Benchmark RadarAI BENCHMARK PROFILE
PPT-Eval
PPT-Eval is a benchmark of 120 PowerPoint tasks across 12 files for computer-use agents. It covers content creation and editing, with rubric-based evaluation that awards partial credit and provides natural language feedback.
- Released
- 2026-06-30
- Readiness
- Inspectable
- Primary field
- General AI
Why it matters
Provides a realistic, multimodal testbed for computer-use agents. The rubric-based scoring captures partial progress and correlates with human judgment, enabling nuanced comparison.
Motivation
Creating and editing slides is a rich, multimodal activity that is ubiquitous in professional and educational settings, making it an ideal testbed for real-world computer-use agents.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.