betaFeedback

WorksPhysics and Astronomy

Certifying information flow in multi-agent behaviours

Abstract

We study which input-output behaviours of several agents can be produced by a declared communication schedule, that is, a fixed sequence of senders, recipients, message alphabets, and output times, together with shared randomness. Since a schedule admits only finitely many deterministic protocols, membership is a linear program, and an infeasible target is excluded by a separating inequality. Under explicit operational assumptions, we prove that the affine hull of the achievable behaviours is described exactly by the equalities stating that each joint output statistic depends only on the inputs that can reach the agents involved, so that a violated equality identifies both the statistic and the missing input. One binary message per event already attains this affine hull, so message capacity restricts which behaviours are achievable without being visible to any equality. We apply the result to one-bit broadcast and relay hierarchies for two and three agents, and to low-order projections of the local polytope.

Introduction

Correlations do not by themselves reveal how information was exchanged. Several agents can respond coherently because they communicate during a task, because they share information established beforehand, or both. Bell scenarios make this distinction precise: deterministic local response tables are the vertices of a polytope, shared randomness generates their convex hull, and Bell inequalities exclude behaviours outside it [1, 2]. The same convex approach has been extended to bounded classical communication [3, 4, 5], to lower bounds on average communication cost [6], and to more general causal correlations [7].

A separating inequality shows that a target is incompatible with a resource model, but it need not indicate which information pathway is missing. Affine equalities carry complementary information. The no-signalling equalities state that a subsystem’s statistics do not depend on a remote input, and the local and no-signalling polytopes have the same affine hull although their inequalities differ [8]. An equality can therefore diagnose a forbidden dependence, whereas satisfying every equality does not establish membership.

In this work the resource is a fixed communication schedule, specified by senders, recipients, message alphabets, and output times. Under explicit operational assumptions, we show that the affine hull of the behaviours achievable with a schedule is determined by reachability: a joint statistic of the outputs of a set of agents can depend only on inputs that reach at least one of those agents, and every such dependence is attained in the affine sense (Theorem 5). In this work the resource is a fixed communication schedule, specified by senders, recipients, message alphabets, and output times. Under explicit operational assumptions, we show that the affine hull of the behaviours achievable with a schedule is determined by reachability: a joint statistic of the outputs of a set of agents can depend only on inputs that reach at least one of those agents, and every such dependence is attained in the affine sense (Theorem 5). In particular, one binary message per event already saturates the affine hull, so no affine equality can detect a shortage of message capacity. Equalities and inequalities therefore answer different questions: a violated equality identifies which input failed to reach which agents, whereas only an inequality can show that a channel was too narrow.

The same moment coordinates show what is lost under low-order projections. We show that every projection of the local polytope onto moments of order at most KK contains an explicit ℓ1\ell_1 ball around the uniform behaviour. For GHZ states, the proper reduced states coincide with those of a separable mixture, a known obstruction to few-body detection [9, 10] that is established for all proper orders in Ref. [10 Sec. VI]. For σx/σy\sigma_x/\sigma_y settings we give an explicit moment-space proof of this obstruction together with a quantitative margin. This complements the detection of other many-body states with two-body Bell inequalities [9, 11].

Finally, we consider agents and observers. Quantum decision theory models choices with Hilbert-space structures and with utility and attraction factors [12, 13, 14], and network extensions include explicit information exchange [15]; recent formulations retain the utility attraction decomposition in classical probabilistic terms and model repeated exchange and imitation among groups of agents [16, 17]. We show that an interference term in an individual response rule does not produce nonlocality when each agent uses only its own input and shared prior information, and that a marginal-preserving joint correction changes only the dependence between outputs. Extended Wigner’s friend arguments [18, 19, 20] also declare which agents measure, which records are passed to whom, and when. We locate them in the present framework: Local Friendliness violations lie in the no-signalling affine hull and can therefore be certified only by inequalities, one message carrying a setting suffices to reproduce them, and in tripartite scenarios where Local Friendliness coincides with locality the GHZ results apply directly. The Frauchiger–Renner [19] argument involves no setting choices and lies outside the behaviour-level certificates considered here. These are conditional mathematical statements about input-output behaviours; they do not assess decision theory or interpretations of quantum theory empirically.

Section 2 introduces behaviours and moment coordinates, and Table 1 collects the principal notation. Section 3 states the main result and the associated certificates, Sec. 4 develops declared hierarchies for two and three agents, Sec. 5 treats low-order projections, Sec. 6 discusses decision rules and extended Wigner’s friend scenarios, and Sec. 7 relates the results to prior work. Appendix A applies these results to a single three-agent selection task, contrasting equality and inequality certificates.

SymbolMeaning
xi∈Xix_i\in X_i, did_iInput of agent ii; alphabet; di=∣Xi∣d_i=|X_i|
aia_i, AiA_iOutput; signed output (−1)ai(-1)^{a_i}
P(a∣x)P(a\mid x), PunifP_{\mathrm{unif}}Behaviour; uniform behaviour
λ\lambda, μ\muShared randomness; its distribution or convex weights
ES(x)E_S(x)Moment of the agents in SS, Eq. (3)
DR(K)D_R(K)Number of moment coordinates, Eq. (5)
ΠK\Pi_K, CKC_KProjection onto moments of order ≤K\le K; CK=ΠK(P0)C_K=\Pi_K(\mathcal{P}_0)
A\mathcal{A}, PA\mathcal{P}_{\mathcal{A}}, P0\mathcal{P}_0Schedule; its class; local class
RA(S)R_A(S)Inputs that can reach the outputs in SS
mm; (c,β)(c,\beta); wwTarget moments; separating inequality; equality certificate

Table 1. Principal notation.

Behaviours and moments

There are NN agents, indexed by [N]={1,…,N}[N] = \{1,\ldots,N\}. Agent ii receives an input xix_i from a finite set XiX_i of size did_i and returns ai∈{0,1}a_i \in\{0,1\}; its signed output is Ai=(−1)aiA_i = (-1)^{a_i}. A behaviour is a family

P(a∣x)≥0,∑aP(a∣x)=1(1)P(a \mid x) \ge0,\qquad\sum_a P(a \mid x) = 1 \tag*{(1)}

on the complete Cartesian input set X=∏iXiX = \prod_i X_i. The shared random variable λ\lambda, with distribution μ\mu, is independent of xx. Mutual independence of the inputs is not needed; input coverage and independence from λ\lambda are the relevant assumptions. The uniform behaviour PunifP_{\mathrm{unif}} has P(a∣x)=2−NP(a \mid x) = 2^{-N}.

In the absence of communication,

P(a∣x)=∫∏ipi(ai∣xi,λ) dμ(λ).(2)P(a \mid x) = \int\prod_i p_i(a_i \mid x_i,\lambda)\,d\mu(\lambda). \tag*{(2)}

Absorbing all private randomness into λ\lambda, each agent follows a deterministic table ai=αi(xi)a_i = \alpha_i(x_i). The local polytope P0\mathcal{P}_0 is the convex hull of the 2∑idi2^{\sum_i d_i} resulting deterministic behaviours, each of which is extreme. For two agents with two binary inputs this is equivalent to Fine’s joint distribution for the four potential outcomes [21].

For nonempty S⊆[N]S \subseteq[N], the moment

ES(x)=∑a[∏i∈S(−1)ai]P(a∣x)(3)E_S(x) = \sum_a \left[\prod_{i\in S}(-1)^{a_i}\right]P(a\mid x) \tag*{(3)}

is the expected product of the signed outputs in SS at context xx; when agents have names we write EAE_A, EABE_{AB}, and so on. Fourier inversion on the binary output cube [22] gives, at every context,

P(a∣x)=2−N[1+∑S≠∅(∏i∈S(−1)ai)ES(x)].(4)P(a\mid x) = 2^{-N}\left[1+\sum_{S\ne\varnothing}\left(\prod_{i\in S}(-1)^{a_i}\right)E_S(x)\right]. \tag*{(4)}

with or without communication. A behaviour is no-signalling if each ESE_S depends only on xSx_S; equivalently, each subsystem marginal ignores the other agents’ inputs. Local behaviours are no-signalling, and PunifP_{\mathrm{unif}} has all moments equal to zero.

If each moment ESE_S may depend only on the inputs of a set R(S)R(S) of agents, then listing each ESE_S once as a function of xR(S)x_{R(S)}, and keeping moments of order at most KK, gives

DR(K)=∑S≠∅∣S∣≤K∏j∈R(S)dj(5)D_R(K) = \sum_{\substack{S\ne\varnothing\\ |S|\le K}}\prod_{j\in R(S)}d_j \tag*{(5)}

coordinates. With R(S)=[N]R(S) = [N] for every SS, Dfull:=DR(N)=(2N−1)∏idiD_{\mathrm{full}} := D_R(N) = (2^N-1)\prod_i d_i is the dimension of the set of all behaviours. With the no-signalling restriction R(S)=SR(S)=S we write D0(K)D_0(K); then D0(N)=∏i(di+1)−1D_0(N)=\prod_i(d_i+1)-1, and for equal di=dd_i=d, D0(K)=∑m=1K(Nm)dmD_0(K)=\sum_{m=1}^{K}\binom{N}{m}d^m, so that D0(2)=2N2D_0(2)=2N^2 for d=2d=2.

The projection ΠK\Pi_K keeps only the moments of order at most KK, and CK=ΠK(P0)\mathcal{C}_K=\Pi_K(\mathcal{P}_0) is the image of the local polytope.

Proposition 1. dim⁡CK=D0(K)\dim\mathcal{C}_K = D_0(K).

Proof. By Ref. [8 Theorem 1], the local polytope has the full no-signalling dimension D0(N)D_0(N) and is therefore full-dimensional in the coordinates ES(x)E_S(x); so is its projection onto the D0(K)D_0(K) retained coordinates.

For three agents with binary inputs, Π2\Pi_2 retains 18 of the 26 independent coordinates. The following definition and lemma make Proposition [1] quantitative.

Definition 2. For a nonempty set SS of agents and an input assignment r∈XSr\in X_S, let eS,re_{S,r} denote the moment vector with ES(xS)=1{xS=r}E_S(x_S)=1\{x_S=r\} and ET(xT)=0E_T(x_T)=0 for every T≠S;thatis,T\ne S; that is,e_{S,r}$ is the standard basis vector along the coordinate ES(r)E_S(r).

For example, for two agents with binary inputs the coordinates are EA(x)E_A(x), EB(y)E_B(y), and EAB(x,y)E_{AB}(x,y), eight in total, and eAB,(1,0)e_{AB,(1,0)} has EAB(1,0)=1E_{AB}(1,0)=1 and all other coordinates equal to zero. In the corresponding behaviour each output is uniformly random; the two outputs always agree at inputs (x,y)=(1,0)(x,y)=(1,0) and are uncorrelated at every other input pair.

