BuseyBench is a public benchmark gallery specifically testing the ability of AI models to draw Gary Busey's face. Its core task is to have large language models (LLMs) attempt to generate a valid and standalone SVG portrait of Gary Busey's face. The website displays 48 results out of 310 published runs, and users can open any portrait to view the evidence, or add up to four cards to a comparison tray for comparison.
Main capabilities:
- Evaluate the ability of language models to generate portraits in SVG format.
- Provide a public gallery displaying published run results.
- Support viewing detailed evidence for a single portrait.
- Support adding up to four cards to a comparison tray for side-by-side comparison.
How to use: Visit the website to browse the portrait works in the gallery. Click any portrait to view its evidence details. By selecting cards, up to four can be added to the comparison tray for intuitive comparison of outputs from different models.
Use cases:
- Researchers or developers compare the performance of different language models on a specific image generation task.
- Users interested in AI-generated art or SVG output browse examples.
- Serve as one of the reference benchmarks for model capability evaluation.
Limitations:
- The task scope is very specific, targeting only SVG generation of Gary Busey's face.
- The website displays published run results, not real-time generation.
- Specific information regarding fees, pricing, platform support, etc. is not yet provided.
