Search papers, labs, and topics across Lattice.
1
0
2
Generative models systematically underestimate data diversity, losing up to X% of entropy, because training on finite datasets inherently biases them towards lower diversity estimates.
