Search papers, labs, and topics across Lattice.
1
0
2
Ditch blurry FID scores: evaluating generative models directly in discrete token space reveals perceptual quality nuances that continuous feature metrics miss.
