Classical shadows with arbitrary group representations
Maxwell West, Frederic Sauvage, Aniruddha Sen, Roy Forestano, David Wierichs, Nathan Killoran, Dmitry Grinko, M. Cerezo
Abstract
Classical shadows (CS) has recently emerged as an important framework to efficiently predict properties of an unknown quantum state. A common strategy in CS protocols is to parametrize the basis in which one measures the state by a random group action; many examples of this have been proposed and studied on a case-by-case basis. In this work, we present a unified theory that allows us to simultaneously understand CS protocols based on sampling from general group representations, extending previous approaches that worked in simplified (multiplicity-free) settings. We identify a class of measurement bases which we call "centralizing bases" that allows us to analytically characterize and invert the measurement channel, minimizing classical post-processing costs. We complement this analysis by deriving general bounds on the sample-complexity necessary to obtain estimates of a given precision. Beyond its unification of previous CS protocols, our method allows us to readily generate new protocols based on other groups, or different representations of previously considered ones. For example, we characterize novel shadow protocols based on sampling from the spin and tensor representations of , symmetric and orthogonal groups, and the exceptional Lie group .
AI Impact Assessments
(3 models)Scientific Impact Assessment
Core Contribution
This paper develops a unified representation-theoretic framework for classical shadows (CS) protocols based on arbitrary compact group representations. The central innovation is the concept of "centralizing bases" — specifically, non-degenerate commuting subgroup eigenbases (NDCSEs) — which guarantee that the measurement channel M decomposes as a sum of scalar multiples of isotypic projectors: M = Σ_λ a^H_λ P^V_λ. This makes channel inversion trivial regardless of whether the underlying representation has multiplicities, which was the key bottleneck in prior work.
The main technical results are Theorem 1 (the measurement channel for any NDCSE protocol takes this diagonal form, with analytically computable coefficients a^H_λ = d^H_λ/d_λ) and Theorem 2 (variance bounds expressed in terms of these coefficients). Together, they turn the recipe "choose a group G, a representation, and an abelian subgroup H" into an off-the-shelf shadows protocol with guaranteed analytical invertibility.
Methodological Rigor
The mathematical framework is carefully constructed. The proofs leverage classical tools from representation theory (Schur's lemma, character orthogonality, Clebsch-Gordan decompositions) in a clean and systematic way. Lemma 4 — showing that dephasing in an NDCSE is equivalent to composing the H-twirl with projection onto G-diagonal operators — is the key technical insight enabling Theorem 1, and its proof is elegant and convincing.
The paper is thorough in verifying that the framework correctly reproduces known results for global/local Cliffords, matchgate circuits, symplectic unitaries, and orthogonal groups, often with substantially less effort than the original derivations required. The explicit Weingarten calculation for orthogonal shadows (Appendix D6) and the exact variance calculations via Clebsch-Gordan machinery (Appendix E) demonstrate serious technical depth.
One important honesty point: the authors clearly acknowledge that their generic variance bounds (Theorem 2) are often loose compared to bespoke calculations — sometimes by a factor of d — and identify conditions (Theorem 4) under which tighter bounds hold. The treatment of Lemma 3 (proving not all S_n irreps admit NDCSEs) and Lemma 2 (showing Gelfand-Tsetlin bases aren't necessarily centralizing) properly delineates the boundaries of the approach.
Potential Impact
Immediate applications: The SU(2) tensor representation result (Theorem 3) is particularly impactful. It provides sample-efficient estimation of any permutation-invariant observable with poly(n) spectral norm, with variance scaling as O((n+1)^4 ||O||²_∞), while simultaneously eliminating the classical post-processing bottleneck identified in Ref. [21]. This addresses a concrete practical need in quantum state estimation for symmetric systems.
Protocol design: The framework provides a constructive recipe for generating new CS protocols. The demonstration with G_2 — yielding a novel protocol for 7-dimensional systems essentially mechanically — illustrates how the approach enables exploration of previously unconsidered groups with minimal effort.
