Quantum Reconstruction Program (QRP)

• Quantum Reconstruction Program (QRP) is a multifaceted research domain that derives quantum formalism from operational data and foundational principles.
• It employs methodologies including axiomatic reconstructions, phase-space and weak-value techniques, and variational tomographic procedures to uncover hidden quantum structures.
• QRP’s approaches enhance practical applications such as fault-tolerant circuit recovery and high-fidelity state reconstruction on NISQ devices.

In the literature, the Quantum Reconstruction Program (QRP) is not a single technique but a family of reconstruction agendas centered on recovering quantum-theoretic structure from operationally accessible data or principles. In one major usage, QRP denotes the derivation of the Hilbert-space formalism from information-theoretic, logical, or symmetry-based postulates. In another, it denotes concrete procedures for reconstructing unknown states, inputs, or circuit structure from measurement statistics, weak values, reservoir outputs, or execution traces. What unifies these otherwise heterogeneous usages is the attempt to replace direct appeal to a finished formalism by an inferential bridge from constrained observables, operational rules, or structural assumptions to the target object of interest (Hoehn, 2014, Gosson et al., 2011, Jabeen et al., 1 Jul 2025).

1. Scope and uses of the term

In the cited literature, “QRP” spans at least five distinct but partially overlapping strands.

Strand	Representative papers	Reconstruction target
Foundational reconstruction	(Hoehn, 2014, Stairs, 2015, Stuckey et al., 2024)	Hilbert-space or event structure from operational, logical, or symmetry principles
Weak-value / phase-space reconstruction	(Gosson et al., 2011)	One state of a pre/postselected pair from a cross-Wigner quasi-distribution
Tomographic reconstruction	(Jabeen et al., 1 Jul 2025, Yu et al., 2018, Koni et al., 26 Sep 2025, Ghosh et al., 2020)	Unknown quantum states from measurement data
Reservoir-based probing / reversible transformation	(Kobayashi et al., 2023, Kobayashi et al., 2024, Wakaura et al., 23 Feb 2026)	Input information, phase structure, or reversible encode–decode maps
Program- and trace-level reconstruction	(Trochatos et al., 20 Aug 2025, Seidel et al., 2023)	Circuit dataflow or inverse circuit fragments

This multiplicity matters. In foundational work, reconstruction is primarily axiomatic: the objective is to show why quantum theory has the mathematical structure it has. In tomography and learning-based work, reconstruction is operational: the objective is to recover a state or process from finite data. In reservoir and architecture papers, reconstruction becomes diagnostic or algorithmic: one infers hidden signal flow, phase structure, or circuit organization from indirect observables (Berghofer, 2024).

2. Axiomatic and principle-theoretic reconstruction

A central foundational strand reconstructs quantum theory from rules governing information acquisition. In the question-based framework of Höhn and Wever, an observer $O$ interrogates a system $S$ with binary questions $Q_i$, and the state of $S$ relative to $O$ is the collection of probabilities $y_i$ for an informationally complete set of questions. The reconstruction is driven by four rules: a finite information limit, the existence of complementary information, conservation of total information between interrogations, and continuous time evolution such that every consistent such evolution is possible. From these rules one derives the compatibility and complementarity structure of questions, the quadratic information measure

$I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$

the Bloch ball and Bloch disc as the single-system state spaces for qubit and rebit cases, and reversible dynamics given by $SO(3)$ or $SO(2)$ in the single-system setting (Hoehn, 2014).

An older reconstruction lineage proceeds through quantum logic. In Stairs’ synopsis, quantum logic is treated not as a replacement theory but as a reconstruction program based on the non-Boolean lattice structure of experimental propositions. Piron lattices—atomic, complete, irreducible, orthomodular lattices satisfying the covering law—support representation results that recover the lattice of closed subspaces of generalized vector spaces. Solér’s theorem then supplies an additional condition that moves the reconstruction toward ordinary Hilbert-space settings over $\mathbb{R}$, $S$0, or $S$1. This approach is presented as a genuine success in revealing noncommutativity, entwinement, and structural features of quantum probability, while also having clear limitations: finite-dimensional reconstruction remains incomplete, the covering law has an uncertain physical interpretation, and standard lattice frameworks do not naturally accommodate POVMs (Stairs, 2015).

