PPTArena: A Benchmark for Agentic PowerPoint Editing

Abstract

We introduce PPTArena, a benchmark for PowerPoint editing that measures reliable modifications to real slides under natural-language instructions. In contrast to image-PDF renderings or text-to-slide generation, PPTArena focuses on in-place editing across 100 decks, 2125 slides, and over 800 targeted edits covering text, charts, tables, animations, and master-level styles. Each case includes a ground-truth deck, a fully specified target outcome, and a dual VLM-as-judge pipeline that separately scores instruction following and visual quality using both structural diffs and slide images.

Building on this setting, we propose PPTPilot, a structure-aware slide-editing agent that plans semantic edit sequences, routes between high-level programmatic tools and deterministic XML operations for precise control, and verifies outputs through an iterative plan-edit-check loop against task-specific constraints. In our experiments, PPTPilot outperforms strong proprietary agents and frontier VLM systems by over 10 percentage points on compound, layout-sensitive, and cross-slide edits, with particularly large gains in visual fidelity and deck-wide consistency.

Features

• Agentic Editing: Edit PowerPoint presentations using natural language instructions.
• Dual-View Evaluation: Compare "Original" vs "Ground Truth" slides side-by-side.
• Iterative Refinement: The agent plans, executes, and verifies edits in a loop.
• Multi-Modal Judge: Automated evaluation using VLM-as-a-judge for both instruction following and visual quality.
• Comprehensive Benchmark: Covers diverse tasks including text editing, chart manipulation, layout adjustments, and image handling.

WebApp: https://ppt-arena.onrender.com/evaluation

Directory Structure

• src/: Core application code (Flask app, LLM handlers, PPT processing) and configuration files (requirements.txt, evaluation_pairs_refined.json).
• Original/: Benchmark dataset - Original PowerPoint files.
• GroundTruth/: Benchmark dataset - Ground Truth PowerPoint files.
• paper/: Contains the paper LaTeX source and figures.

Installation

1. Clone the repository:

   git clone https://github.com/michaelofengenden/PPTArena.git
   cd PPTArena

2. Install dependencies:

   pip install -r src/requirements.txt

3. Configure API Keys: Create a credentials.env file in the root directory:

   OPENAI_API_KEY=your_openai_key
   GEMINI_API_KEY=your_gemini_key

Usage

1. Start the Application:

   python src/app.py

2. Access the Web Interface: Open http://localhost:5000 in your browser.

3. Evaluate:

   • Go to the Evaluation tab.
   • Select a test case to see the Original and Ground Truth.
   • Click "Generate prediction" to run the agent on the task.
   • Use "Call LLM Judge" to score the result.

4. Chat & Edit:

   • Go to the Chat tab.
   • Upload any .pptx file.
   • Type instructions like "Change the title font to Arial" or "Add a bar chart with this data...".

Citation

If you find this work useful, please cite our paper:

@article{ofengenden2025pptarena,
  title={PPTArena: A Benchmark for Agentic PowerPoint Editing},
  author={Ofengenden, Michael and Man, Yunze and Pang, Ziqi and Wang, Yu-Xiong},
  journal={arXiv preprint arXiv:2512.03042},
  year={2025}
}