Nikita Romanov

Identifying the interactions and dynamics of open quantum systems is essential for characterizing quantum hardware, designing robust simulation protocols, and developing tailored error-correction approaches. Motivated by this, we study the task of learning an unknown Lindbladian generator — a superoperator that fully specifies both coherent (Hamiltonian) and dissipative dynamics. Prior protocols assume known interaction structure, which can be restrictive when the relevant error mechanisms or control imperfections are not known in advance. In this paper, we present the first efficient protocol for learning sparse Lindbladians without any structural or locality assumptions. Our protocol is ancilla-free and uses only product-state preparations and Pauli-basis measurements, making it compatible with near-term experimental capabilities. Moreover, it achieves a nearly optimal time resolution in the regime where the Lindbladian contains at most polynomially many terms. Together, this provides a systematic route to scalable characterization of open-system quantum dynamics, especially when the error mechanisms of interest are not known in advance.