Search papers, labs, and topics across Lattice.
1
0
2
Transformers can provably learn Bayesian posterior predictive distributions in-context, revealing the architectural inductive biases that enable this capability.
