I didn’t expect that one day I’ll cherish someone the way I cherish and love you
talking about what hurt me many times and not stepping back not even thinking of going away, I want all the tangled parts to be solved and fighting for being the best versions of ourselves
I was raised to believe that light always conquers darkness — but I grew up in a world where the darkness slowly swallowed the light inside me.
Still, somewhere, a spark remains.
In our new work, we ask a simple question:
Which design choices actually enable (or prevent) compositional generalization?
We study this in a controlled setting across visual modalities—cutting down the search space for anyone training or using these models.
Generalization is the goal. A core piece is compositional generalization: recombining known concepts into new combinations.
It’s central to human intelligence, but we still don’t know what drives/hinders it in generative models and today’s design choices are not driven by it.
Seriously, what is the goal of today’s visual generative models?
Are pretty videos/images and low FIDs enough – or should we also demand something closer to human-like creativity? Our paper tries to answer this question 🧵
If you just know your worth, I could hold the stars in my palm & it won't be enough to describe how much I love you, but sometimes it slips out of my grasp unintentionally my inner world is much messier and nuanced, sth forced me in that dark place but no worries i'm fighting.