Many-Body Contextuality Game

• Many-body contextuality game is a multiplayer nonlocal game defined by engineered parity constraints that witness quantum advantage in complex quantum states.
• It employs local two-qubit Pauli measurement contexts on cluster states, linking string order parameters from one-dimensional SPT order to measurable benchmarks.
• The game establishes a clear quantum-versus-classical separation and offers an operational tool for testing long-range order under noisy, thermal conditions.

A many-body contextuality game is a multiplayer nonlocal game whose winning conditions are engineered so that certain many-body quantum states admit perfect or classically inaccessible strategies, thereby turning contextuality into an operational witness of quantum advantage. In the formulation tied to one-dimensional symmetry-protected topological order, the game is built from local two-qubit Pauli measurement contexts on a ring, and its performance is controlled by symmetry data, string order parameters, and—away from zero temperature—twisted string order parameters together with symmetry representation expectation values. In this setting, the game serves simultaneously as a contextuality test, a bounded-depth quantum-versus-classical separation, and a benchmark for preparing long-range ordered cluster/SPT states on noisy devices (Fagan et al., 13 Mar 2026).

1. Base game and many-body generalization

The base construction is a three-player triangle game. Players $p \in \{1,2,3\}$ receive inputs $x_p \in \{0,1\}$ uniformly at random and must output three bits $y_p(x_p)$. If $x_p=0$, then

$y_p(0)=(a_p,b_p,c_p),$

and if $x_p=1$, then

$y_p(1)=(d_p,b_p,e_p).$

The bit $b_p$ is shared between the two local contexts. The global winning constraints are

$|\mathbf{y}|=0,$

$a_1+a_2+a_3=0,$

$x_p \in \{0,1\}$0

$x_p \in \{0,1\}$1

where the Hamming weight is used in the first condition, addition is modulo $x_p \in \{0,1\}$2, and $x_p \in \{0,1\}$3 is cyclic (Fagan et al., 13 Mar 2026).

The classical non-communicating bound follows from the observation that the $x_p \in \{0,1\}$4 input cannot satisfy the induced parity constraint. With all $x_p \in \{0,1\}$5 inputs uniformly random, this gives

$x_p \in \{0,1\}$6

A stronger variant using a subset of inputs yields a $x_p \in \{0,1\}$7 bound, although that variant is not pursued in the mixed-state SPT analysis (Fagan et al., 13 Mar 2026).

A perfect quantum strategy exists already in the three-player setting. Using the six-qubit cyclic cluster state $x_p \in \{0,1\}$8, each player $x_p \in \{0,1\}$9 holds qubits $y_p(x_p)$0 and performs commuting two-qubit Pauli measurements. For $y_p(x_p)$1, the measurement context is

$y_p(x_p)$2

producing $y_p(x_p)$3. For $y_p(x_p)$4, the context is

$y_p(x_p)$5

producing $y_p(x_p)$6. These contexts enforce $y_p(x_p)$7, and the remaining win conditions are identified with cluster-state stabilizers or their negatives, so

$y_p(x_p)$8

for the cluster-state strategy (Fagan et al., 13 Mar 2026).

The multiplayer, or genuinely many-body, triangle game places $y_p(x_p)$9 players on a ring of $x_p=0$0 qubits. Three special players $x_p=0$1, equally spaced, receive random inputs in $x_p=0$2; all other players receive $x_p=0$3 and output only $x_p=0$4. For each edge $x_p=0$5 between special players, one defines edge strings $x_p=0$6, with $x_p=0$7, as the sum of the corresponding symbols along the players on that edge. The many-body winning conditions are the three-player ones augmented by these edge-string contributions. This embeds the triangle game into a ring geometry while preserving a local constant-depth quantum strategy (Fagan et al., 13 Mar 2026).

2. Classical depth bounds and quantum strategy

The many-body game is designed so that the relevant classical comparison class is not unrestricted centralized computation but restricted-geometry classical processing. Classical players are allowed $x_p=0$8 rounds of nearest-neighbor communication, equivalently depth-$x_p=0$9 classical circuits with nearest-neighbor fan-in. If

$y_p(0)=(a_p,b_p,c_p),$0

the classical bound remains

$y_p(0)=(a_p,b_p,c_p),$1

Only when $y_p(0)=(a_p,b_p,c_p),$2 reaches $y_p(0)=(a_p,b_p,c_p),$3 can players share all inputs and achieve $y_p(0)=(a_p,b_p,c_p),$4. Consequently, any constant-depth quantum strategy with

$y_p(0)=(a_p,b_p,c_p),$5

establishes a separation from sublinear-depth classical circuits (Fagan et al., 13 Mar 2026).

