Classical Shadows State Tomography

Short answer: Classical shadow tomography predicts many properties of a quantum state from few randomized measurements, without reconstructing the full state. Two standard protocols sit at opposite ends: single-qubit Pauli measurements, which suit local operators, and global Clifford measurements, which are efficient for low-rank operators but need heavy gate overhead. Recent papers from 2023 to 2025 report middle-ground and refined versions that cut the number of samples needed. They are theoretical and numerical results.

The two standard protocols

A 2023 Quantum paper calls classical shadow tomography a powerful randomized measurement protocol for predicting many properties of a state with few measurements. It describes the two protocols studied so far. Local Pauli measurement is well suited to local operators but inefficient for large operators. Global Clifford measurement is efficient for low-rank operators but infeasible on near-term devices because of its gate overhead. A 2024 Physical Review Research paper adds that a shadow is an efficiently representable, compressed classical description built from random measurements on copies of a state, followed by classical processing. Shadows do not give a complete description of the state, but contain enough to estimate a wide range of properties.

Shallow circuits in between

The 2023 paper demonstrates a scalable approach for generic randomized measurements implemented with finite-depth local Clifford random circuits, interpolating between the Pauli and Clifford limits. It combines the locally scrambled classical shadow framework with tensor network techniques to compute the reconstruction map and evaluate properties at scale. The authors say it allows shadow tomography on shallow circuits with superior sample efficiency and minimal gate overhead, suits NISQ devices, and gives an exponential advantage over Pauli measurement for quasi-local operators and a more efficient fidelity estimation.

Better post-processing

The 2024 paper uses an underused resource in the inversion map of classical shadows to refine the estimation cost of target observables without extra overhead. For local measurement strategies they show optimization leading to an exponential separation in sample complexity compared with the standard approach, and in an exceptional case optimized post-processing for local measurements achieves the same efficiency as global Clifford shadows.

Real-valued variants

A 2025 paper in Journal of Physics A studies protocols using local or global orthogonal Clifford gates, with real-valued vectors. For estimating expectation values of arbitrary real-valued observables, it finds global orthogonal Cliffords typically halve the required samples. For k-local observables made of real-valued Pauli operators, local orthogonal Cliffords give a reduction by an exponential-in-k factor over local unitary Cliffords. By measuring in a basis with complex-valued vectors, orthogonal shadows can in the large-system limit reproduce the original unitary protocol exactly.

What this does not tell you

Sample complexity improvements are relative to the baseline each paper names, and they apply to the classes of observables stated. They do not show how a given experiment will perform with real noise.

FAQ

What is a classical shadow?

A compact classical description of a quantum state built from random measurements, enough to estimate many properties.

Why not use global Clifford measurements?

They need extensive gate overhead, which is infeasible on near-term devices, per the 2023 paper.

What do shallow circuits offer?

They interpolate between Pauli and Clifford measurements with minimal gate overhead.

Can post-processing help?

A 2024 paper reports an exponential separation in sample complexity for local measurements.

Is this experimental?

These papers are theoretical or numerical.