- Glif tests
- Edit from markup (garden)
Edit from markup (garden)
How 20 models handled “Edit from markup (garden)” from our eval suite, GPT Image 2 first, then newest first.
What we look for
Edits only the marked up area. Removes markup. Knows plant names and appearances.
Prompt
“Change the plant in the marked up area to Nepeta faassenii. Then remove the markup.”
Tests for
Image editing, world knowledge