Broader connections: The observation that operators in exponentially large G-modules are exponentially expensive to estimate, while those in polynomially-sized modules are efficient, connects directly to quantum resource theories (as noted in the discussion). This creates a bridge between shadow tomography complexity and the resource-theoretic structure of quantum states.
Theoretical unification: Having a single framework that encompasses Cliffords, matchgates, symplectic/orthogonal groups, and exotic groups like G_2 provides significant conceptual clarity. The insight that existing protocols implicitly measure in NDCSEs is retrospectively illuminating.
Timeliness & Relevance
Classical shadows is one of the most active areas in quantum information, and the proliferation of case-by-case analyses (global Cliffords, local Cliffords, matchgates, shallow circuits, etc.) has created a clear need for unification. The paper directly addresses the limitation identified in Ref. [17], which required multiplicity-free decompositions. The connection to quantum resource theories and the growing interest in symmetry-adapted quantum protocols makes this timely.
Strengths
1. True unification: Not merely a review but a genuine framework that subsumes and extends multiple prior results
2. Constructive nature: The recipe (choose G, H, representation → get protocol) is concrete and actionable
3. Discovery of genuinely new protocols: The orthogonal weight-basis protocol and G_2 protocol are new, not just re-derivations
4. The Bell basis example (Eq. 12) beautifully illustrates how different abelian subgroups of the same parent group lead to protocols favoring different observable classes
5. Comprehensive appendices: Over 40 pages of detailed calculations ensure reproducibility
Limitations
1. Variance bounds are often loose: The factor-of-d gap for global unitaries is acknowledged but not fully resolved
2. The open question of whether centralizing bases always exist remains unanswered — the framework provides sufficient but not necessary conditions
3. Practical circuit implementation: While the Schur transform has efficient circuits, implementing Haar-random elements of exotic groups like G_2 on a quantum computer is not discussed
4. Restriction to non-adaptive single-copy measurements: As the authors note, extensions to entangled or adaptive settings remain open
5. Limited numerical validation: Only a few numerical experiments (Fig. 6) are presented
Overall Assessment
This is a substantial theoretical contribution that provides the natural mathematical language for understanding group-based classical shadows. It simultaneously unifies a growing body of literature and enables straightforward construction of new protocols. The SU(2) tensor result alone has significant practical value. While the variance bounds could be tighter and some implementation questions remain, the framework's elegance and broad applicability make it an important advance.
Generated Apr 3, 2026
Comparison History (61)
Paper 1 establishes a profound and unexpected connection between quantum measurement theory and score-based diffusion models from machine learning, identifying the feedback Hamiltonian as exactly the score function required by Anderson's reverse-time diffusion theorem. This bridges two major active research areas (quantum control and generative AI), offers immediate practical applications (ML-based score estimation for realistic experiments), and reveals new physics (continuous one-parameter family of path measures unique to quantum systems). Paper 2 provides a valuable unification of classical shadow protocols under general group representations, but is more incremental — extending known frameworks rather than discovering a fundamentally new cross-disciplinary connection.
Paper 2 is more likely to have higher impact: it delivers a conceptual identification (feedback Hamiltonian as a score function) that bridges quantum measurement/feedback control with diffusion-model theory in ML, enabling new algorithmic and experimental directions (trajectory reversal, learning-based score estimation under realistic nonidealities). Its methodological toolkit is rigorous and broadly relevant across quantum control, stochastic processes, and ML. Paper 1 is mathematically unifying and useful for designing new classical-shadow protocols, but its impact is more specialized within quantum tomography/estimation and likely more incremental relative to existing CS frameworks.
Paper 2 presents a unified theoretical framework for classical shadows with arbitrary group representations, which is a foundational contribution to quantum information science. It generalizes and unifies many existing protocols while enabling new ones, offering broad applicability across quantum computing, state tomography, and quantum learning. Its mathematical generality (covering SU(2), symmetric, orthogonal groups, and G_2) ensures wide adoption. Paper 1, while novel in applying quantum algorithms to higher-order coupled oscillators, addresses a more niche intersection of quantum computing and nonlinear dynamics with narrower immediate applicability.