Lemma 3. For every nonempty SS, every r∈XSr \in X_S, and both signs, ±eS,r\pm e_{S,r} (Definition 2) is the moment vector of a local behaviour. Consequently, for every KK,

{u∈RD0(K):∥u∥1≤1}⊆CK,(6)\left\{u \in\mathbb{R}^{D_0(K)} : \lVert u\rVert_1 \le1\right\} \subseteq C_K, \tag*{(6)}

where ∥u∥1=∑j∣uj∣\lVert u\rVert_1 = \sum_j |u_j|. Moreover, let c∈RD0(K)c \in\mathbb{R}^{D_0(K)} and β∈R\beta\in\mathbb{R} define a valid inequality for CKC_K, that is, c⊤u≤βc^\top u \le\beta for all u∈CKu \in C_K. If c≠0c \ne0, then β≥∥c∥∞=max⁡j∣cj∣>0\beta\ge\lVert c\rVert_\infty= \max_j |c_j| > 0.

Proof. Fix SS, rr, and a sign τ∈{±1}\tau\in\{\pm1\}. Let the shared variable consist of signs (ξi)i∈S(\xi_i)_{i\in S}, uniformly distributed on the set {ξ:∏i∈Sξi=τ}\{\xi: \prod_{i\in S}\xi_i=\tau\}, together with independent unbiased signs ξ1′,…,ξN′\xi'_1,\ldots,\xi'_N. Agent i∈Si\in S outputs ξi\xi_i if xi=rix_i=r_i and ξi′\xi'_i otherwise, and agent i∉Si\notin S outputs ξi′\xi'_i. Each output depends only on the agent’s own input and the shared variable, so the behaviour is local.

Let T≠∅T\ne\varnothing. If T⊈ST\not\subseteq S, or if xi≠rix_i\ne r_i for some i∈Ti\in T, the product of the outputs in TT contains a factor ξi′\xi'_i that is unbiased and independent of the other factors, so ET=0E_T=0. If T⊆ST\subseteq S and xT=rTx_T=r_T, then ET=E∏i∈TξiE_T=\mathbb{E}\prod_{i\in T}\xi_i, which equals τ\tau for T=ST=S. For ∅≠T⊊S\varnothing\ne T\subsetneq S, choose i∈Ti\in T and j∈S∖Tj\in S\setminus T; flipping ξi\xi_i and ξj\xi_j together preserves the distribution of ξ\xi and reverses the sign of ∏k∈Tξk\prod_{k\in T}\xi_k, so ET=0E_T=0. The moment vector is therefore τeS,r\tau e_{S,r}.

The vectors ±eS,r\pm e_{S,r} are the signed standard basis vectors of RD0(N)\mathbb{R}^{D_0(N)}, and their convex hull is the unit ℓ1\ell_1 ball. Since the local polytope is convex, it contains this ball, and ΠK\Pi_K maps it onto the unit ℓ1\ell_1 ball of RD0(K)\mathbb{R}^{D_0(K)}, which proves Eq. (6). Finally, let (S,r)(S,r) with ∣S∣≤K|S|\le K index a coordinate at which ∣cS,r∣=∥c∥∞|c_{S,r}|=\lVert c\rVert_\infty. The point u=sign⁡(cS,r)ΠK(eS,r)u=\operatorname{sign}(c_{S,r})\Pi_K(e_{S,r}) lies in the ball and hence in CKC_K, and c⊤u=∥c∥∞c^\top u=\lVert c\rVert_\infty, so β≥∥c∥∞\beta\ge\lVert c\rVert_\infty. □

Information flow, affine hulls, and certificates

A two-agent example. Take binary inputs xx for Alice and yy for Bob. Alice sends one message m1m_1, a function of xx, to Bob before he responds, and Alice receives nothing. Conditional on λ\lambda, Alice’s output is a function of xx, while Bob’s is a function of yy and m1m_1. Hence EAE_A cannot depend on yy, whereas EBE_B and EABE_{AB} may depend on both inputs. Of the 3×4=123\times4=12 moments, the two equalities

EA(x,1)−EA(x,0)=0,x∈{0,1},(7)E_A(x,1)-E_A(x,0)=0,\qquad x\in\{0,1\}, \tag*{(7)}

are forced, leaving 10 free coordinates, and Theorem 5 below shows that there are no further affine constraints.

The target in which Alice outputs yy and Bob outputs 0 has EA(0,1)−EA(0,0)=−2E_A(0,1)-E_A(0,0)=-2 and so violates Eq. (7) at x=0x=0. The violation identifies the statistic (Alice’s marginal) and the input (yy) that the schedule does not deliver to Alice; a separating inequality would also exclude this target but need not point to yy. In this example the equalities together with positivity describe the class exactly (Proposition 8); this is not true in general, as shown at the end of this section.

Definition 4. A schedule A\mathcal{A} fixes a finite sequence of message events, each with a sender, a set of recipients, and a finite alphabet, together with the time at which each agent’s output is produced. Inputs arrive before any message. Agents remember their input and everything they have received, and may apply any deterministic function to this history. Shared randomness is independent of the inputs. The class PA\mathcal{P}_{\mathcal{A}} is the convex hull of the deterministic behaviours that the schedule can produce.

Definition 4 specifies the schedule but leaves implicit several properties of the protocol family on which the main result depends. We state them separately, because each enters a specific step of the proof of Theorem 5 below and each corresponds to a modelling choice that one may wish to vary.

(i) Outputs are final: no message uses a produced output as an argument, so flipping any agent’s output, with all messages unchanged, gives another protocol of the same schedule. This is used to isolate individual moments.

(ii) Reachability is structural: which inputs can reach which outputs is read off the schedule, not off the functions a particular protocol uses, and a protocol may ignore what it receives. This makes the equalities depend on the schedule alone.

(iii) Messages are free: each message may be any function of the sender’s history, and each alphabet has at least two noiseless symbols that the sender can choose between. This is used to transmit the one-bit flags in the spanning argument.

(iv) Timing and routing are fixed: who speaks, to whom, and when do not depend on the inputs, directly or through earlier messages, so the presence or absence of a message carries no information.

All four properties hold for the schedules of § 4. They concern the affine hull only; two schedules with the same reachability have the same affine hull but can have different polytopes.

Each schedule A\mathcal{A} determines a directed acyclic graph, its time-ordered information graph, with one node for each input, message event, and output, and with edges representing memory and message delivery (Figure 1). Input xjx_j reaches output ii if this graph contains a directed path from xjx_j to output ii. We denote by RA({i})R_{\mathcal{A}}(\{i\}) the set of such jj and set RA(S)=⋃i∈SRA({i})R_{\mathcal{A}}(S)=\bigcup_{i\in S}R_{\mathcal{A}}(\{i\}). A message that arrives after an agent’s output has been fixed does not reach that output. We write DA(K)D_{\mathcal{A}}(K) for Eq. (5) with R=RAR=R_{\mathcal{A}}.

Figure 1. Time-unrolled two-bit relay of § 4.3, with m1=m1(x)m_1=m_1(x) and m2=m2(y,m1)m_2=m_2(y,m_1). Grey lines are memory and red arrows are messages. Each output’s reach is shown on the right; the reach of a set of agents is the union, so EACE_{AC} may depend on {x,y,z}\{x,y,z\} while EAE_A may depend only on xx.

For example, consider three agents with inputs xx, yy, zz and the relay schedule of Figure 1: Alice sends m1(x)m_1(x) to Bob, Bob then sends m2(y,m1)m_2(y,m_1) to Charlie, and all outputs are produced after these messages. Labelling inputs by name, Alice’s output is reached only by xx, Bob’s by {x,y}\{x,y\} through m1m_1, and Charlie’s by {x,y,z}\{x,y,z\} through the chain x→m1→m2x\to m_1\to m_2. Reach sets of groups are unions, so RA({A,C})={x,y,z}R_{\mathcal{A}}(\{A,C\})=\{x,y,z\}: the joint statistic EACE_{AC} may depend on all three inputs, although EAE_A may depend only on xx. Timing matters: if Bob produced his output before m1m_1 arrived, his reach would shrink to {y}\{y\}. Summing 2∣R(S)∣2^{|R(S)|} over the seven sets SS gives DA(3)=2+4+8+4+8+8+8=42D_{\mathcal{A}}(3)=2+4+8+4+8+8+8=42, the value listed for this schedule in Table 3.

ScheduleAABBCCABABACACBCBCABCABCDimensionVertices
No messagesxxyyzzxyxyxzxzyzyzxyzxyz2664
One broadcastxxxyxyxzxzxyxyxzxzxyzxyzxyzxyz341024
Two broadcastsxyxyxyxyxyzxyzxyxyxyzxyzxyzxyzxyzxyz4465536
One relay bitxxxyxyzzxyxyxzxzxyzxyzxyzxyz32256
Two relay bitsxxxyxyxyzxyzxyxyxyzxyzxyzxyzxyzxyz425632

Table 3. Reach sets, affine dimensions, and numbers of distinct deterministic behaviours for three-agent schedules. A string such as xyxy denotes the input set {x,y}\{x,y\}.

Theorem 5. Let A\mathcal{A} satisfy Definition 4 and assumptions (i)–(iv). Every behaviour in PA\mathcal{P}_{\mathcal{A}} satisfies

ES(x)=ES(x′)ifxRA(S)=xRA(S)′(8)E_S(x)=E_S(x')\quad\text{if}\quad x_{R_{\mathcal{A}}(S)}=x'_{R_{\mathcal{A}}(S)} \tag*{(8)}

and these are the only affine constraints. In moment coordinates, the affine hull of PA\mathcal{P}_{\mathcal{A}} is the set of moment vectors in which each ESE_S is an arbitrary function of xRA(S)x_{R_{\mathcal{A}}(S)}, and

dim⁡PA=DA(N)=∑S≠∅∏j∈RA(S)dj.(9)\dim\mathcal{P}_{\mathcal{A}}=D_{\mathcal{A}}(N)=\sum_{S\ne\emptyset}\prod_{j\in R_{\mathcal{A}}(S)}d_j. \tag*{(9)}

Points of this affine hull need not be valid probability distributions.

Proof. Conditional on λ\lambda, each output depends only on its ancestors, so the product over SS depends only on xRA(S)x_{R_{\mathcal{A}}(S)}; input-independent mixtures preserve Eq. (8).