The cluster-state measurement strategy underlying this separation is local and uses only two-qubit Pauli measurements. In the general $y_p(0)=(a_p,b_p,c_p),$6 notation, the onsite representation is

$y_p(0)=(a_p,b_p,c_p),$7

and the strategy uses the contexts

$y_p(0)=(a_p,b_p,c_p),$8

and

$y_p(0)=(a_p,b_p,c_p),$9

with $x_p=1$0 the fixed-point edge operators. In the three-player reduction this becomes the explicit Pauli pair of contexts $x_p=1$1 and $x_p=1$2 (Fagan et al., 13 Mar 2026).

This construction fits into a broader body of work in which graph states generate many-body pseudo-telepathy games. In the graph-state formalism, players correspond to vertices, questions select local Pauli $x_p=1$3 measurements, and the parity constraints are indexed by involved subsets $x_p=1$4. For non-bipartite graphs, the resulting many-body contextuality graph game is pseudo-telepathic, while the associated contextuality scenario can exhibit large multipartiteness width (Anshu et al., 2016). A related stabilizer-testing viewpoint samples stabilizer elements directly and asks players to reproduce stabilizer eigenvalue constraints; this yields a universal classical cap at $x_p=1$5 whenever a full-query stabilizer-testing game has non-zero quantum advantage (Zhao et al., 18 Dec 2025).

A salient feature of the SPT-based many-body triangle game is that the quantum strategy uses the same constant-depth resource-preparation and local-measurement pattern that is natural for cluster states and MBQC. This suggests that the game is not merely a Bell-type witness but an operational reformulation of the nonclassical computational structure present in the phase (Fagan et al., 13 Mar 2026).

3. Relation to one-dimensional SPT order

The many-body contextuality game analyzed for mixed states is anchored in one-dimensional SPT order with onsite finite abelian symmetry $x_p=1$6. Such phases are classified by $x_p=1$7 via projective edge representations. For the spin-$x_p=1$8 chain of even length $x_p=1$9 with $y_p(1)=(d_p,b_p,e_p).$0, the onsite representation on two-site blocks is

$y_p(1)=(d_p,b_p,e_p).$1

The model Hamiltonian, with periodic boundary conditions, is

$y_p(1)=(d_p,b_p,e_p).$2

At the cluster fixed point $y_p(1)=(d_p,b_p,e_p).$3, the ground state is the one-dimensional cluster state

$y_p(1)=(d_p,b_p,e_p).$4

with commuting stabilizers

$y_p(1)=(d_p,b_p,e_p).$5

These states are canonical MBQC resources and form the reference point for the game (Fagan et al., 13 Mar 2026).

The relevant order parameter is the string order parameter

$y_p(1)=(d_p,b_p,e_p).$6

where $y_p(1)=(d_p,b_p,e_p).$7 is the product of onsite symmetry operators on the bulk of the segment $y_p(1)=(d_p,b_p,e_p).$8. In symmetric ground states $y_p(1)=(d_p,b_p,e_p).$9, one has

$b_p$0

so the expectation value of $b_p$1 diagnoses the SPT phase (Fagan et al., 13 Mar 2026).

For pure symmetric SPTO states, contextual advantage is linked directly to these SOPs. If all three nontrivial cluster SOPs exceed $b_p$2, then the many-body contextuality game with equidistant $b_p$3 and block length matching $b_p$4 satisfies

$b_p$5

This is the pure-state threshold connecting SPT long-range order to a bounded-depth quantum-classical separation (Fagan et al., 13 Mar 2026).

The mixed-state extension requires a refinement of this criterion. At the fixed point, the twisted SOP is defined by

$b_p$6

Using the projective edge-representation relations, it equals

$b_p$7

so on fixed-point symmetric pure states it acts with the twist phase. This is the key object allowing the many-body contextuality game to probe thermal and generally non-symmetric states, where conventional SOP expectations alone are no longer sufficient (Fagan et al., 13 Mar 2026).

4. Mixed states, winning probability, and thermal thresholds

For a general state $b_p$8, possibly mixed and non-symmetric, the cluster-state measurement strategy yields a quantum winning probability

$b_p$9

and the worst-case performance over distinct nontrivial $|\mathbf{y}|=0,$0 is

$|\mathbf{y}|=0,$1

In this formulation, quantum advantage in mixed states is measured by a combination of twisted SOP and symmetry representation expectation values rather than by a pure-state SOP threshold alone. The paper states in particular that “a nontrivial SPTO state is necessary to pass the value $|\mathbf{y}|=0,$2” (Fagan et al., 13 Mar 2026).

Thermal states are taken as Gibbs states

$|\mathbf{y}|=0,$3

with $|\mathbf{y}|=0,$4 and $|\mathbf{y}|=0,$5. At the cluster point, the Gibbs state is equivalent to applying independent single-qubit dephasing with probability

$|\mathbf{y}|=0,$6

to $|\mathbf{y}|=0,$7. For this thermal cluster state, the minimum winning probability has the exact form

$|\mathbf{y}|=0,$8

This beats the classical $|\mathbf{y}|=0,$9 bound whenever $a_1+a_2+a_3=0,$0, where

$a_1+a_2+a_3=0,$1

Equivalently, at fixed $a_1+a_2+a_3=0,$2, the maximum system size supporting advantage is

$a_1+a_2+a_3=0,$3

For $a_1+a_2+a_3=0,$4 and $a_1+a_2+a_3=0,$5, the critical temperature is

$a_1+a_2+a_3=0,$6

corresponding to $a_1+a_2+a_3=0,$7 dephasing (Fagan et al., 13 Mar 2026).

A central conclusion is finite-size robustness. Although

$a_1+a_2+a_3=0,$8

it is “relatively flat with $a_1+a_2+a_3=0,$9 for moderately large $x_p \in \{0,1\}$00,” and for fixed small $x_p \in \{0,1\}$01, $x_p \in \{0,1\}$02 grows exponentially as $x_p \in \{0,1\}$03. Numerical METTS data across the full SPT phase similarly show a nonzero $x_p \in \{0,1\}$04 for finite $x_p \in \{0,1\}$05, yielding a “quantum-advantage surface” in $x_p \in \{0,1\}$06-space that shrinks to zero only in the thermodynamic limit (Fagan et al., 13 Mar 2026).

The paper also reports a geometric insensitivity on the quantum side: for sufficiently large separation $x_p \in \{0,1\}$07 relative to the correlation length, $x_p \in \{0,1\}$08 depends only on $x_p \in \{0,1\}$09 and $x_p \in \{0,1\}$10 and is essentially independent of the geometric placement of $x_p \in \{0,1\}$11, in contrast to the classical hardness, which depends on the minimum separation between the three special players. This suggests that the same finite-size state may remain a viable witness of contextuality across a range of experimental layouts (Fagan et al., 13 Mar 2026).

5. Numerical method and operational benchmark

The finite-temperature analysis is carried out with the minimally entangled typical thermal states algorithm. The required observables are

$x_p \in \{0,1\}$12

which are then inserted into the mixed-state winning formula. METTS approximates thermal expectation values by sampling typical-thermal MPS states according to

$x_p \in \{0,1\}$13

where

$x_p \in \{0,1\}$14

The sampling steps are: start from a random product state $x_p \in \{0,1\}$15; imaginary-time evolve via TDVP within an MPS manifold to get $x_p \in \{0,1\}$16 and record $x_p \in \{0,1\}$17; then collapse to a new product state $x_p \in \{0,1\}$18 with probability $x_p \in \{0,1\}$19, which satisfies detailed balance (Fagan et al., 13 Mar 2026).

The implementation specifics are explicit. The simulations use periodic boundary conditions and focus primarily on $x_p \in \{0,1\}$20, with other $x_p \in \{0,1\}$21 values also studied. The imaginary-time TDVP step size is

$x_p \in \{0,1\}$22

with Krylov expansion of order $x_p \in \{0,1\}$23 between TDVP steps. The maximum MPS bond dimension is $x_p \in \{0,1\}$24, the cutoff is $x_p \in \{0,1\}$25, the sampling uses $x_p \in \{0,1\}$26 METTS iterations, and $x_p \in \{0,1\}$27 warm-up iterations are discarded. The standard error is estimated as

$x_p \in \{0,1\}$28

with

$x_p \in \{0,1\}$29

and $x_p \in \{0,1\}$30, where $x_p \in \{0,1\}$31 is the first zero-crossing lag. The collapse basis matters: $x_p \in \{0,1\}$32-only local collapse strongly outperformed $x_p \in \{0,1\}$33-only, $x_p \in \{0,1\}$34-only, alternating, and fully random collapse choices for this SPT model (Fagan et al., 13 Mar 2026).

The game also admits a direct fidelity interpretation. Let $x_p \in \{0,1\}$35 be the graph-state basis, $x_p \in \{0,1\}$36, and let $x_p \in \{0,1\}$37 be the global fidelity with the cyclic cluster state $x_p \in \{0,1\}$38. Since the winning probability depends only on the graph-diagonal components of $x_p \in \{0,1\}$39, one obtains the state-independent lower bound

$x_p \in \{0,1\}$40

Equality holds only for the pure cluster state. Operationally, any device state with fidelity $x_p \in \{0,1\}$41, together with sufficiently accurate contextual measurements, is guaranteed to beat the classical bound in the multiplayer triangle game. The game therefore provides an operational meaning to benchmarking the capacity to create long-range cluster/SPT order (Fagan et al., 13 Mar 2026).