Paper 1 likely has higher impact: it provides a unified, representation-theoretic framework for classical shadows beyond multiplicity-free cases, introduces “centralizing bases” to enable analytic inversion with reduced post-processing, and gives general sample-complexity bounds—tools broadly useful across quantum information, tomography, and near-term experimental protocols. It also systematically generates new measurement schemes across many groups (e.g., SU(2), orthogonal, G2), increasing cross-domain uptake. Paper 2 is timely and promising but hinges on stronger data-access assumptions and targets a narrower application area, which may limit near-term, broad adoption.
Paper 2 presents a unified theoretical framework for classical shadows with arbitrary group representations, which is a foundational contribution to quantum information science. It generalizes and unifies many existing protocols while enabling new ones, providing both analytical characterization and sample-complexity bounds. This broad theoretical advance will impact quantum state tomography, quantum computing benchmarking, and quantum machine learning. Paper 1, while practically impressive in solving large-scale optimization problems, is more of an engineering/systems contribution combining known quantum and classical techniques. Paper 2's foundational nature gives it broader and longer-lasting scientific impact.
Paper 2 likely has higher impact due to clearer near-term real-world applicability and demonstrated scalability: it targets dense large-scale HUBO problems, integrates HPC with quantum circuits, reports concrete performance (500 variables, 170s), and validates on an applied domain (metamaterial design). This positions it as a practical computational paradigm relevant across engineering and scientific optimization. Paper 1 is methodologically rigorous and conceptually novel in unifying classical shadow protocols via general group representations, but its impact is more specialized within quantum information theory and may translate to applications less immediately.
Paper 1 offers a broadly novel, unifying theoretical framework for classical shadows over arbitrary group representations (beyond multiplicity-free cases), introducing “centralizing bases” with analytical inversion and sample-complexity bounds. This both consolidates many existing protocols and enables systematically generating new ones across diverse groups (e.g., SU(2), orthogonal/symmetric groups, G2), suggesting wide cross-cutting impact in quantum information, tomography, and randomized measurement design. Paper 2 is timely and practically relevant for early FTQC systems, but its impact is more architecture-specific and incremental (compiler/benchmark optimizations) with narrower theoretical generality.
Paper 1 provides a foundational theoretical advancement by unifying and extending classical shadow protocols, a crucial technique in quantum information and simulation. Its mathematical generality and introduction of novel protocols give it broad applicability across quantum computing subfields. Paper 2, while highly relevant for near-term fault-tolerant systems, focuses on compiler optimizations for a specific modular architecture, giving it a narrower scope and potentially less long-term scientific breadth compared to the fundamental theoretical contributions of Paper 1.
Paper 2 likely has higher impact due to broader real-world applicability (classical simulation of physically motivated bosonic systems like driven Bose-Hubbard and photonic/Kerr circuits), clear algorithmic contributions with rigorous complexity/error guarantees, and immediate relevance to near-term quantum advantage claims. Its framework can influence quantum optics, many-body physics, and quantum computing benchmarking. Paper 1 is novel and mathematically unifying for classical shadows with general group representations, but its impact is more specialized to quantum tomography/estimation and may have a narrower near-term application footprint despite strong theoretical value.
Paper 2 likely has higher impact: it introduces a concrete computational framework with provable runtime/accuracy guarantees for classically simulating a physically motivated, universal bosonic model (linear optics + Kerr), including quasi-polynomial simulation for log-many Kerr gates and polynomial-time in a weak-nonlinearity regime, plus benchmarks on Bose–Hubbard dynamics. This has immediate relevance to quantum advantage claims, near-term photonics/analog simulators, and many-body physics. Paper 1 is novel and unifying for classical shadows with broad theoretical reach, but its applications are more indirect and specialized to quantum tomography/estimation.