For the converse, flipping output signs preserves the schedule by assumption (i). Averaging over all flips gives PunifP_{\mathrm{unif}}, whose moment vector is zero, so the affine hull is the linear span of the attainable moment vectors. If FϵF_\epsilon flips outputs according to ϵ∈{±1}N\epsilon\in\{\pm1\}^N, it multiplies ETE_T by ∏i∈Tϵi\prod_{i\in T}\epsilon_i, and the signed average

LS=2−N∑ϵ(∏i∈Sϵi)Fϵ(10)L_S=2^{-N}\sum_{\epsilon}\left(\prod_{i\in S}\epsilon_i\right)F_\epsilon \tag*{(10)}

keeps the moments ESE_S and removes all others by character orthogonality, so each SS can be treated separately. Fix SS and an assignment rr of the inputs in RA(S)R_{\mathcal{A}}(S). Give input jj the flag hj=1{xj=rj}h_j=\mathbf{1}\{x_j=r_j\} if j∈RA(S)j\in R_{\mathcal{A}}(S) and hj=1h_j=1 otherwise. At each message event the sender transmits the AND of its own flag and all flags received so far, which assumption (iii) permits. By induction over the schedule, each message is the AND of the flags of its ancestral inputs, and binary symbols suffice even for broadcasts. At output i∈Si\in S, let gig_i be the AND of its own and received flags, so gi=∏j∈RA({i})hjg_i=\prod_{j\in R_{\mathcal{A}}(\{i\})}h_j and

∏i∈Sgi=1{xRA(S)=r}.(11)\prod_{i\in S}g_i=\mathbf{1}\{x_{R_{\mathcal{A}}(S)}=r\}. \tag*{(11)}

For each T⊆ST\subseteq S, let the agents in TT output the sign 2gi−12g_i-1 and all others +1+1; the resulting SS-moment is ∏i∈T(2gi−1)\prod_{i\in T}(2g_i-1), and

∏i∈Sgi=2−∣S∣∑T⊆S∏i∈T(2gi−1)(12)\prod_{i\in S}g_i=2^{-|S|}\sum_{T\subseteq S}\prod_{i\in T}(2g_i-1) \tag*{(12)}

places each indicator (11) in the span, from which LSL_S isolates it. These indicators span all functions of xRA(S)x_{R_{\mathcal{A}}(S)}, and summing over SS gives Eq. (9). □

Remark 6. One might expect a narrow channel to narrow the affine hull, since a single binary message cannot transmit an input alphabet with more than two symbols. Theorem 5 shows otherwise: one binary symbol per message event already gives each statistic its full reach. The spanning argument uses many protocols, each transmitting only the one-bit flag for a single assignment rr, and the resulting indicators span all functions of xRA(S)x_{R_{\mathcal{A}}(S)}, although no single protocol carries the whole input. Capacity therefore restricts the polytope but not its affine

hull, and schedules with equal reachability but different alphabets can simulate different behaviours. The ternary example below and the two-bit relay of § 4.3 are instances.

Corollary 7. Fix SS, write R=RA(S)R=R_A(S) and x=(xR,xRc)x=(x_R,x_{R^c}), and choose a reference value xRc∘x_{R^c}^{\circ} of the unreachable inputs. The equalities

ES(xR,xRc)=ES(xR,xRc∘),xRc≠xRc∘,(13)E_S(x_R,x_{R^c})=E_S(x_R,x_{R^c}^{\circ}),\qquad x_{R^c}\ne x_{R^c}^{\circ}, \tag*{(13)}

over all SS and xRx_R, are independent and complete, and there are Dfull−DA(N)D_{\mathrm{full}}-D_A(N) of them. Keeping only moments of order at most KK, the image of PA\mathcal{P}_A has dimension DA(K)D_A(K).

Proof. Each equality involves a distinct non-reference coordinate, so the equalities are independent; their number equals the codimension in Theorem 5, so they are complete. Projection onto the moments with ∣S∣≤K|S|\le K maps the affine hull onto the sum of the retained function spaces.

For the local schedule R(S)=SR(S)=S, and Eq. (9) reproduces D0(N)D_0(N). If every input reaches every output, the dimension is DfullD_{\mathrm{full}}, which does not imply that every behaviour is achievable. In the example above, R({A})={x}R(\{A\})=\{x\} and R({B})=R({A,B})={x,y}R(\{B\})=R(\{A,B\})=\{x,y\}, giving 2+4+4=102+4+4=10, and Eq. (13) with reference y=0y=0 is Eq. (7).

Let x,y∈{0,1,2}x,y\in\{0,1,2\} and allow one message from Alice to Bob. Binary and ternary message alphabets both give dimension 3+9+9=213+9+9=21. Nevertheless, the target

a=0,b=1{x=y}(14)a=0,\qquad b=1\{x=y\} \tag*{(14)}

requires three message symbols for exact simulation. With two symbols, two values of xx share a message, so Bob’s response rows for these values coincide, whereas the target rows differ; a mixture that reproduces a deterministic behaviour must reproduce it in every positive-weight component, so shared randomness does not help. This is a one-way instance of the equality function, a standard example in communication complexity [23].

Convex certificates

Choose a coordinate map and let uλu_\lambda be the image of each distinct deterministic behaviour of a declared schedule. A target with moment vector mm belongs to the class exactly when

m=∑λμλuλ,∑λμλ=1,μλ≥0.(15)m=\sum_\lambda\mu_\lambda u_\lambda,\qquad \sum_\lambda\mu_\lambda=1,\qquad\mu_\lambda\ge0. \tag*{(15)}

The weights give a simulation in which shared randomness selects protocol λ\lambda with probability μλ\mu_\lambda; in projected coordinates they certify only the retained statistics. Otherwise, linear-programming duality gives a separating inequality [2],

cTuλ≤β∀λ,cTm>β.(16)c^{\mathsf T}u_\lambda\le\beta\quad\forall\lambda,\qquad c^{\mathsf T}m>\beta. \tag*{(16)}

An equality of Corollary 7, written wTuλ=0w^{\mathsf T}u_\lambda=0 for all λ\lambda, that the target violates is a separator with β=0\beta=0 after choosing the sign of ww; its contexts identify a dependence on an unreachable input. Figure 2 summarizes the procedure.

Certification procedure flowchart

Figure 2. Certification procedure. Passing the affine equalities permits a convex membership test but does not imply simulation. A decomposition in complete coordinates certifies the full behaviour; after projection it certifies only the retained moments.

For a noisy interpolation mη=(1−η)m+ηmrefm_\eta=(1-\eta)m+\eta m_{\mathrm{ref}} with cTmref≤βc^{\mathsf T}m_{\mathrm{ref}}\le\beta, the witness remains violated exactly for 0≤η<η⋆0\le\eta<\eta^\star, where

η⋆=cTm−βcTm−cTmref.(17)\eta^\star=\frac{c^{\mathsf T}m-\beta}{c^{\mathsf T}m-c^{\mathsf T}m_{\mathrm{ref}}}. \tag*{(17)}

This is the robustness of that certificate, not necessarily the exact membership threshold; an equality violation diluted toward an equality-compatible reference remains nonzero for every η<1\eta<1.

For estimated moments m^\widehat{m} with estimated covariance Σ^\widehat{\Sigma}, a standardized margin is

ζ=cTm^−βcTΣ^c,cTΣ^c>0.(18)\zeta=\frac{c^{\mathsf T}\widehat{m}-\beta}{\sqrt{c^{\mathsf T}\widehat{\Sigma}c}},\qquad c^{\mathsf T}\widehat{\Sigma}c>0. \tag*{(18)}

It is defined only when the estimated variance is positive and is a signal-to-noise ratio rather than a calibrated significance level. For a mean E^\widehat{E} of nn independent signed products at a fixed context, Hoeffding’s inequality [24] gives Pr⁡{∣E^−E∣≥ε}≤2exp⁡(−nε2/2)\Pr\{|\widehat{E}-E|\ge\varepsilon\}\le2\exp(-n\varepsilon^2/2), so, for two contexts with equal true moments a union bound gives, with probability at least 1−α1-\alpha,

∣E^−E^′∣≤2log⁡(4/α)n+2log⁡(4/α)n′.(19)|\widehat{E}-\widehat{E}'|\le\sqrt{\frac{2\log(4/\alpha)}{n}}+\sqrt{\frac{2\log(4/\alpha)}{n'}}. \tag*{(19)}

Testing many equalities, or choosing a witness on the same data, requires multiplicity control or independent validation data, and serially dependent trials require a different analysis. Positive rescaling of (c,β)(c,\beta) leaves ζ\zeta and η⋆\eta^\star unchanged. Minimizing ∥c∥1\lVert c\rVert_1 at fixed margin is one way to obtain sparser witnesses, although it does not guarantee fewer settings or a larger ζ\zeta.

Communication hierarchies

All inputs and outputs in this section are binary. A “round” is one prescribed message event, not a simultaneous distributed round, and every level can ignore any message, so the levels are nested.

Canonical targets

For two agents, define

B=EAB(00)+EAB(01)+EAB(10)−EAB(11).(20)\mathcal{B}=E_{AB}(00)+E_{AB}(01)+E_{AB}(10)-E_{AB}(11). \tag*{(20)}

Local behaviours satisfy ∣B∣≤2|\mathcal{B}|\le2 [1], and in this scenario the eight relabelled CHSH inequalities together with positivity characterize the local polytope [2, 21]. The PR behaviour PPR(ab∣xy)=121{a⊕b=xy}P_{\mathrm{PR}}(ab\mid xy)=\frac{1}{2}\mathbf{1}\{a\oplus b=xy\} [25] has zero marginals, EAB(xy)=(−1)xyE_{AB}(xy)=(-1)^{xy}, and B=4\mathcal{B}=4, above the quantum bound 222\sqrt{2} [26]. The family PPR,v=vPPR+(1−v)PunifP_{\mathrm{PR},v}=vP_{\mathrm{PR}}+(1-v)P_{\mathrm{unif}} has correlators v(−1)xyv(-1)^{xy} and largest CHSH value 4v4v, so it is local exactly for v≤1/2v\le1/2. It is quantum-realizable exactly for v≤1/2v\le1/\sqrt{2}, being the mixture 2vPTs+(1−2v)Punif\sqrt{2}vP_{\mathrm{Ts}}+(1-\sqrt{2}v)P_{\mathrm{unif}} for such vv, where the Tsirelson target PTsP_{\mathrm{Ts}} has zero marginals and EAB(xy)=(−1)xy/2E_{AB}(xy)=(-1)^{xy}/\sqrt{2}. The latter is realized on ∣Φ+⟩=(∣00⟩+∣11⟩)/2|\Phi^+\rangle=(|00\rangle+|11\rangle)/\sqrt{2} with Alice measuring σz\sigma_z or σx\sigma_x and Bob measuring (σz±σx)/2(\sigma_z\pm\sigma_x)/\sqrt{2}.

