Mastodon Feed
baldur@toot.cafe ("Baldur Bjarnason") wrote:
“[2603.21687] MIRAGE: The Illusion of Visual Understanding”
https://arxiv.org/abs/2603.21687
> Frontier models readily generate detailed image descriptions and elaborate reasoning traces, including pathology-biased clinical findings, for images never provided
And
> Second, without any image input, models also attain strikingly high scores across general and medical multimodal benchmarks, bringing into question their utility and design