initial capability alone does not guarantee effective self-correction. (ii) Automated generation pipelines exhibit a trade-off between visual design and content construction. (iii) Multi-task RL training shows stronger generalization in aesthetic awareness tasks.

Problem

• Automated academic presentation methods lack storytelling and aesthetic quality
• Existing systems cannot self-adjust without reliable aesthetic evaluation metrics
• Current approaches struggle with visual design and content construction balance

Innovation

• Self-improvement agent framework for academic presentations
• Multi-task reinforcement learning aesthetic model PresAesth
• Comprehensive benchmark with multimodal resources for evaluation