Figure 3 shows the projection onto B\mathcal{B} and B′=EAB(00)+EAB(01)−EAB(10)+EAB(11)\mathcal{B}'=E_{AB}(00)+E_{AB}(01)-E_{AB}(10)+E_{AB}(11). The projections of the local, quantum, and no-signalling sets are

Figure 3. Images in the (B,B′)(\mathcal{B},\mathcal{B}') plane: the local square, quantum disk, and no-signalling diamond. The dashed line is the witness B=2\mathcal{B}=2. The PR, Tsirelson, and noisy-PR points lie on B′=0\mathcal{B}'=0.

∣B∣,∣B′∣≤2,|\mathcal{B}|,|\mathcal{B}'|\le2,
B2+B′2≤8,\mathcal{B}^{2}+\mathcal{B}'^{2}\le8,
∣B±B′∣≤4.(21)|\mathcal{B}\pm\mathcal{B}'|\le4. \tag*{(21)}

At a local vertex with signed tables A(x)A(x), B(y)B(y), either B(0)=B(1)B(0)=B(1) and B=B′=2A(0)B(0)\mathcal{B}=\mathcal{B}'=2A(0)B(0), or B(0)=−B(1)B(0)=-B(1) and B=−B′=2A(1)B(0)\mathcal{B}=-\mathcal{B}'=2A(1)B(0), so the local image is the square with corners (±2,±2)(\pm2,\pm2). With U=E00+E01U=E_{00}+E_{01} and V=E10−E11V=E_{10}-E_{11}, the no-signalling image is ∣U∣,∣V∣≤2|U|,|V|\le2, attained by zero-marginal correlators. The unit-vector representation Exy=sx⋅tyE_{xy}=s_x\cdot t_y of quantum correlators [26] gives U2+V2≤∥t0+t1∥2+∥t0−t1∥2=4U^2+V^2\le\lVert t_0+t_1\rVert^2+\lVert t_0-t_1\rVert^2=4, with equality attained by coplanar Pauli measurements on a maximally entangled pair; convex mixing fills the disk. Membership in a projected set does not establish membership of the unprojected behaviour.

For three agents, Mermin’s expression is

M=EABC(000)−EABC(011)−EABC(101)−EABC(110),(22)\mathcal{M}=E_{ABC}(000)-E_{ABC}(011)-E_{ABC}(101)-E_{ABC}(110), \tag*{(22)}

and local behaviours satisfy ∣M∣≤2|\mathcal{M}|\le2 [27]. The state ∣GHZ3⟩=(∣000⟩+∣111⟩)/2|\mathrm{GHZ}_3\rangle=(|000\rangle+|111\rangle)/\sqrt{2} [28], with input 0 denoting σx\sigma_x and input 1 denoting σy\sigma_y, has vanishing one- and two-body moments and

EABC(x,y,z)=cos⁡[π2(x+y+z)],(23)E_{ABC}(x,y,z)=\cos\left[\frac{\pi}{2}(x+y+z)\right], \tag*{(23)}

so M=4\mathcal{M}=4. The family PGHZ,v=vPGHZ+(1−v)PunifP_{\mathrm{GHZ},v}=vP_{\mathrm{GHZ}}+(1-v)P_{\mathrm{unif}} is local exactly for v≤1/2v\le1/2. Necessity follows from M(PGHZ,v)=4v\mathcal{M}(P_{\mathrm{GHZ},v})=4v. For sufficiency, the weights μλ=[1+M(λ)/2]/64\mu_\lambda=[1+\mathcal{M}(\lambda)/2]/64 on the 64 local vertices are nonnegative by Mermin’s bound and sum to one because M\mathcal{M} averages to zero; since distinct monomials in the signed responses are orthogonal under the uniform distribution on vertices, the mixture has half of each nonzero GHZ moment and zero for every other moment.

Two agents

Let Pk(2)\mathcal{P}^{(2)}_k be the two-agent class with Alice speaking first:

k=0:a=a(x),b=b(y),k=1:m1=m1(x),a=a(x),b=b(y,m1),k=2:m2=m2(y,m1),a=a(x,m2),b=b(y,m1).(24)\begin{aligned} k=0:\quad&a=a(x),\qquad b=b(y),\\ k=1:\quad&m_1=m_1(x),\qquad a=a(x),\qquad b=b(y,m_1),\\ k=2:\quad&m_2=m_2(y,m_1),\\ &\qquad a=a(x,m_2),\qquad b=b(y,m_1). \tag*{(24)} \end{aligned}

We retain EAE_A, EBE_B, and EABE_{AB} at all four contexts, giving twelve coordinates without a no-signalling assumption.

Proposition 8. The hierarchy is strict, P0(2)⊊P1(2)⊊P2(2)\mathcal{P}^{(2)}_0\subsetneq\mathcal{P}^{(2)}_1\subsetneq\mathcal{P}^{(2)}_2, with 16, 64, and 256 vertices and affine dimensions 8, 10, and 12. The one-message class consists exactly of the behaviours whose Alice marginal is independent of yy; the two-message class is the full normalized behaviour polytope.

Proof. With m1=xm_1=x, bb is an arbitrary function of (x,y)(x,y), and with m1=xm_1=x, m2=ym_2=y both outputs are; this gives the vertex counts, the k=1k=1 count agreeing with the one-bit vertex formula of Ref. [5]. The dimensions follow from Theorem 5 with reach sets ({x},{y},{x,y})\left(\{x\},\{y\},\{x,y\}\right), ({x},{x,y},{x,y})\left(\{x\},\{x,y\},\{x,y\}\right), and ({x,y},{x,y},{x,y})\left(\{x,y\},\{x,y\},\{x,y\}\right) for (EA,EB,EAB)(E_A,E_B,E_{AB}), and give strictness. For the one-message class, necessity is immediate because Alice’s output depends only on xx and λ\lambda. For sufficiency, let PA(a∣x)=P(a∣x,y)P_A(a\mid x)=P(a\mid x,y) and PB∣A(b∣a,x,y)=P(a,b∣x,y)/PA(a∣x)P_{B\mid A}(b\mid a,x,y)=P(a,b\mid x,y)/P_A(a\mid x) whenever PA(a∣x)>0P_A(a\mid x)>0. Let λ\lambda consist of independent components αx∼PA(⋅∣x)\alpha_x\sim P_A(\cdot\mid x) and βx,y,a∼PB∣A(⋅∣a,x,y)\beta_{x,y,a}\sim P_{B\mid A}(\cdot\mid a,x,y); Alice outputs αx\alpha_x and sends m1=xm_1=x, and Bob, who knows xx and λ\lambda, outputs βx,y,αx\beta_{x,y,\alpha_x}. Each λ\lambda fixes a deterministic k=1k=1 protocol, and

Pr⁡(a,b∣x,y)=PA(a∣x)PB∣A(b∣a,x,y)=P(a,b∣x,y).(25)\Pr(a,b\mid x,y)=P_A(a\mid x)P_{B\mid A}(b\mid a,x,y)=P(a,b\mid x,y). \tag*{(25)}

When PA(a∣x)=0P_A(a\mid x)=0 any distribution may be assigned to βx,y,a\beta_{x,y,a}, since that branch has probability zero and P(a,b∣x,y)=0P(a,b\mid x,y)=0 for every bb. In the two-message class both agents know (x,y)(x,y), and sampling (a,b)(a,b) from P(⋅,⋅∣x,y)P(\cdot,\cdot\mid x,y) with shared randomness yields every behaviour.

The finite-input observation agrees with earlier communication-assisted treatments [3, 5]. It differs from Toner and Bacon’s one-bit simulation of arbitrary projective measurements on a singlet [29]: a binary input is sent here, and no claim is made that one bit transmits an arbitrary finite input alphabet. Figure 4 shows the information flow.

Figure 4. Two-agent hierarchy with Alice speaking first. Choosing the messages to be the speakers’ inputs realizes all deterministic tables with the displayed dependencies; the affine dimensions are 8, 10, and 12.

For PR, a uniform shared bit ss gives the one-message protocol

a=s,m1=x,b=s⊕xy.(26)a=s,\qquad m_1=x,\qquad b=s\oplus xy. \tag*{(26)}

The targets a=0,b=xa=0,b=x and a=y,b=0a=y,b=0 diagnose the direction of information flow, with equality violations

EB(0,0)−EB(1,0)=2,EA(0,0)−EA(0,1)=2.(27)E_B(0,0)-E_B(1,0)=2,\qquad E_A(0,0)-E_A(0,1)=2. \tag*{(27)}

The first difference vanishes on P0(2)\mathcal{P}^{(2)}_0 and the second on both P0(2)\mathcal{P}^{(2)}_0 and P1(2)\mathcal{P}^{(2)}_1 (Table 2). The last target needs two events only because of the prescribed order, so the minimum within a fixed hierarchy is not an intrinsic cost of the target over all protocols.

TargetP0(2)\mathcal{P}^{(2)}_0P1(2)\mathcal{P}^{(2)}_1P2(2)\mathcal{P}^{(2)}_2kmin⁡k_{\min}
UniformYYY0
Noisy PR, v=1/2v=1/2YYY0
Noisy PR, v=7/10v=7/10NYY1
TsirelsonNYY1
PRNYY1
a=0a=0, b=xb=xNYY1
a=ya=y, b=0b=0NNY2

Table 2. Two-agent target membership. kmin⁡k_{\min} counts events in Eq. (24). Noisy PR rows use PPR,v=vPPR+(1−v)PunifP_{\mathrm{PR},v}=vP_{\mathrm{PR}}+(1-v)P_{\mathrm{unif}}.

Three agents: broadcast and relay

In the broadcast hierarchy Pkbc\mathcal{P}^{\mathrm{bc}}_k, one bit is one symbol delivered to both other agents; Alice broadcasts first and Bob second:

k=0:a(x), b(y), c(z),k=1:m1(x);a(x), b(y,m1), c(z,m1),k=2:m2(y,m1);a(x,m2), b(y,m1),c(z,m1,m2).(28)\begin{aligned} k&=0:\quad a(x),\ b(y),\ c(z),\\ k&=1:\quad m_1(x);\quad a(x),\ b(y,m_1),\ c(z,m_1),\\ k&=2:\quad m_2(y,m_1);\quad a(x,m_2),\ b(y,m_1),\\ &\qquad c(z,m_1,m_2). \tag*{(28)} \end{aligned}

In the point-to-point relay hierarchy Pkpt\mathcal{P}^{\mathrm{pt}}_k, the events are Alice to Bob and then Bob to Charlie (Figure 1):

k=0:a(x), b(y), c(z),k=1:m1(x);a(x), b(y,m1), c(z),k=2:m2(y,m1);a(x), b(y,m1), c(z,m2).(29)\begin{aligned} k&=0:\quad a(x),\ b(y),\ c(z),\\ k&=1:\quad m_1(x);\quad a(x),\ b(y,m_1),\ c(z),\\ k&=2:\quad m_2(y,m_1);\\ &\qquad a(x),\ b(y,m_1),\ c(z,m_2). \tag*{(29)} \end{aligned}

Charlie is silent in both families, and all seven moments at all eight contexts are retained.

