Skip to content
Better HN
Top
New
Best
Ask
Show
Jobs
Search
⌘K
DatBench fixes VLM evals: 70% blindly solvable, 42% mislabeled, 35% prod gap | Better HN
story
DatBench fixes VLM evals: 70% blindly solvable, 42% mislabeled, 35% prod gap
(opens in new tab)
datologyai.com
5 points
hurrycane
4mo ago
0 comments
Share
0 comments
No comments yet.
