Diluted mean-field models are spin systems whose geometry of interactions is induced by a sparse random graph or hypergraph. Such models play an eminent role in the statistical mechanics of disordered systems as well as in combinatorics and computer science. In a path-breaking paper based on the non-rigorous 'cavity method', physicists predicted not only the existence of a replica symmetry breaking phase transition in such models but also sketched a detailed picture of the evolution of the Gibbs measure within the replica symmetric phase and its impact on important problems in combinatorics, computer science and physics [Krzakala et al.: PNAS 2007]. In this paper we rigorise this picture completely for a broad class of models, encompassing the Potts antiferromagnet on the random graph, the k-XORSAT model and the diluted k-spin model for even k. We also prove a conjecture about the detection problem in the stochastic block model that has received considerable attention [Decelle et al.: Phys. Rev. E 2011].1 k h=1 ρ π h (σ h ) . 2 and d cond (k, β) = inf{d > 0 : sup π∈P 2 * ({1,−1}) B k−spin (d, β, π) > ln 2}. Then 0 < d cond (k, β) < ∞ and (k, β).From now on we assume that k ≥ 4 is even. The regime d < d cond (k, β) is called the replica symmetric phase. According to the cavity method, its key feature is that with probability tending to 1 in the limit n → ∞, two independent samples σ 1 , σ 2 ('replicas') chosen from the Gibbs measure µ H,J ,β are "essentially perpendicular". To formalize this define for σ, τ : V n → {±1} the overlap as ̺ σ,τ = x∈V n σ(x)τ(x)/n. We write 〈 · 〉 H,J ,β for the average on σ 1 , σ 2 chosen independently from µ H,J ,β and denote the expectation over the choice of H and J by E [ · ]. Theorem 1.2. For all β > 0 and k ≥ 4 even we have d cond (k, β) The corresponding statement for k = 2 was proved by Guerra and Toninelli, but as they point out their argument does not extend to larger k [37]. Theorem 1.2 implies the absence of extensive long-range correlations in the replica symmetric phase. Indeed, for two vertices x, y ∈ V n and s, t ∈ {+1, −1} letbe the joint distribution of the spins assigned to x, y. Further, letρ be the uniform distribution on {±1} × {±1}. Then the total variation distance µ H,J ,β,x,y −ρ TV is a measure of how correlated the spins of x, y are. Indeed, in the case that k is even for every x ∈ V n the Gibbs marginals satisfy µ H,J ,β,x (±1) = 〈1{σ 1 (x) = ±1}〉 H,J ,β = 1/2 because µ H,J ,β (σ) = µ H,J ,β (−σ) for every σ ∈ {−1, +1} n . Therefore, if the spins at x, y were independent, then µ H,J ,β,x,y = µ H,J ,β,x ⊗ µ H,J ,β,y =ρ. Furthermore, it is well known (e.g., [13, Section 2]) thatThus, Theorem 1.2 implies that for d < d cond (k, β), with probability tending to 1, the spins assigned to two random vertices x, y of H are asymptotically independent. By contrast, Theorem 1.2 and (1.2) show that extensive longrange dependencies occur beyond but arbitrarily close to d cond (k, β).1.3. The Potts antiferromagnet. Let q ≥ 2 be an integer, let Ω = {1, . . . , q} be a set...