Personalization as Inverse Planning: Learning Latent Design Intents for Agentic Slide Generation via Structural Denoising

2026-07-01

AI Summary

Existing approaches struggle to achieve fine-grained, page-level slide personalization due to their inability to effectively capture users’ latent design intent. This work formulates the task as an inverse planning problem and introduces structural denoising as a verifiable proxy objective—a first in this domain. To optimize executable design proposals without relying on tool-specific knowledge, the authors propose SPIRE, a multi-agent reinforcement learning framework in which agents collaboratively refine designs. The approach substantially reduces policy gradient variance and significantly outperforms current state-of-the-art methods in experiments, empirically validating the alignment between structural denoising and personalization objectives.

Abstract

Slide design requires personalizing both deck themes and page layouts. Yet, current AI agent-based methods struggle with fine-grained, page-level design. Solely relying on prespecified templates or user verbose instructions, they fail to capture latent design intents, leaving Page-level Slide Personalization (PSP) unresolved. To close this gap, this work formulates PSP as an inverse planning problem. We propose to learn a design intent without assuming any knowledge of the specific executing tools (e.g., PowerPoint, Beamer) being used. However, relinquishing control over these tools makes the problem intractable to optimize end-to-end. To overcome this, we propose SPIRE, a principled framework to solve PSP approximately. By intentionally corrupting the visual structures of clean slides, SPIRE creates a verifiable task to denoise the corruption, whereby two agents learn to collaboratively refine executable designs via reinforcement learning (RL). We present a proof that structural denoising is a consistent surrogate for PSP, and that the multi-agent formulation strictly reduces policy gradient variance in RL. Extensive experiments demonstrate the superiority of SPIRE.

Problem

• Page-level Slide Personalization
• inverse planning
• latent design intents
• structural denoising
• AI agents

Innovation

• inverse planning
• structural denoising
• slide personalization
• multi-agent reinforcement learning
• latent design intent