A more recent and philosophically ambitious line argues that QRP becomes complete only when its information-theoretic postulates are supplied with a deeper symmetry principle. On this view, the relativity principle, formulated as no preferred reference frame (NPRF), plays for quantum theory the role it plays in special relativity. The proposed analogy is

$S$2

After “spatializing” measurement, Information Invariance & Continuity is read as a Planck postulate: all observers measure the same value of Planck’s constant $S$3, regardless of relative orientation or location. Darrigol’s discreteness requirement, Brukner and Zeilinger’s information invariance, and Khrennikov’s quantum action invariance are then treated as equivalent formulations of the same underlying principle. This line also presents “average-only” conservation formulas for Stern–Gerlach and Bell-state experiments as derivations of the usual qubit and bipartite probabilities from discreteness plus correspondence or symmetry assumptions. These claims are explicitly framed as a proposal for “completing” QRP rather than as a settled consensus of the field (Stuckey et al., 2024, Stuckey et al., 29 Jun 2026).

3. Weak values, cross-Wigner transforms, and phase-space reconstruction

A mathematically sharp reconstruction problem arises in the weak-value formalism. For preselected state $S$4, postselected state $S$5, and observable $S$6 with Weyl symbol $S$7, the weak value is

$S$8

De Gosson shows that this quantity can be written as a phase-space average over the normalized complex quasi-distribution

$S$9

where $Q_i$0 is the cross-Wigner transform

$Q_i$1

The weak value then takes the form

$Q_i$2

This identifies weak values as averages over a complex phase-space density encoding interference between the preselected and postselected states (Gosson et al., 2011).

The reconstruction theorem is stronger than the usual single-state Wigner-function statement. The cross object $Q_i$3, together with one factor $Q_i$4 or $Q_i$5, uniquely determines the other factor, not merely up to phase. The quasi-distribution has marginals

$Q_i$6

and therefore simultaneously encodes position-space and momentum-space overlaps. The proof exploits the Grossmann–Royer operator, its relation to the cross-Wigner transform, and Moyal’s identity. Explicit reconstruction formulas are given for either state using any auxiliary $Q_i$7 with nonvanishing overlap in the denominator. In this formulation, Lundeen et al.’s direct weak-measurement wavefunction-reconstruction scheme becomes a special case of a more general phase-space theorem for arbitrary pre/postselected pairs and arbitrary observables (Gosson et al., 2011).

4. State reconstruction from measurement data

A large contemporary strand of QRP is algorithmic tomography from measurement statistics. A representative example is variational quantum-circuit tomography on NISQ devices. Here the unknown target state is measured in selected Pauli bases $Q_i$8, producing classical distributions $Q_i$9, while a variational ansatz $S$0 produces matched distributions $S$1. Training minimizes a symmetrized KL-divergence loss

$S$2

with

$S$3

The default optimizer is SPSA. Reported simulation fidelities are about $S$4 for 3-qubit GHZ and 3-qubit spin-chain ground states, $S$5 for 6-qubit GHZ, around $S$6 for a 6-qubit XXZ spin-chain ground state with a 16-layer ansatz, and typically above $S$7 for 6-qubit random-circuit states. On hardware, the protocol reconstructs a 3-qubit GHZ state with fidelity $S$8 on IBM Brisbane and $S$9 on IonQ Forte. In incomplete-basis experiments, using 15 of 27 bases for 3 qubits or 200 of 729 for 6 qubits, median fidelities remain $O$0 for 3-qubit GHZ and spin-chain states, $O$1 for 6-qubit GHZ, and $O$2 for the 6-qubit spin-chain ground state (Jabeen et al., 1 Jul 2025).

A different learning-based protocol uses semi-quantum reinforcement learning to reconstruct an unknown photonic qubit from a limited number of copies. An environment qubit $O$3, a register $O$4, and an agent qubit $O$5 are coupled by a CNOT from environment to register, followed by a single-shot measurement of $O$6. Outcome $O$7 is a reward, leaving the agent essentially unchanged, while $O$8 is a punishment, inducing a random unitary update. The exploration range $O$9 evolves according to

$y_i$0

with $y_i$1. The reported photonic implementation achieves fidelities over $y_i$2 within 50 iterations, and above $y_i$3 for one test state after about 20 iterations. The paper explicitly compares the protocol with tomography under the same copy budget and reports regimes in which the RL procedure extracts more useful information from limited copies than maximum-likelihood qubit tomography (Yu et al., 2018).