Proposition 9. Both three-agent hierarchies are strictly nested at levels 0,1,20,1,2, with the dimensions and vertex counts of Table 3, and Pkpt⊊Pkbc\mathcal{P}^{\mathrm{pt}}_k\subsetneq\mathcal{P}^{\mathrm{bc}}_k for k=1,2k=1,2.

Proof. The dimensions follow from Theorem 5. For broadcast, m1=xm_1=x and m2=ym_2=y give arbitrary tables with the indicated dependencies. For two relay bits, a(x)a(x) and b(x,y)b(x,y) are arbitrary, but c(x,y,z)c(x,y,z), viewed as four rows indexed by (x,y)(x,y), has at most two distinct rows, giving 4+(42)(24−2)=884+\binom{4}{2}(2^4-2)=88 Charlie tables and 4⋅16⋅88=56324\cdot16\cdot88=5632 behaviours. Ignoring a received copy of a broadcast emulates the relay, and the unequal dimensions give strictness. □

The relay illustrates the difference between affine dimension and capacity: its 42 dimensions are those of a schedule in which Charlie receives the full label (x,y)(x,y), yet one bit permits only 88 of 256 Charlie tables. The table whose rows are 00, 01, 10, 11 for $(x,y)=00, 01, 10, 11 satisfies every relay equality but is not achievable.

One-bit GHZ simulation. The σx/σy\sigma_x/\sigma_y GHZ target belongs to both one-message classes. That one bit suffices follows from Tessier et al.{?}, who reproduce all Pauli-product correlations of the nn-qubit GHZ state with n−2n-2 bits, optimal for that measurement set [30]. We give a realization within the declared hierarchies. Let s0,s1,s2s_0,s_1,s_2 be independent unbiased shared signs and

\begin{array}{c|cc} G_{xy} & y=0 & y=1\\ \hline x=0 & s_1 & -s_2\\ x=1 & -s_2 & -s_1 \end{array} \qquad C(0)=s_1,\qquad C(1)=s_2. \tag*{(30)}\end{array}

Alice outputs s0s_0 and sends xx to Bob, Bob outputs s0Gxys_0G_{xy}, and Charlie outputs C(z)C(z). Every one- and two-body moment contains an unpaired unbiased sign, and E(GxyC(z))\mathbb{E}(G_{xy}C(z)) equals +1+1 at 000, −1-1 at 011, 101, and 110, and 0 elsewhere, which is Eq. (23).

Further targets. For PR on A,BA,B with c=0c=0, Eq. (26) applies to either one-message schedule. For a=b=0a=b=0, c=x⊕yc=x\oplus y,

EC(0,0,0)−EC(0,1,0)=2,(31)E_C(0,0,0)-E_C(0,1,0)=2, \tag*{(31)}

which vanishes in both one-message classes; two broadcasts send xx and yy, and the two-bit relay sends xx and then x⊕yx\oplus y. Broadcast and relay differ for a=b=0a=b=0, c=xc=x, where one broadcast suffices but two relay bits are needed, and for a=ya=y, b=c=0b=c=0, which the prescribed relay never achieves (Table 4). For a=za=z, b=c=0b=c=0,

TargetP0\mathcal{P}_0P1bc\mathcal{P}_1^{\mathrm{bc}}P2bc\mathcal{P}_2^{\mathrm{bc}}kmin⁡bck_{\min}^{\mathrm{bc}}P1pt\mathcal{P}_1^{\mathrm{pt}}P2pt\mathcal{P}_2^{\mathrm{pt}}kmin⁡ptk_{\min}^{\mathrm{pt}}
UniformYYY0YY0
Noisy GHZ, v=1/2v = 1/2YYY0YY0
GHZ, σx/σy\sigma_x/\sigma_yNYY1YY1
PR on A,BA, B; c=0c = 0NYY1YY1
a=b=0a = b = 0, c=xc = xNYY1NY2
a=b=0a = b = 0, c=x⊕yc = x \oplus yNNY2NY2
a=ya = y, b=c=0b = c = 0NNY2NN–
a=za = z, b=c=0b = c = 0NNN–NN–

Table 4. Three-agent membership and scheduled minimum event counts. A dash means that no level k≤2k \le2 realizes the target. PGHZ,v=vPGHZ+(1−v)PunifP_{\mathrm{GHZ},v} = vP_{\mathrm{GHZ}} + (1-v)P_{\mathrm{unif}}.

EA(0,0,0)−EA(0,0,1)=2(32)E_A(0,0,0)-E_A(0,0,1)=2 \tag*{(32)}

vanishes in every schedule in which Charlie is silent. The certificate identifies the required dependence, information about zz reaching Alice before her output, but not a unique physical channel. If Charlie also broadcasts zz after the two broadcasts of Eq. (28) and all outputs follow this event, every deterministic table becomes achievable and the class is the full product of eight seven-dimensional simplices.

Low-order projections

For N≥2N\ge2, consider ∣GHZN⟩=(∣0⟩⊗N+∣1⟩⊗N)/2\lvert\mathrm{GHZ}_N\rangle=(\lvert0\rangle^{\otimes N}+\lvert1\rangle^{\otimes N})/\sqrt{2} with each agent measuring σx\sigma_x (input 0) or σy\sigma_y (input 1). Every proper reduced state equals the corresponding reduction of the separable mixture ρmax⁡=12(∣0⟩⟨0∣⊗N+∣1⟩⟨1∣⊗N)\rho_{\max}=\frac{1}{2}(\lvert0\rangle\langle0\rvert^{\otimes N}+\lvert1\rangle\langle1\rvert^{\otimes N}), so correlations among fewer than NN parties do not reveal the GHZ coherence. This obstruction, for every order k<Nk<N and without a symmetry restriction, is given in Ref. [10, Sec. VI]; the related Bell-inequality observation appears in Ref. [9]. Since ρmax⁡\rho_{\max} is separable, the order-KK projection of the GHZ behaviour lies in CKC_K for every K<NK<N and every choice of settings.

For σx/σy\sigma_x/\sigma_y settings the moments are explicit. The reductions of ρmax⁡\rho_{\max} are diagonal in the computational basis, whereas nonempty products of σx\sigma_x and σy\sigma_y are off-diagonal, so all proper moments vanish. The GHZ state is an eigenvector of each NN-fold product with an even number nyn_y of σy\sigma_y factors [27], with eigenvalue (−1)ny/2(-1)^{n_y/2}, and products with odd nyn_y map it to an orthogonal state. Hence

ES(xS)=0(∅≠S⊆[N]),E_S(x_S)=0\quad(\varnothing\ne S\subseteq[N]),
E[N](x)=cos⁡(π2∑ixi).(33)E_{[N]}(x)=\cos\left(\frac{\pi}{2}\sum_i x_i\right). \tag*{(33)}

Proposition 10. For every K<NK<N, ΠK(PGHZ)=ΠK(Punif)=0\Pi_K(P_{\mathrm{GHZ}})=\Pi_K(P_{\mathrm{unif}})=0. *This point is the centre of the ℓ1\ell_1 ball of Eq. (6), so every valid inequality cTu≤β\mathbf{c}^{\mathsf{T}}\mathbf{u}\le\beta for CKC_K with c≠0\mathbf{c}\ne0 has slack at least ∥c∥∞\lVert\mathbf{c}\rVert_\infty there.

Proof. The retained GHZ moments vanish by Eq. (33), the moments of PunifP_{\mathrm{unif}} vanish, and Lemma 3 gives the remaining statements.

The qualitative content of Proposition 10 is the obstruction of Ref. [10]. The proposition adds an explicit moment-space proof for σx/σy\sigma_x/\sigma_y settings and a quantitative margin: every valid inequality has slack at least ∥c∥∞\lVert\mathbf{c}\rVert_\infty at the common image, and a Euclidean ball of radius 1/D0(K)1/\sqrt{D_0(K)} lies inside CKC_K. At N=3N=3, the target violates Mermin’s inequality maximally but lies at the origin of the 18-dimensional second-order projection (Figure 5), with guaranteed Euclidean radius 1/181/\sqrt{18}; this radius is a lower bound, not the inradius.

Distinct complete behaviours with identical one- and two-body moments

Figure 5. Distinct complete behaviours with identical one- and two-body moments. The ellipse schematically represents a neighbourhood of the origin in an 18-dimensional polytope; the three-body moments distinguish the two sources.

Definition 11. A no-signalling target PP is KK-body detectable if some valid inequality for the local model supported on moments of order at most KK is violated by PP, equivalently if ΠK(P)∉CK\Pi_K(P) \notin C_K.

Detectability is decided by the membership program (15), and detectability at order KK implies detectability at every larger order. The three-agent σx/σy\sigma_x/\sigma_y GHZ target requires K=3K = 3, and at N=2N = 2 the σx/σy\sigma_x/\sigma_y GHZ behaviour is itself local. The quantitative statement depends on the settings: measurements involving σz\sigma_z can give nonzero proper moments, which still agree with those of ρmix\rho_{\mathrm{mix}} but need not lie at an interior point. Low-order Bell tests do detect other entangled states [11]. A classical analogue shows the same loss: if independent unbiased bits aa, bb determine c=a⊕bc = a \oplus b, every pair is independent while a⊕b⊕c=0a \oplus b \oplus c = 0 exactly.

Agents and observers

Local decision rules

In decision terms, xix_i is the information available to agent ii and aia_i its choice; a behaviour is local if it can be explained by information shared before the task, without exchange during it. Consider a binary response rule motivated by quantum decision theory,

pi(1∣xi,λ)=fi(xi,λ)+κqi(xi,λ)∈[0,1],(34)p_i(1 \mid x_i,\lambda) = f_i(x_i,\lambda) + \kappa q_i(x_i,\lambda) \in[0,1], \tag*{(34)}

where the utility and attraction decomposition follows the decision-theory literature [12, 13, 14], and κ\kappa is an interpolation parameter introduced here, not a parameter claimed for those models. A value is admissible if every response probability lies in [0,1][0,1].

