InfoVis: Unified Slide Image Generation and Editing
InfoVis, a 9B slide model, nearly matches Nano-Banana-2 on a 1,527-instance benchmark.
Researchers introduce InfoVis, a 9B model for unified presentation-slide generation and editing, trained with more than 13 million continual-pretraining records and 1.9 million supervised fine-tuning records. VisSlideBench contains 1,527 screened instances covering four tasks, and its automatic rankings match blinded human rankings. InfoVis leads every evaluated open model on all eleven metrics, scoring 7.94 overall against 7.99 for Nano-Banana-2. Starting from FLUX.2-klein-base-9B, slide-specific post-training adds 1.6 points, and a document-to-slide workflow ranks second in aesthetics on SlidesGen-Bench.
- InfoVis is a unified 9B slide generation and editing model.
- Training used over 13M pretraining and 1.9M fine-tuning records.
- VisSlideBench has 1,527 instances across four tasks.
- InfoVis scores 7.94 overall versus 7.99 for Nano-Banana-2.
- Slide post-training adds 1.6 points over FLUX.2-klein-base-9B.
- Products
- VisSlideBenchSlidesGen-Bench
From the original article
1 min read at huggingface.co
Generating and editing presentation slides requires models to combine accurate text rendering, instruction following, visual grounding, aesthetic design, and precise editing within one canvas. Existing work covers only subsets of these capabilities and lacks large-scale slide-native supervision and diagnostic evaluation.
By Jiahao Wang, Xiyan Jiang, Chenyuan Wu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.