Another variational route reformulates tomography as an Ising optimization problem. Given projective measurements $y_i$4 and a tomography matrix $y_i$5, the least-squares objective

$y_i$6

is encoded into an Ising Hamiltonian

$y_i$7

which is minimized by VQE. In a proof of concept for OAM-entangled photons in a two-dimensional $y_i$8 subspace with 36 overcomplete measurements, simulator fidelities are approximately $y_i$9, $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$0, $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$1, and $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$2, while IBM hardware reconstructions with level-1 error mitigation reach $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$3 on $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$4 and $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$5 on $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$6, relative to MLE reconstructions. The paper explicitly presents this as methodological proof of concept rather than evidence of quantum advantage (Koni et al., 26 Sep 2025).

Quantum reservoir state tomography takes yet another route. An unknown state $I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$7 is injected into a fixed random quantum reservoir; only average site occupations

$I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$8

are measured; and a trained linear readout reconstructs the density matrix or Wigner function via

$I_{O\to S}(\vec y)=\sum_i (2y_i-1)^2,$9

For finite-dimensional systems, the paper argues that a reservoir with size $SO(3)$0 can reconstruct arbitrary $SO(3)$1-dimensional states with unit fidelity, and that time multiplexing reduces the effective hardware size requirement to $SO(3)$2, where $SO(3)$3 is the number of time samples. The same framework is extended to continuous-variable states by reconstructing the Wigner function on a phase-space grid, again from occupation-number measurements in a single physical setup (Ghosh et al., 2020).

5. Reservoir-based probing, phase diagnostics, and reversible information transformation

In quantum reservoir probing, QRP is explicitly defined as the inverse paradigm of quantum reservoir computing. Instead of using a quantum many-body system as a feature map for an external machine-learning task, the task performance itself is used as a diagnostic of the system’s internal physics. An input sequence $SO(3)$4 is injected by local quenches, and a single observable $SO(3)$5 is used to reconstruct a delayed input $SO(3)$6 through the linear readout

$SO(3)$7

The determination coefficient

$SO(3)$8

measures how much information about the input is retained in that observable. Applied to the one-dimensional Ising chain, this operator-resolved metric distinguishes ballistic propagation in the free-fermion case from diffusive propagation in the chaotic case, identifies the dominant information channels, and is reported to be more sensitive than OTOCs and TMI to small symmetry-breaking perturbations (Kobayashi et al., 2023).

A closely related use of the same idea targets quantum phase transitions. Here a local quench parameter $SO(3)$9 is encoded into a single site, the system evolves under its Hamiltonian, and a linear estimator attempts to recover $SO(2)$0 from a single-site observable elsewhere in the system. The central empirical claim is that near a quantum critical point, enhanced quantum fluctuations suppress the imprint of the local quench, causing the averaged reconstruction score $SO(2)$1 to dip sharply. This identifies the phase boundary in the transverse-field Ising model at $SO(2)$2, the ANNNI model near $SO(2)$3, the spin-1 XXZ Haldane–Néel transition near $SO(2)$4, and the extended SSH topological transition near $SO(2)$5, all using local quenches and single-site observables (Kobayashi et al., 2024).

In a different direction, reversible information transformation via quantum reservoir computing treats reconstruction as bidirectional recovery of an encoded data sequence. The proposed four-equation protocol uses two reservoirs, two distributed keys, and two secret keys with cross-key pairing: $SO(2)$6

$SO(2)$7

Using a full XYZ Hamiltonian reservoir with $SO(2)$8 data qubits and one ancilla, the feature dimension is

$SO(2)$9

Under ideal conditions, the paper reports machine-precision reconstruction, with MSE in the $\mathbb{R}$0 to $\mathbb{R}$1 range for $\mathbb{R}$2, and identifies the rank condition $\mathbb{R}$3 as necessary. It also reports a clear noise hierarchy—shot noise dominant, depolarizing noise adding a moderate penalty—and a large improvement from asymmetric shot allocation, namely 10 shots for encoding and $\mathbb{R}$4 for decoding, which yields roughly two orders of magnitude MSE improvement relative to the symmetric 1,000-shot setting (Wakaura et al., 23 Feb 2026).