Corollary 12. If each response uses only its own input and input-independent shared randomness, and residual random choices are conditionally independent given that randomness, the joint behaviour lies in P0\mathcal{P}_0 for every admissible κ\kappa, and its communication cost in the present resource model is zero.

Under these assumptions an interference term can move the behaviour within the local polytope but cannot violate a Bell inequality. Yukalov and Sornette describe prospect operators as entangling because of the composite structure of a decision maker’s action prospects [12]; this does not by itself identify the Hilbert-space factors with separate agents holding independently chosen inputs, and an entangled representation alone does not imply Bell nonlocality [2]. A distributed model that specifies a Bell-nonlocal joint law must account for a shared quantum resource, joint access to private inputs, or information exchange; network models with repeated exchange [15] correspond to a different schedule. The same applies to the classical networks of Refs. [16, 17]: there the initial single-step choices are made before any exchange, whereas each later step updates a group’s choice probability using the other groups’ probabilities, which in the present terms is a message event. Those models describe aggregate choice fractions without setting choices, so the behaviour-level certificates of § 3.1 would apply only after inputs are introduced.

A complementary construction corrects the joint distribution rather than the individual rules. For two binary responses, initially independent with common marginal (ν,1−ν)(\nu,1-\nu), 0<ν<10 < \nu< 1, add δ\delta to the diagonal and subtract it from the off-diagonal entries:

Pδ=(ν2+δ, ν(1−ν)−δ, ν(1−ν)−δ, (1−ν)2+δ)(35)P_{\delta} = \left( \nu^2+\delta,\ \nu(1-\nu)-\delta,\ \nu(1-\nu)-\delta,\ (1-\nu)^2+\delta \right) \tag*{(35)}

in the order 00, 01, 10, 11. Both marginals remain (ν,1−ν)(\nu,1-\nu), and for these marginals Eq. (35) is the general joint distribution, with nonnegativity requiring −min⁡{ν2,(1−ν)2}≤δ≤ν(1−ν)-\min\{\nu^2,(1-\nu)^2\}\leq\delta\leq\nu(1-\nu). In signed moments, EA=EB=2ν−1E_A=E_B=2\nu-1 and EAB=(2ν−1)2+4δE_{AB}=(2\nu-1)^2+4\delta, so only the connected two-body moment changes, and the Pearson correlation coefficient is δ/[ν(1−ν)]\delta/[\nu(1-\nu)]. At δ=ν(1−ν)\delta=\nu(1-\nu) the outputs agree on every trial and their mutual information equals the binary entropy H2(ν)H_2(\nu). Using the same PδP_\delta at every context, the shared variable λ=(λA,λB)∼Pδ\lambda=(\lambda_A,\lambda_B)\sim P_\delta, with Alice returning λA\lambda_A and Bob returning λB\lambda_B, gives a decomposition of the form of Eq. (2) for every admissible δ\delta. In Corollary 12, dependence between outputs can enter only through λ\lambda, and rules that do not depend on λ\lambda yield product behaviours for every κ\kappa; the joint correction instead fixes the marginals and changes only the dependence, which shared randomness suffices to produce. Identifying δ\delta with an attraction or interference factor of a specific decision-theory formulation would require a derivation within that formulation, which we do not supply; the example is an illustrative composite decision model. It also shows that a nonfactorized description need not imply a communication cost, since locality requires only the existence of some decomposition of the form of Eq. (2).

Extended Wigner’s friend scenarios

Extended Wigner’s friend scenarios [18, 19, 20, 31] declare which agents measure, which records pass to whom, and when. Here the behaviour consists of the superobservers’ recorded outputs given their settings.

Observer-independent facts. Brukner shows that locality, freedom of choice, and observer-independent facts imply a joint distribution over the outcomes of both settings at each site, which the quantum prediction contradicts through a CHSH violation [18]. In the present terms the joint distribution places the behaviour in P0\mathcal{P}_0 [21], and the result is a separation by Eq. (20); see Refs. [32, 33, 34] for an experimental test and for later critiques.

Local Friendliness. Local Friendliness (LF) combines Absoluteness of Observed Events with Local Agency [20, 34]. In an LF model the friends’ outcomes act like the shared variable λ\lambda of Eq. (2), and LF behaviours form a polytope between P0\mathcal{P}_0 and NS\mathcal{N}\mathcal{S} that equals P0\mathcal{P}_0 in the canonical multipartite scenarios identified in Ref. [34]. Table 5 compares it with the declared classes, and three consequences follow.

LF polytope [34]PA\mathcal{P}_{\mathcal{A}} (this work)
GeneratorsQuery outputs equal the friends’ recorded outcomes; remaining outputs follow a no-signalling extreme point on the non-query contexts [34 Thm. 8]Deterministic protocols of the declared schedule (Definition 4)
Affine equalitiesNormalization and no-signalling; dim⁡LF=dim⁡P0=dim⁡NS\dim\mathrm{LF}=\dim\mathcal{P}_0=\dim\mathcal{N}\mathcal{S} [34 Cor. 12]No-dependence equalities, Eq. (8); complete by Theorem 5
AssumptionsAbsoluteness of Observed Events and Local Agency; space-like separated superobservers; query input reads the friend’s record; no communicationFixed acyclic schedule; finite noiseless alphabets; input-independent timing and routing; shared randomness independent of inputs

Table 5. Comparison of the LF polytope with the declared classes. P0\mathcal{P}_0 and NS\mathcal{N}\mathcal{S} denote the Bell-local and no-signalling polytopes.

First, every LF polytope shares the affine hull of P0\mathcal{P}_0 and NS\mathcal{N}\mathcal{S} [8, 34], so an LF-violating no-signalling behaviour satisfies every affine equality valid on it and can be certified only by an inequality.

only by an inequality. A violated no-signalling equality excludes LF through Local Agency and names the remote setting involved, but says nothing about Absoluteness of Observed Events.

Second, the conditional-sampling construction in the proof of Proposition 8 applies to any finite number of settings when the message alphabet has one symbol per setting of Alice. Every no-signalling behaviour, LF-violating or not, is then reproduced by one message from Alice to Bob; in the scenario with two superobservers and three settings each, where LF inequalities are violated [20, 34], one ternary message suffices.

Third, for three superobservers with a friend and two settings each, the LF polytope equals P0\mathcal{P}_0 [34], so Mermin’s inequality is an LF inequality that the oX/oYo_X/o_Y GHZ behaviour violates when the friends’ measurements are modelled unitarily [34]. The noisy GHZ family is then LF-compatible exactly for ν≤1/2\nu\leq1/2, and by Proposition 10 no LF inequality of order at most two detects the violation. Because P0=LF\mathcal{P}_0=\mathrm{LF}, the behaviours of Corollary 12 also satisfy every LF inequality.

Scenarios without settings. The Frauchiger–Renner protocol [19] involves no setting choices, so its announced results have a single context and are reproduced by shared randomness; none of the present certificates applies.

Possibilistic arguments. Hardy-type arguments [35, 36, 37] have a convex counterpart for the declared classes. Write P(e)P(e) for P(a∣x)P(a\mid x) at e=(a,x)e=(a,x). If a target assigns probability zero to every ee in a set ZZ and positive probability to e∗e^\ast, and every deterministic protocol of the class that assigns zero to ZZ also assigns zero to e∗e^\ast, then every protocol satisfies

P(e⋆)≤∑e∈ZP(e),(36)P(e^\star) \le\sum_{e\in Z} P(e), \tag*{(36)}

because its probabilities are 0 or 1, and the target violates this valid inequality. The argument requires deterministic vertices and so does not apply directly to the LF polytope [34].

Related reformulations and extensions of LF appear in Refs. [38, 39, 40].

Discussion

The convex certificate of Eq. (15) is the standard Bell-polytope construction [2]. Communication-assisted polytope methods already distinguish fixed direction, input alphabets, and message budgets [3, 4, 5], and the two-agent counts and input-transmission protocol of § 4.2 are elementary instances of that framework. Pironio showed that the violation of a Bell inequality lower-bounds the average communication needed to reproduce a behaviour, and that each behaviour has an optimal inequality for which that communication also suffices [6]; average cost is a different resource measure from the fixed event budgets used here. Toner and Bacon address a more extensive measurement family on an entangled pair [29], and Tessier et al. give the (n−2)(n-2)-bit model for Pauli GHZ correlations [30].

Other generalizations of the local polytope relax different assumptions. Partially deterministic polytopes [41], which include the LF polytopes of § 6.2, predetermine the outputs of only some inputs at each site and therefore lie between P0P_0 and NS\mathcal{NS}; a declared schedule instead restricts which inputs can reach which outputs and, once an input reaches a remote output, extends beyond NS\mathcal{NS}. Multipartite causal-correlation polytopes allow more general causal orders [7], and causal-structure analysis distinguishes models whose observable constraints are exhausted by equalities from models that also require inequalities [42]. Theorem 5 concerns a narrower model, a fixed acyclic schedule with one shared source, arbitrary local functions, and binary outputs, and provides an explicit equality basis for it; we do not claim that equality constraints or the diagnostic use of no-signalling are new in general. Conditional-independence methods in causal discovery also relate observed distributions to graph structure under stated distributional and faithfulness conditions [43]. Here the schedule is supplied rather than discovered, and the equalities hold for every admissible distribution without a Gaussian assumption, although independence of the inputs from the shared randomness, timing, retained histories, and the allowed local functions are part of the model. Applied to every T⊆ST \subseteq S, Eq. (8) gives

P(aS∣x)=P(aS∣xRA(S)),(37)P(a_S \mid x) = P(a_S \mid x_{R_A(S)}), \tag*{(37)}

the conditional independence of aSa_S from the unreachable inputs given the reachable ones; a single moment equality generally tests less than this.

The non-obvious content of Theorem 5 is Remark 6: the affine hull depends on reachability alone, so capacity limits, which decide simulability, can be certified only by inequalities. Several questions remain open. Theorem 5 determines affine hulls but not facets, and the examples of § 3 and § 4.3 show that capacity constraints can make a declared class strictly smaller than the set of normalized nonnegative behaviours satisfying its equalities. A facet description of declared classes, and the extension of Theorem 5 to communication-assisted LF classes in which the space-like separation of Local Agency is replaced by a declared schedule, are left for future work.

Acknowledgments

We thank V. I. YukaIov for helpful comments.

References

Worked example

This appendix applies the main results to one three-agent task. It illustrates that a violated equality identifies a dependence on an input that cannot reach the agents involved, whereas a target that satisfies every equality, such as one limited only by message capacity, can be excluded only by an inequality.

