//
sign in
Post
by @danabra.mov
PostEmbed
by @danabra.mov
Record
by @jimpick.com
Record
by @atsui.org
+ new component
Post
āœ… This shows that, starting from the mid-layers, a single token effectively summarizes all 1024 image tokens! āŒ This does not occur in models fine-tuned for visual understanding (such as Pixtral).