Search papers, labs, and topics across Lattice.
This paper characterizes the conditions under which inexact score oracle access allows for sampling with negligible total variation bias for a standard target family. The authors demonstrate that any error weaker than the sub-Gaussian assumption precludes the possibility of unbiased sampling, thereby extending the implications of previous work to a broader set of error scenarios. This finding is significant as it provides a clearer understanding of the limitations of sampling methods under inexact score conditions, making it algorithm-agnostic.
Weakening error assumptions reveals that unbiased sampling may be impossible, reshaping our understanding of tractable sampling methods.
We provide a simple and tight characterization of the types of inexact score oracle access that permit sampling with vanishing total variation bias, for a standard, well-behaved target family. Our main result shows that any weaker error than the sub-Gaussian assumption used by [YW26] rules out the tractability of unbiased sampling. This strengthens the conclusion of [CCSW26] to be algorithm-agnostic, and to hold for a wider range of error assumptions.