Alice and Bob hold bits xx and yy, and Charlie’s bit zz selects which of them matters. With u=(u0,u1)=(x,y)u=(u_0,u_1)=(x,y), the answer and its sign are

uz=x⊕z(x⊕y),T(x,y,z)=(−1)uz.(38)u_z=x\oplus z(x\oplus y),\qquad T(x,y,z)=(-1)^{u_z}. \tag*{(38)}

In the direct readout Charlie reports the answer: a=b=0a=b=0, c=uzc=u_z. In the parity readout the group reports it: a⊕b⊕c=uza\oplus b\oplus c=u_z, with aa and bb uniform and independent. In PvdirP_v^{\mathrm{dir}}, Charlie’s output is replaced with probability 1−v1-v by a fresh unbiased bit. In PvparP_v^{\mathrm{par}}, the whole behaviour is replaced with probability 1−v1-v by PunifP_{\mathrm{unif}}.

We show three things. Equalities exclude the direct readout whenever xx or yy fails to reach Charlie. The parity readout passes every equality and is excluded only by inequalities. Once all inputs reach Charlie, the size of the last message alone decides the direct readout. Table 6 summarizes the results; in the schedule “Broadcast of zz by C”, Charlie sends zz to both others before any output is produced.

ScheduleRA({C})R_A(\{C\})Dim.vdir⋆v_{\mathrm{dir}}^\starvpar⋆v_{\mathrm{par}}^\star
No messageszz2601/2
One relay bitzz3201/2
One broadcast by Axzxz3401/2
Broadcast of zz by Czz3401
Two relay messagesxyzxyz42q/4q/41
Two broadcastsxyzxyz4411

Table 6. The selection task. v⋆v^\star is the largest vv for which the noisy target lies in the class. Zero entries are certified by Eq. (39), entries 1/21/2 by Ipar≤4\mathcal{I}_{\mathrm{par}}\le4, and q/4q/4 by Eq. (41), where m2m_2 has q∈{2,3,4}q\in\{2,3,4\} symbols.

Write A=(−1)aA=(-1)^a, B=(−1)bB=(-1)^b, and C=(−1)cC=(-1)^c for the signed outputs. Eq. (4) gives

Pvdir=18(1+A)(1+B)(1+vTC),Pvpar=18(1+vTABC).P_v^{\mathrm{dir}}=\frac{1}{8}(1+A)(1+B)(1+vTC),\qquad P_v^{\mathrm{par}}=\frac{1}{8}(1+vTABC).

The direct readout has EA=EB=EAB=1E_A=E_B=E_{AB}=1 and ES=vTE_S=vT for every S∋CS\ni C. Charlie’s marginal therefore depends on xx and $y, and the behaviour signals. The parity readout has EABC=vTE_{ABC}=vT and all other moments zero, so it is no-signalling. Flipping the outputs by a shared uniform sign pattern ϵ\epsilon with ϵAϵBϵC=1\epsilon_A\epsilon_B\epsilon_C=1 is permitted by assumption (i) and shared randomness. It removes every moment of PvdirP_v^{\mathrm{dir}} except EABCE_{ABC} and yields PvparP_v^{\mathrm{par}}. Every class that contains the direct readout therefore contains the parity readout.

The direct readout fails at the equality step. Its moments satisfy

EC(0,0,0)−EC(1,0,0)=2v,E_C(0,0,0)-E_C(1,0,0)=2v,
EC(0,0,1)−EC(0,1,1)=2v.(39)E_C(0,0,1)-E_C(0,1,1)=2v. \tag*{(39)}

The first difference violates an equality of Corollary 7 whenever xx does not reach Charlie, and the second whenever yy does not. Each violation names the missing input and persists for every v>0v>0, which gives the zero entries of Table 6. The parity readout satisfies every equality of every schedule, so only an inequality can exclude it.

At a local vertex EABC=A(x)B(y)C(z)E_{ABC}=A(x)B(y)C(z), where A(x)A(x), B(y)B(y), C(z)C(z) are the signed response tables, and

Ipar:=∑x,y,zTEABC=C(0)α−β++C(1)α+β−,(40)\mathcal{I}_{\mathrm{par}}:=\sum_{x,y,z}T E_{ABC}=C(0)\alpha_-\beta_+ + C(1)\alpha_+\beta_-, \tag*{(40)}

with α±=A(0)±A(1)\alpha_{\pm}=A(0)\pm A(1) and β±=B(0)±B(1)\beta_{\pm}=B(0)\pm B(1). One of α±\alpha_{\pm} vanishes and the other equals ±2\pm2, so Ipar≤4\mathcal{I}_{\mathrm{par}}\le4 on P0P_0, while Ipar(Pvpar)=8v\mathcal{I}_{\mathrm{par}}(P_v^{\mathrm{par}})=8v. Equivalently, every local strategy gives the wrong parity in at least two of the eight contexts, an instance of Eq. (36).

Guessing zz attains the bound. Take shared unbiased bits g,r1,r2,r3g,r_1,r_2,r_3. If g=0g=0, Alice outputs x⊕r1x\oplus r_1 and Bob outputs r2r_2; if g=1g=1, Alice outputs r1r_1 and Bob outputs y⊕r2y\oplus r_2. Charlie outputs r1⊕r2r_1\oplus r_2 if z=gz=g and r3r_3 otherwise. The parity is uzu_z when the guess is right and uniform otherwise, and all other moments vanish. Hence Pvpar∈P0P_v^{\mathrm{par}}\in P_0 exactly for v≤1/2v\le1/2.

At v=1v=1 the target is also not quantum. Suppose the outputs come from ±1\pm1-valued observables Ax,By,CzA_x,B_y,C_z on separate tensor factors acting on a state ψ\psi. Perfect parity at contexts 000 and 100 gives B0C0ψ=A0ψB_0C_0\psi=A_0\psi and A1B0C0ψ=−ψA_1B_0C_0\psi=-\psi, hence A1A0ψ=−ψA_1A_0\psi=-\psi. The same argument at 001 and 101 gives A1A0ψ=ψA_1A_0\psi=\psi, a contradiction. We do not determine the quantum threshold.

Now consider the relay with two messages, where m2m_2 takes qq values. Every input reaches Charlie, so the direct readout passes all 14 equalities for every qq. Write Charlie’s row at uu as ρ(u)=(c(u,0),c(u,1))\rho(u)=(c(u,0),c(u,1)). The target requires ρ(u)=u\rho(u)=u, which is the table quoted after Proposition 9. Since Charlie sees uu only through m2m_2, ρ\rho takes at most qq values. Each row contributes ∑zT(u,z)(−1)c(u,z)=2−2 d(ρ(u),u)\sum_z T(u,z)(-1)^{c(u,z)}=2-2\,d(\rho(u),u), where dd is the Hamming distance, so at most qq rows contribute 2 and the rest at most 0:

Idir:=∑x,y,zTEABC≤2q,Idir(Pvdir)=8v.(41)\mathcal{I}_{\mathrm{dir}}:=\sum_{x,y,z}T E_{ABC}\le2q,\qquad\mathcal{I}_{\mathrm{dir}}(P_v^{\mathrm{dir}})=8v. \tag*{(41)}

The bound is attained. In all three protocols Alice sends m1=xm_1=x and Alice and Bob output 0.

  • q=2q=2: for a shared random jj, Bob sends uju_j. Charlie outputs it if z=jz=j and a shared unbiased bit otherwise, giving EC=T/2E_C=T/2.

  • q=3q=3: for shared random kk and tt, Bob sends uu unless uk=tu_k=t, in which case he sends a third symbol. On that symbol Charlie outputs tt if z=kz=k and a shared unbiased bit otherwise, giving EC=3T/4E_C=3T/4.

  • q=4q = 4: Bob sends uu.

Mixing with P0dirP_0^{\mathrm{dir}} covers smaller vv, so vdir⋆=q/4v_{\mathrm{dir}}^\star= q/4, although the affine hull is the same for every qq. The equalities cannot see this difference. The spanning argument of Theorem 5 isolates one indicator at a time, each needing one flag bit, whereas the task needs all four rows at once. Two broadcasts deliver xx and yy separately and realize the direct readout exactly.

The parity readout needs less communication. Alice can absorb the term xx, which leaves b⊕c=z(x⊕y)b \oplus c = z(x \oplus y). This is a PR task between Bob, with input x⊕yx \oplus y once he knows xx, and Charlie, with input zz. Without a message between them it obeys the local bound: repeating Eq. (40) with BB depending on (x,y)(x,y), or with BB and CC depending on xx, still gives 4. One relay bit or one broadcast by Alice therefore does not help. Two relay messages suffice even with binary m2m_2: Alice sends m1=xm_1 = x, Bob sends m2=x⊕ym_2 = x \oplus y, and the outputs are a=r1a = r_1, b=x⊕r1⊕r2b = x \oplus r_1 \oplus r_2, and c=r2⊕zm2c = r_2 \oplus z m_2. A broadcast of zz by Charlie also suffices, with a=xzˉ⊕r1a = x\bar{z} \oplus r_1, b=yz⊕r2b = yz \oplus r_2, and c=r1⊕r2c = r_1 \oplus r_2, where zˉ=1⊕z\bar{z} = 1 \oplus z.

The weights and inequalities above meet at v⋆v^\star, so the membership program (15) needs no numerics. Eq. (17) gives η⋆=1/2\eta^\star= 1/2 for Ipar≤4\mathcal{I}_{\mathrm{par}} \le4 and η⋆=1−q/4\eta^\star= 1 - q/4 for Eq. (41), so both witnesses are optimal along their noise lines. The equalities (39) instead detect every v>0v > 0. With nn trials per context, Hoeffding’s inequality and a union bound give ∣I^−I∣≤82log⁡(16/α)/n|\widehat{\mathcal{I}} - \mathcal{I}| \le8\sqrt{2\log(16/\alpha)/n} with probability at least 1−α1 - \alpha, which yields a level-α\alpha test for either inequality.

Low-order statistics miss the parity readout. For every vv, Π2(Pvpar)=0\Pi_2(P_v^{\mathrm{par}}) = 0 is the centre of the ball of Eq. (6), so every valid inequality on moments of order at most two has slack at least ∥c∥∞\lVert c\rVert_\infty there. Like the σx/σy\sigma_x/\sigma_y GHZ target, the parity readout is detectable only at order three, whereas the direct readout is exposed by the single-agent moment ECE_C.

For agents (§ 6), Corollary 12 gives Ipar≤4\mathcal{I}_{\mathrm{par}} \le4 for every admissible κ\kappa when each response uses only its own input and shared prior information; the guessing strategy is such a rule and attains the bound. The marginal-preserving correction of Eq. (35), applied to Alice and Bob, changes only EABE_{AB}, by 4δ4\delta at ν=1/2\nu= 1/2, and so leaves Ipar\mathcal{I}_{\mathrm{par}} unchanged.

