Minimal Scenarios Witnessing Genuine Multipartite Nonlocality Possibilistically

Technical Summary: Minimal Scenarios Witnessing Genuine Multipartite Nonlocality Possibilistically

Problem Statement
The paper addresses the challenge of certifying Genuine Multipartite Nonlocality (GMNL) in its strongest possible form: possibilistic incompatibility (also known as strong contextuality or pseudotelepathy). While standard nonlocality rules out local hidden variable theories, and previous GMNL definitions (e.g., Svetlichny) rule out mixtures of lower-partite distributions, a stronger certification is sought. This certification must demonstrate that a quantum strategy can win a nonlocal game with probability 1 (certainty), whereas any model utilizing only resources shared among at most $k < n$ parties (specifically bipartite resources, $k=2$) fails to do so with positive probability.

The central open question investigated is whether such a "strong" separation exists in minimal scenarios. Specifically, can an $n$-party quantum strategy achieve a winning probability of 1 in a scenario where every model built from $(n-1)$-party resources fails? Previous work established this for complex scenarios or specific resource models (e.g., wiring of PR boxes), but often lacked robustness against noise or did not rule out the broader class of Local Operations and Shared Randomness (LOSR) models.

Methodology
The authors employ the inflation technique, a method originally developed for bounding correlations in quantum networks, to derive robustness bounds for nonlocal games.

1. The 5-Cycle Game: The primary focus is the simplest nontrivial 5-party Bell scenario, denoted $(5,2,2)$, involving five parties ($A, B, C, D, E$), two measurement settings per party ($X, Z$), and two outcomes ($0, 1$). The game is defined by six predicates (P1–P6) involving modulo-2 sums of outcomes.
2. Inflation Proofs: The authors construct specific "inflation networks" (copies of the original network with shared randomness) to derive contradictions.
   • They first prove that if all predicates are satisfied with probability 1, specific parties must output constant values, leading to a contradiction with the game's constraints (Proposition 1).
   • They then utilize a higher-order inflation to convert this logical obstruction into a quantitative robustness bound. This involves solving linear programs to bound the winning probability of models restricted to pairwise (bipartite) resources.
3. Simulability Analysis: The paper investigates whether the 5-cycle game's statistics can be simulated by resources shared among 3 or 4 parties. This involves formulating linear programs to find no-signaling distributions that satisfy the game's conditions under restricted resource constraints.
4. $\epsilon$-Close Separation: For the general $n$-party case, the authors construct a family of 3-party games (and generalize to $n$) where quantum strategies achieve a winning probability of $1 - O(1/N^2)$, while bipartite models are bounded by $1 - O(1/N)$.

Key Contributions and Results

1. Strong LOSR-GMNL in the Minimal 5-Party Scenario:
   The paper demonstrates that the 5-cycle game admits a quantum strategy (using a 5-qubit state) that wins with probability 1. Crucially, using the inflation technique, the authors prove that any model built from bipartite nonclassical resources (under the LOSR framework, which includes arbitrary measurements and shared randomness) has a winning probability bounded above by $121/126 \approx 0.960$.

   • This result improves upon previous bounds (e.g., $53/54 \approx 0.981$ for wiring models) and applies to the broader, more comprehensive LOSR-GMNL definition, ruling out entangled measurements on bipartite resources.

2. Hierarchy of Simulability:
   A comprehensive analysis reveals the specific resource requirements for the 5-cycle game:

   • Bipartite Resources: Cannot simulate the behavior (Winning prob $\leq 0.960$).
   • Tripartite Resources: The status remains an open question for general LOSR models, though it is proven to be non-simulable under the stricter no-signaling Svetlichny definition.
   • Quadpartite Resources: The game is simulable. The authors construct a no-signaling Svetlichny-type model using 4-party resources that replicates the full quantum statistics of the 5-party winning strategy.
   • Implication: The 5-cycle game is not "genuinely" 5-partite nonlocal in the sense that it requires 5-party resources; 4-party resources suffice. However, it is genuinely nonlocal against all bipartite models.

3. $\epsilon$-Optimal $n|(n-1)$ Separation:
   While the paper does not find a perfect (probability 1) separation for $n=3$ against $n-1=2$ resources, it constructs a family of 3-party games where the quantum winning probability approaches 1 arbitrarily closely ($1 - \delta/N^2$) while bipartite models are bounded away from 1 ($1 - \gamma/N$).

   • This establishes that for any desired precision $\epsilon$, there exists a game where the quantum strategy is experimentally indistinguishable from a perfect pseudotelepathic strategy, while bipartite models fail with positive probability.

4. Technical Advancement:
   The authors provide explicit robustness bounds derived via inflation, noting that the complexity of the associated linear programs grows exponentially with the number of parties. Their ability to derive these bounds for the 5-party case represents a significant computational and theoretical achievement in the study of strong nonlocality.

Significance and Claims
The paper claims to provide the first demonstration of strong LOSR-GMNL in the simplest possible nontrivial 5-party scenario $(5,2,2)$. By establishing a robust separation ($\approx 0.960$) between quantum strategies and all bipartite-resource models, the work advances the certification of network resource structures beyond simple local hidden variables.

The authors highlight an intriguing parallel between their findings and the Mermin-GHZ game:

• In the GHZ game $(3,2,2)$, quantum mechanics requires the third party to exhibit pseudotelepathy, as bipartite superquantum resources (PR boxes) can simulate the behavior.
• In the 5-cycle game $(5,2,2)$, quantum mechanics requires the "extra" fifth party to exhibit strong LOSR-GMNL, as 4-party no-signaling resources (even superquantum ones) can simulate the behavior.

Finally, the paper addresses the practical limitations of "perfect" experiments. By demonstrating $\epsilon$-close separations, the authors argue that while true probability-1 winning is experimentally elusive due to noise, there exist families of games where the quantum advantage is so robust that the behavior is observationally indistinguishable from true pseudotelepathy, provided experimental precision is sufficient. This lays the groundwork for future applications in device-independent protocols and the foundational understanding of multipartite quantum networks.