Learnijoy
ResearchAI Engineering & DevToolsAI in Marketing

ACE: Self-Correcting Agent for Presentation Automation.

JooYoung Jang, Taegyeong Lee, Jihyeon Park, Nojun Kwak· August 26, 2026 View original

Key takeaways

  • ACE automates multi-slide presentation editing using a hierarchical scene-graph.
  • Content-aware routing and self-correction significantly improve agent performance.
  • The system achieves higher instruction following, speed, and cost efficiency.
  • Human raters strongly prefer ACE's self-corrected design outputs.

Who benefits

MarketingSalesMedia & PublishingDesignEdTech

Summary

This paper introduces ACE, a self-correcting agentic canvas editor for multi-slide presentation automation, which uses a hierarchical scene-graph and a content-aware router. ACE significantly improves instruction following, speed, and cost efficiency compared to previous methods, with human raters preferring its self-corrected outputs.

Automating document editing in commercial design platforms using Large Language Model (LLM) agents faces two primary challenges: handling flat, absolutely positioned legacy document formats that lead to layout breakage, and the lack of a unique "ground truth" for design, making traditional evaluation difficult. This research presents "ACE" (Agentic Canvas Editor), a solution designed for multi-slide presentation automation. ACE operates on a hierarchical scene-graph with a specialized action space of 98 tools, enabling more robust layout management. ACE is paired with "CARE" (Content-Aware Router), which reduces input tokens by feeding the agent only relevant document slices. A key innovation is its self-correction loop, driven by a "ground-truth-free" instruction-following (IF) judge that provides natural-language critique back to the agent. This self-correction mechanism significantly boosts ACE's instruction following capabilities, achieving higher scores at 1.75x speed and 44% lower cost than an iterative HTML pipeline. Blind human raters overwhelmingly preferred ACE's overall output and its self-corrected versions, validating the effectiveness of the approach.

Why it matters

Professionals in marketing, sales, and content creation can leverage ACE to automate presentation design, significantly reducing manual effort, improving consistency, and accelerating content production with higher quality outputs.

How to implement this in your domain

  1. 1Explore agentic design tools that utilize hierarchical scene-graphs for document editing.
  2. 2Implement content-aware routing mechanisms to optimize LLM input for design tasks.
  3. 3Develop self-correction loops for AI agents using natural-language feedback from judges.
  4. 4Evaluate agent performance using human preference ratings when objective ground truth is absent.
  5. 5Consider integrating such automation into marketing and sales content creation workflows.

Original post by JooYoung Jang, Taegyeong Lee, Jihyeon Park, Nojun Kwak

"arXiv:2608.24103v1 Announce Type: new Abstract: Commercial design platforms increasingly edit documents through large language model (LLM) agents, but two practical problems block reliable deployment: legacy document formats expose only \emph{flat}, absolutely positioned elements…"

View on X

Originally posted by JooYoung Jang, Taegyeong Lee, Jihyeon Park, Nojun Kwak on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Engineering & DevTools

AI Engineering & DevTools

Zapier vs. Tray: Enterprise Automation Platform Comparison for 2026

This post compares Zapier and Tray.io, evaluating which platform is better suited for enterprise automation needs by balancing power and ease of use. It argues that the best tools scale for complex requirements while remaining intuitive for all users.

Ryan KaneJul 28, 2027
AI ResearchAI Engineering & DevToolsAI Investing

FraudBench Benchmarks Adversarial Robustness in Financial Risk Assessment

This paper introduces FraudBench, a protocol-sensitive benchmark for evaluating the adversarial robustness of machine learning models in financial fraud and credit-risk detection. It demonstrates that robustness conclusions are highly dependent on how domain-specific constraints and attacker capabilities are incorporated into the evaluation protocol.

Xitong Zeng, Zhaoge Bi, Yitian Yang, Huaming Chen, Quan Z. ShengAug 26, 2026
AI ResearchAI Engineering & DevTools

Persistent Cross Entropy Extends Topological Data Analysis

This paper introduces Persistent Cross Entropy (PCE), a novel extension of cross-entropy to persistence diagrams, which are used in topological data analysis. PCE bridges different event spaces of diagrams using an induced probability, enabling new applications like distinguishing diagrams with similar persistent entropy and separating causal directions in dynamical systems.

Sijin Yeom, Jae-Hun JungAug 26, 2026