| paper | d4-conf-atal-2018-0341 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
Theorem 4.2
statement extracted from the paper’s text layer
Given core and periphery masses \(\mu_C,\mu_P\) with \(\mu_C+\mu_P=1\), per-capita degree-volume weights \(d_C,d_P\), a rational \(q\in[0,1]\) for the periphery’s probability of sampling the frozen-blue core on each draw, a target \(\kappa\), and a horizon \(H\), initialize the core blue and the periphery red. Let \(x_t\) be the blue fraction of the periphery, \(y_t=q+(1-q)x_t\), and \(x_{t+1}=(1-x_t)y_t^2+x_t(1-(1-y_t)^2)\). Decide whether, for some \(t\le H\), the degree-volume-weighted blue share \((\mu_Cd_C+\mu_Pd_Px_t)/(\mu_Cd_C+\mu_Pd_P)\) is at least \(1-\kappa\).
A continuum of agents in core and periphery types, carrying masses \(\mu_C,\mu_P\) and degree-volume weights \(d_C,d_P\). The core acts as a frozen blue reservoir for the horizon; periphery mass follows the deterministic 2-Choices update induced by its core-sampling probability \(q\).
The finite theorem covers every graph satisfying its block-degree ratios, while the proposed recurrence assumes role-level exchangeability; the paper does not prove that all such graphs converge to this recurrence.
fatal: False
The mirror covers Theorem 4.2’s supercritical consensus question and the related subcritical persistence regime. It leaves the general graph breadth of Theorem 3.4, the critical clique’s finite-population hitting-time bounds, and the experiments and core-extraction heuristic outside the model.
The strongest honest case is that this paper has a very natural high-multiplicity population setting, but it has no qualifying computational anchor for the ChoCo programme. Its numbered results analyze the probabilistic behavior of a stochastic process; none asserts that a problem is in \(\mathrm{P}\), NP-hard, FPT, or otherwise computationally classified. I therefore can’t claim a continuous computational mirror is anchored in a named result of the required kind.
The closest results are Theorem 3.4 and Theorem 4.2, both proved in the paper. Theorem 3.4 says that on an \(n\)-agent graph with minimum degree \(\omega(\log n)\), the biased 2-Choices process started with everyone red has a threshold \(p^\star=3-2\sqrt{2}\): above it, blue reaches a \(\kappa(n)\)-consensus within \(O(\log(1/\kappa(n)))\) rounds with high probability, for \(\kappa(n)\geq \lambda\log n/d_{\min}\); below it, blue’s volume stays at most \(\frac{1-3p}{4(1-p)}\mathrm{vol}(V)\) for polynomially many rounds with high probability; at the threshold, on a clique, consensus takes between \(\Omega((n/\log n)^{1/4})\) and \(O(n\log n)\) rounds with high probability. Theorem 4.2 transfers this phase-transition picture to the paper’s core-periphery networks: with high robustness \(c_r>n^{(\epsilon+\delta)/2}\) and dominance \(c_d>(\sqrt{2}-1)/2\) by a constant, blue reaches almost-consensus in \(O(\log(1/\kappa(n)))\) rounds; with \(c_r>1/c^\star\), at least three quarters of the core’s volume remains blue for polynomially many rounds; and with \(c_d<c^\star\), at least three quarters of the periphery’s volume remains red. These are substantial analytic results, but they are not complexity results.
A plausible *new* continuous question suggested by them would be Core-Periphery Continuum Reachability. Agents have one of a fixed number of roles, such as core or periphery, and one of two opinions. Each type has a rational population mass, degree weight, and probabilities of sampling each role as a neighbor. Given those quantities, the initial opinion masses, a bias parameter, a target \(\kappa\), and a horizon \(H\), ask whether the induced continuum update reaches blue mass at least \(1-\kappa\) by round \(H\); a solution is a yes/no answer and, on yes instances, the earliest such round. This is a type-level dynamic question, not an optimization objective.
The regime is credible: the paper studies many agents with only two structural roles and common local interaction ratios. In its model, however, the core may have \(n^\epsilon\) agents against \(n\) in the periphery, so its population share vanishes for \(\epsilon<1\) even while its edge volume and influence remain substantial. A faithful continuum would need to carry interaction or degree volume alongside agent mass. That co-modification seems defensible, and the authors could recognize the resulting question as related to their phase analysis.
The weakness is that this reachability problem is not a result of the paper. Its continuum recurrence would abstract away the finite network’s stochastic fluctuations and topology; recovering the paper’s high-probability conclusions would require additional analysis. It may be a useful follow-on computational question, but assigning it a tractability or hardness class would be speculation, not a mirror of one of the paper’s named results. The paper’s continuous-population motivation is therefore strong; its case as a ChoCo computational anchor is not.
The strongest objection is that this paper has no computational result to mirror. Theorem 3.4 and Theorem 4.2 prove probabilistic and asymptotic behavior of a dynamics: they locate a phase transition, bound consensus time, and show metastability under structural assumptions. They do not classify the complexity of a computational problem. Calling their conclusions “anchors” for ChoCo would stretch the programme from computational questions to analytic ones.
The proposed Core-Periphery Continuum Reachability problem does not fix that gap. On a fixed, small set of roles with specified interaction rates, the natural continuum model is a low-dimensional deterministic recurrence. Asking whether its blue mass crosses a threshold by round \(H\) is a new reachability problem, not a computational question established by either theorem. Expanding the model to preserve arbitrary network structure makes node types carry increasingly detailed neighborhood information; the support can then grow with the network, erasing the proposed type-level compression. Either way, the paper supplies no reason to expect a useful complexity landscape.
Theorem 3.4’s critical case is an especially poor target for a faithful continuum mirror. Its clique result concerns finite-population random fluctuations: the proof uses variance of order \(n\) to bound the time to consensus. In a population continuum, those fluctuations vanish. The deterministic fraction recurrence instead approaches the critical fixed point; it does not reproduce the finite-\(n\) consensus-time phenomenon. One could study a rescaled fluctuation limit, but that is a different limit question, not a computational mirror of the theorem.
Theorem 4.2 offers the better case for a continuous population model: many peripheral agents share a role, and a small core can exert substantial influence. But the core has \(n^\epsilon\) agents against \(n\) in the periphery, so for \(\epsilon<1\) its population mass vanishes even as its influence remains important. Adding degree or interaction volume can preserve that influence, as the proponent observes. Yet with a fixed number of roles this yields another small mean-field system; retaining the heterogeneous graph details that drive the theorem requires richer types and forfeits that compression. The theorem’s phase analysis may motivate such a model, but it does not turn that model into a computational problem.
I would therefore reject these as ChoCo anchors, while not claiming that no continuous model of opinion dynamics could be worthwhile. The honest negative case is strongest on the programme’s stated criterion: this paper analyzes stochastic behavior rather than computational complexity, and its most faithful continuum limits either discard the finite-size phenomenon or reduce to analytic dynamics. The broader claim that *no scenario* could produce a valuable computational mirror is weaker; a new intervention or network-design problem might be interesting, but it would need its own motivation rather than being credited as a mirror of these results.
The adversarial triple: the proponent anchors on up to three named results; the opponent sees that case and must defeat every anchor; the judge decides which case convinced it. These are the pipeline’s own outputs, generated by tools/triple_run.py — no human edited them. The paper’s own text is not reproduced here beyond the quoted statement above.