This operational use of contextuality is consonant with other recent many-body proposals. Stabilizer-testing games make the observed winning probability a linear function of the fidelity to a target stabilizer state, and for cyclic cluster states the classical value approaches $x_p \in \{0,1\}$42 exponentially fast, implying that exponentially small fidelity can still suffice to witness contextuality (Zhao et al., 18 Dec 2025). Experimental work on superconducting processors has also demonstrated bounded-resource contextuality advantages through GHZ parity games, Kochen–Specker–Bell inequalities, and hidden linear function tasks, treating contextuality-based scores as device benchmarks (Kumar et al., 1 Dec 2025).

6. Broader context, misconceptions, and open questions

The many-body contextuality game should not be conflated with arbitrary many-player XOR or two-body-correlation inequalities. One important limitation established elsewhere is that multi-party XOR-type inequalities involving only two-body correlation functions cannot exhibit pseudo-telepathy: finite perfect quantum value together with strictly subunit classical value is impossible in that class (Gnaciński et al., 2015). By contrast, the triangle-game construction tied to SPT order uses structured many-body parity constraints realized through local two-qubit contexts on a graph-like resource state, and the resulting classical hardness is expressed in bounded-depth nearest-neighbor communication rather than in unrestricted classical value alone (Fagan et al., 13 Mar 2026).

A second misconception is that many-body contextuality games are intrinsically about pure states. The mixed-state analysis shows otherwise: near-term devices naturally prepare noisy or thermal states, and the relevant mixed-state witness is not just the untwisted string order parameter but a combination of twisted SOP and symmetry representation expectations. This suggests that contextuality can remain operationally visible well away from the pure fixed point, even though finite-temperature onsite-symmetry SPT order does not survive in the thermodynamic limit (Fagan et al., 13 Mar 2026).

The limitations are explicit. The main theorem and simulations assume translationally invariant states on even-length rings, equally spaced $x_p \in \{0,1\}$43 for classical hardness, and the cluster measurement strategy with two-qubit dichotomic $x_p \in \{0,1\}$44 observables. Finite-$x_p \in \{0,1\}$45 SPT order does not survive in the thermodynamic limit for onsite symmetries, and correspondingly the critical temperature vanishes as $x_p \in \{0,1\}$46. It is known that thermal cluster states are constant-depth preparable, but whether generic Gibbs states of one-dimensional $x_p \in \{0,1\}$47 SPT Hamiltonians are preparable by constant-depth quantum channels remains open. Extending the theorem to spin-1 Haldane SPT phases would require mapping to dichotomic edge observables via isometries (Fagan et al., 13 Mar 2026).

Several open directions are highlighted. One is whether nonzero SOP expectation values imply contextuality more generally in settings that allow some communication. Another is extension to three-dimensional cluster SPT order protected by higher-form symmetries, where finite-temperature order exists in the thermodynamic limit and could yield more noise-robust contextuality tests. A further question concerns the constant-depth preparability of broader classes of Gibbs states of SPT Hamiltonians (Fagan et al., 13 Mar 2026).

Related lines of work suggest complementary generalizations. CSS-code nonlocal games quantify many-body contextuality through Boolean nonlinearity and Walsh–Hadamard spectra, while CSS submeasurement games enable self-testing, illustrated explicitly for the $x_p \in \{0,1\}$48D toric code (Hart et al., 18 Dec 2025). Infinite, translation-invariant one-dimensional systems display a different set of strengths and limits: some nearest-neighbor scenarios show strong evidence of no contextuality even beyond quantum theory, whereas nearest-and-next-to-nearest-neighbor scenarios can reach ultimate quantum limits and support self-testing-oriented interpretations (Yang et al., 2021). From a single-device perspective, arbitrary contextuality games can also be compiled into operational tests by computational assumptions, using cryptography to enforce temporal separation between sequential measurements (Arora et al., 2024).

Taken together, these results place the many-body contextuality game at the intersection of MBQC resource theory, SPT order diagnostics, stabilizer and graph-state nonlocal games, and experimental benchmarking. In the one-dimensional SPT setting, its defining contribution is to convert long-range order in noisy many-body states into a quantitatively analyzable winning probability, with explicit finite-size thresholds, a bounded-depth classical comparison, and a fidelity lower bound that gives the game direct operational significance (Fagan et al., 13 Mar 2026).