Out-of-domain generalization from rephrasability and stability
Produced by Generalization Bounds for Autoregressive Processes and In-Context Learning
Autoregressive process
A sequential stochastic model whose conditional distribution at each step is parameterized as a function of preceding observations.
A sequential stochastic model whose conditional distribution at each step is parameterized as a function of preceding observations.
The dependence on prior observations gives an ordered factorization that is useful for both probabilistic analysis and sequential generation.