Transferring IPP query‑to‑sample reductions to PAC‑verification

Investigate whether the query‑to‑sample reduction techniques for interactive proofs of proximity developed by Goldberg and Rothblum (2022) can be extended to PAC‑verification of agnostic learning tasks, where the verifier must output a hypothesis with guaranteed approximation error.

Background

Interactive proofs of proximity (IPPs) paper delegating sublinear‑time property testing with a prover, and recent work established query‑to‑sample reductions in that setting.

The paper questions whether analogous reductions apply in the PAC‑verification setting, which involves different goals (producing hypotheses with error guarantees) and may face different complexity trade‑offs.