| paper | d4-conf-ijcai-2021-0047 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The proposed role-mass winner question is plausible, but it does not rescue the paper’s only anchor: Theorem 3 proves an outcome condition for the dynamics, not a result about computing that outcome. The paper therefore lacks a named computational result paired with the proposed continuous question.
fails bit a — no named computational result to mirror
The proposed mirror covers only Theorem 3’s sufficient non-winning guarantee; it leaves the elite experiments, random initial colorings, and stabilization-time results without a mirror.
The strict anchor test yields no qualifying computational-complexity result. Theorem 2, Theorem 3, Theorem 4, and Theorem 5 are proved here, but they establish guarantees about the dynamics; none classifies a problem as tractable, hard, or parameterized. In particular, Theorem 5’s stabilization-time bound is a bound on rounds of the process, not on the computational time needed to solve a problem. So I can’t honestly offer a paper result that meets your anchor requirement.
The strongest nearby result is Theorem 3, proved here. It says: for a graph \(G=(V,E)\), let \(Z\subset V\) have \(|Z|<n/2\), and suppose every \(v\notin Z\) has at most an \(f\)-fraction of its neighbors in \(Z\). In the majority model where nodes in \(Z\) have influence factor \(r\in\mathbb N\) and all other nodes have influence factor \(1\), if each node outside \(Z\) has stubbornness factor greater than
\[
\frac{r}{r+(1-f)/f},
\]
then \(Z\) is not a winning set. This is a useful structural guarantee for a winner-determination question, but it is not a complexity result.
If that structural theorem is admitted as an anchor, its most faithful continuous mirror is Continuous Elite-Coalition Winner. An instance has a finite set \(T\) of network roles, rational masses \(\mu_i>0\) summing to \(1\), an elite subset \(E\subseteq T\) with \(\sum_{i\in E}\mu_i<1/2\), a symmetric nonnegative contact matrix \(A\), and influence and stubbornness factors. Each row of \(A\) gives a role’s contact exposure to the roles in \(T\); every agent of a role has the same exposure row and the same model parameters. Initially, all elite-role mass is black and all other mass is white.
Let \(x_i(t)\in\{0,\mu_i\}\) be the black mass in role \(i\) at round \(t\). Each round applies the paper’s majority update to these role-level states: contacts from role \(j\) are weighted by \(r_j\), and a nonelite role changes color only when the weighted fraction of opposite-color contacts reaches its stubbornness threshold. The question is whether the initial elite coalition wins—whether black holds more than half the mass in the terminal configuration. A solution is a yes/no answer, with the eventual trajectory or cycle as a certificate. This asks whether a specified coalition can control the outcome, rather than optimizing a surrogate such as its size.
A plausible regime is a large online platform whose users fall into a modest number of roles—communities, elite status, and exposure profiles—with many users per role. Here mass is a population share, and the contact matrix records the role-to-role exposure pattern. The paper itself studies elites and countermeasures in social networks, so the underlying question should be recognizable to its authors. It covers only the countermeasure guarantee in Theorem 3, not the paper’s random-initial-coloring or stabilization-time results.
In this mirror, the theorem’s condition becomes a direct mass-level certificate. If every nonelite role gets at most an \(f\)-fraction of its contacts from \(E\), then an initially white nonelite role sees at most
\[
\frac{rf}{rf+1-f}
\]
of its influence-weighted contacts as black. When its stubbornness threshold is strictly above that value, it does not turn black; as the elite mass is below \(1/2\), black cannot win. Checking this certificate is tractable from the rational input. For the unrestricted winner-determination problem, though, the paper gives no complexity classification, and I would not infer one from Theorem 3. The most I would predict is tractability for the theorem’s promise regime; the general problem remains open here.
The weakest point is also the largest one: role-level exposure replaces a graph’s individual neighborhoods with shared, deterministic contact profiles. That is a plausible high-multiplicity regime for some platforms, but it is an additional modeling assumption, and the paper does not formulate the resulting winner-determination question as a computational problem. This is a credible population-level extension of its elite-resistance result, but it cannot supply the requested computational anchor.
The strongest case against a paper-anchored mirror is that the paper proves no computational-complexity result. Theorem 3 is a sufficient condition for a coalition not to win; Theorems 4 and 5 bound the number of rounds the dynamics take. Neither classifies an algorithmic problem as tractable or hard. Theorem 3 is the closest anchor, but its finite-role version simply replaces neighborhood counts with contact masses: the condition becomes \(\gamma>\frac{rf}{rf+1-f}\). Checking that condition is a direct calculation, not a new complexity question.
The proposed contact-matrix model also needs strong structure to stay closed at the role level. Unless agents in each role have compatible neighborhoods and move in sync, role masses alone do not determine the next state. Encoding individual neighborhood patterns can make the number of types grow with the population; keeping few types means imposing an equitable-block or mixing assumption. Under that assumption, Theorem 3’s guarantee transfers cleanly, but it remains the same sufficient-condition argument. Asking for the smallest mass that can seed a winning cascade would be a more substantial computational problem, but that is a new influence-optimization problem, not a mirror of the theorem.
That is the strongest honest negative case, but it does not establish the requested universal claim. The paper explicitly motivates finding a small set to seed, and a large platform with repeated user roles and stable contact patterns is a plausible high-multiplicity setting. A continuous minimum-seed problem there could be worthwhile. I would reject Theorem 3 as a computational anchor, while conceding that this paper may inspire a valuable adjacent mirror.
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.