6. Reconstruction of circuit structure and inverse computation

QRP also appears in work on fault-tolerant architectures and quantum software tooling. In surface-code lattice surgery, access traces are defined as space-time patterns of patch activity. TraceQ takes as input the minimal Level-1 trace

$\mathbb{R}$5

where each binary matrix records which patches are active at each time step. From these traces it reconstructs the program’s dataflow DAG through heuristic endpoint detection, BFS decomposition of known subgraphs, DFS with backtracking for ambiguous routing regions, and ambiguity-aware DAG assembly, followed by exact subgraph matching with VF3. On 20 synthetic benchmarks with 30 random perturbations each, subroutine recovery rates are reported as $\mathbb{R}$6 for Square Sparse layouts, $\mathbb{R}$7 for Compact layouts, and $\mathbb{R}$8 for Intermediate layouts, with no false positives observed. The paper explicitly frames this as a side-channel result: even minimal binary activity traces can reveal subroutines or entire fault-tolerant programs (Trochatos et al., 20 Aug 2025).

At the software level, automated uncomputation in Qrisp treats reconstruction as recovery of inverse circuit fragments from program structure. The framework exposes two interfaces, @auto_uncompute and .uncompute(), both based on an improved version of Unqomp. The aim is to restore ancillas to $\mathbb{R}$9, disentangle temporaries, and enable safe deallocation without measurement-based deletion. The paper emphasizes two extensions to the original Unqomp logic: first, synthesized gates can be uncomputed by checking whether the whole wrapped gate is qfree rather than whether its decomposition is qfree; second, the notion of “controlled” is generalized to “permeable,” defined by

$S$00

If a unitary is permeable on its first $S$01 qubits, it admits the block-diagonal form

$S$02

and overlapping permeable operators commute under the stated conditions. In this sense, reconstruction is not of an unknown state but of the inverse computational structure required for qubit recycling and algorithmic correctness (Seidel et al., 2023).

7. Interpretation, philosophical significance, and open constraints

Several papers argue that reconstruction and interpretation are distinct but inseparable. One line of defense maintains that interpreting quantum mechanics usually takes the formalism as given, whereas reconstructing quantum theory starts from simple physical principles and derives the formalism instead. On that view, reconstructions help explain why complex Hilbert spaces, Born-rule structure, and operational constraints arise, and therefore place pressure on standard $S$03-ontic or objectivist interpretations that do not naturally explain the form of the theory from within their own conceptual vocabulary (Berghofer, 2024).

A more systematic methodological proposal introduces an “interpretation-free zone,” a preliminary descriptive stage in which formal results, operational assumptions, experimental procedures, and modelling heuristics are surveyed without prior metaphysical commitment. Reconstruction is then used to identify which facts are interpretationally relevant through “stratification” and “operationalization.” This proposal treats QRP not only as a derivational program but also as a tool for discovering new physical principles and reorganizing ontology; examples include local tomography, symmetric transition probabilities, the Operational Indistinguishability Postulate, and the derived principle of complementarity of persistence and nonpersistence (Goyal, 19 Dec 2025).

Another interpretive proposal argues that if QRP is taken seriously and understood as anti-$S$04-ontic, then its natural ally may not be $S$05-epistemicism in the factive sense of knowledge, but $S$06-doxastic or experiential approaches. In that proposal, Rovelli-style information postulates are reformulated as “Limited perspective” and “Unlimited perspectives,” in order to ground reconstruction in non-factive experiential terms more congenial to QBist-style views and to a possible “Degrees of Epistemic Justification Interpretation” (Berghofer, 2024).

QBist-inspired qplex reconstructions sharpen the issue by showing both the power and the incompleteness of current geometric constraints. In the SIC-based qplex framework, bipartite correlators can be written as inner products of $S$07-vectors. For CHSH, the shared inner-product geometry enforces the Tsirelson bound $S$08. For the three-outcome CGLMP inequality $S$09, however, the same norm and inner-product constraints still allow the algebraic maximum $S$10. The stated implication is that qplex geometry captures a genuinely quantum-like core but does not by itself recover the full set of quantum correlation constraints; further principles are required to distinguish Hilbert qplexes from general qplexes (Gupta et al., 5 Jun 2026).

Taken together, these strands show that QRP is best understood as a plural research domain rather than a monolithic doctrine. Its targets range from the axiomatic reconstruction of quantum kinematics, to exact phase-space recovery from weak values, to practical tomography on NISQ devices, to operator-resolved many-body diagnostics, to trace-based recovery of fault-tolerant circuit structure. The recurrent claim across these otherwise disparate contexts is that quantum structure can often be recovered more transparently from operational, geometric, informational, or algorithmic constraints than from the bare formalism alone.