Pinned
Are neural nets across modalities really converging to the same representation as they scale, as the Platonic Representation Hypothesis suggests?
We show that common representational similarity metrics are confounded by network width & depth. We propose a permutation-based
00:00