For three superobservers with a friend and two settings each, the LF polytope equals P0P_0 [34], so Ipar≤4\mathcal{I}_{\mathrm{par}} \le4 is an LF inequality. For v>1/2v > 1/2, PvP_v violates it while satisfying every affine equality of the LF polytope. No LF inequality of order at most two detects the violation, and a broadcast of zz reproduces it. We do not determine whether quantum theory violates Ipar≤4\mathcal{I}_{\mathrm{par}} \le4. With all inputs fixed, as in the Frauchiger–Renner protocol, a single context remains, and shared randomness reproduces both readouts.

References

  1. [1]J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt. Proposed experiment to test local hidden-variable theories. Phys. Rev. Lett., 23:880–884, 1969. doi: 10.1103/PhysRevLett.23.880.DOI
  2. [2]N. Brunner, D. Cavalcanti, S. Pironio, V. Scarani, and S. Wehner. Bell nonlocality. Rev. Mod. Phys., 86:419–478, 2014. doi: 10.1103/RevModPhys.86.419.DOI
  3. [3]D. Bacon and B. F. Toner. Bell inequalities with auxiliary communication. Phys. Rev. Lett., 90:157904, 2003. doi: 10.1103/PhysRevLett.90.157904.DOI
  4. [4]K. Maxwell and E. Chitambar. Bell inequalities with communication assistance. Phys. Rev. A, 89:042108, 2014. doi: 10.1103/PhysRevA.89.042108.DOI
  5. [5]E. Zambrini Cruzeiro and N. Gisin. Bell inequalities with one bit of communication. Entropy, 21(2):171, 2019. doi: 10.3390/e21020171.
  6. [6]S. Pironio. Violations of Bell inequalities as lower bounds on the communication cost of nonlocal correlations. Phys. Rev. A, 68:062102, 2003. doi: 10.1103/PhysRevA.68.062102.DOI
  7. [7]A. A. Abbott, C. Giarmatzi, F. Costa, and C. Branciard. Multipartite causal correlations: Polytopes and inequalities. Phys. Rev. A, 94:032131, 2016. doi: 10.1103/PhysRevA.94.032131.DOI
  8. [8]S. Pironio. Lifting Bell inequalities. J. Math. Phys., 46:062112, 2005. doi: 10.1063/1.1928727.DOI
  9. [9]J. Tura, R. Augusiak, A. B. Sainz, B. Lücke, C. Klempt, M. Lewenstein, and A. Acín. Nonlocality in many-body quantum systems detected with two-body correlators. Ann. Phys., 362:370–423, 2015. doi: 10.1016/j.aop.2015.07.021.DOI
  10. [10]E. Y.-Z. Tan, D. Kaszlikowski, and L. C. Kwek. Entanglement witness via symmetric two-body correlations. Phys. Rev. A, 93:012341, 2016. doi: 10.1103/PhysRevA.93.012341.DOI
  11. [11]J. Tura, R. Augusiak, A. B. Sainz, T. Vértesi, M. Lewenstein, and A. Acín. Detecting nonlocality in many-body quantum states. Science, 344(6189):1256–1258, 2014. doi: 10.1126/science.1247715.DOI
  12. [12]V. I. Yukalov and D. Sornette. Mathematical structure of quantum decision theory. Adv. Complex Syst., 13:659–698, 2010. doi: 10.1142/S0219525910002803.DOI
  13. [13]M. Favre, A. Wittwer, H. R. Heinemann, V. I. Yukalov, and D. Sornette. Quantum decision theory in simple risky choices. PLOS ONE, 11(12):e0168045, 2016. doi: 10.1371/journal.pone.0168045.DOI
  14. [14]V. I. Yukalov. Evolutionary processes in quantum decision theory. Entropy, 22(6):681, 2020. doi: 10.3390/e22060681.DOI
  15. [15]V. I. Yukalov, E. P. Yukalova, and D. Sornette. Information processing by networks of quantum decision makers. Physica A, 492:747–766, 2018. doi: 10.1016/j.physa.2017.11.004.DOI
  16. [16]V. I. Yukalov and E. P. Yukalova. Emotions as intrinsic colored noise in biological systems. BioSystems, 264:105795, 2026. doi: 10.1016/j.biosystems.2026.105795.DOI
  17. [17]V. I. Yukalov and E. P. Yukalova. Emotions explain typical paradoxes in the behavior of biological systems. BioSystems, 267:105890, 2026. doi: 10.1016/j.biosystems.2026.105890.
  18. [18]Č. Brukner. A no-go theorem for observer-independent facts. Entropy, 20(5):350, 2018. doi: 10.3390/e20050350.DOI
  19. [19]D. Frauchiger and R. Renner. Quantum theory cannot consistently describe the use of itself. Nat. Commun., 9:3711, 2018. doi: 10.1038/s41467-018-05739-8.
  20. [20]K.-W. Bong, A. Utreras-Alarcón, F. Ghafari, Y.-C. Liang, N. Tischler, E. G. Cavalcanti, G. J. Pryde, and H. M. Wiseman. A strong no-go theorem on the Wigner’s friend paradox. Nat. Phys., 16:1199–1205, 2020. doi: 10.1038/s41567-020-0990-x.
  21. [21]A. Fine. Hidden variables, joint probability, and the Bell inequalities. Phys. Rev. Lett., 48:291–295, 1982. doi: 10.1103/PhysRevLett.48.291.DOI
  22. [22]R. O’Donnell. Analysis of Boolean Functions. Cambridge University Press, Cambridge, 2014. doi: 10.1017/CBO9781139814782.DOI
  23. [23]E. Kushilevitz and N. Nisan. Communication Complexity. Cambridge University Press, Cambridge, 1997. doi: 10.1017/CBO9780511574948.DOI
  24. [24]W. Hoeffding. Probability inequalities for sums of bounded random variables. J. Am. Stat. Assoc., 58(301):13–30, 1963. doi: 10.1080/01621459.1963.10500830.DOI
  25. [25]S. Popescu and D. Rohrlich. Quantum nonlocality as an axiom. Found. Phys., 24:379–385, 1994. doi: 10.1007/BF02058098.DOI
  26. [26]B. S. Cirel’son. Quantum generalizations of Bell’s inequality. Lett. Math. Phys., 4:93–100, 1980. doi: 10.1007/BF00417500.DOI
  27. [27]N. D. Mermin. Extreme quantum entanglement in a superposition of macroscopically distinct states. Phys. Rev. Lett., 65(15):1838–1840, 1990. doi: 10.1103/PhysRevLett.65.1838.DOI
  28. [28]D. M. Greenberger, M. A. Horne, A. Shimony, and A. Zeilinger. Bell’s theorem without inequalities. Am. J. Phys., 58:1131–1143, 1990. doi: 10.1119/1.16243.DOI
  29. [29]B. F. Toner and D. Bacon. Communication cost of simulating Bell correlations. Phys. Rev. Lett., 91:187904, 2003. doi: 10.1103/PhysRevLett.91.187904.DOI
  30. [30]T. E. Tessier, C. M. Caves, I. H. Deutsch, B. Eastin, and D. Bacon. Optimal classical-communication-assisted local model of n-qubit Greenberger–Horne–Zeilinger correlations. Phys. Rev. A, 72:032305, 2005. doi: 10.1103/PhysRevA.72.032305.
  31. [31]E. P. Wigner. Remarks on the mind-body question. In J. J. Good, editor, The Scientist Speculates, pages 284–302. Heinemann, London, 1961.
  32. [32]M. Proietti, A. Pickston, F. Graffitti, P. Barrow, D. Kundys, C. Branciard, M. Ringbauer, and A. Fedrizzi. Experimental test of local observer independence. Sci. Adv., 5(9):eaaw9832, 2019. doi: 10.1126/sciadv.aaw9832.DOI
  33. [33]R. Healey. Quantum theory and the limits of objectivity. Found. Phys., 48:1568–1589, 2018. doi: 10.1007/s10701-018-0216-6.DOI
  34. [34]M. Haddara and E. G. Cavalcanti. Local friendliness polytopes in multipartite scenarios. Phys. Rev. A, 111(1):012206, 2025. doi: 10.1103/PhysRevA.111.012206.DOI
  35. [35]L. Hardy. Quantum mechanics, local realistic theories, and Lorentz-invariant realistic theories. Phys. Rev. Lett., 68:2981–2984, 1992. doi: 10.1103/PhysRevLett.68.2981.DOI
  36. [36]L. Hardy. Nonlocality for two particles without inequalities for almost all entangled states. Phys. Rev. Lett., 71:1665–1668, 1993. doi: 10.1103/PhysRevLett.71.1665.DOI
  37. [37]M. Haddara and E. G. Cavalcanti. A possibilistic no-go theorem on the Wigner’s friend paradox. New J. Phys., 25:093028, 2023. doi: 10.1088/1367-2630/aceea3.DOI
  38. [38]H. M. Wiseman, E. G. Cavalcanti, and E. G. Rieffel. A “thoughtful” Local Friendliness no-go theorem: a prospective experiment with new assumptions to suit. Quantum, 7:1112, 2023. doi: 10.22331/q-2023-09-14-1112.DOI
  39. [39]Y. Ying, M. Maciel Ansanelli, A. Di Biagio, E. Wolfe, D. Schmid, and E. G. Cavalcanti. Relating Wigner’s friend scenarios to nonclassical causal compatibility, monogamy relations, and fine tuning. Quantum, 8:1485, 2024. doi: 10.22331/q-2024-09-26-1485.DOI
  40. [40]L. Walleghem, Y. Ying, R. Wagner, and D. Schmid. Connecting extended Wigner’s friend arguments and noncontextuality. Quantum, 9:1819, 2025. doi: 10.22331/q-2025-07-31-1819.DOI
  41. [41]M. Haddara, H. M. Wiseman, and E. G. Cavalcanti. Properties and applications of partially deterministic polytopes. arXiv:2510.25127, 2025.arxiv.org/abs/2510.25127
  42. [42]S. Khanna, M. Maciel Ansanelli, M. F. Pusey, and E. Wolfe. Classifying causal structures: Ascertaining when classical correlations are constrained by inequalities. Phys. Rev. Research, 6(2):023038, 2024. doi: 10.1103/PhysRevResearch.6.023038.DOI
  43. [43]M. Kalisch and P. Bühlmann. Estimating high-dimensional directed acyclic graphs with the PC-algorithm. J. Mach. Learn. Res., 8(22):613–636, 2007. URL https://www.jmlr.org/papers/v8/kalisch07a.html.

Paper details

vlnc:W_ca3wsxzx
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 7

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents