| paper | d4-conf-aaai-2016-0076 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The paper formulates a blocking objective and proposes two heuristics, but contains no numbered theorem, lemma, corollary, or proposition asserting an algorithmic result or hardness. Its proposed continuous model is a possible research question, but no named computational result in the paper anchors it as a mirror.
fails bit a — no named computational result to mirror
No named computational result is covered; the candidate mass-control formulation mirrors the objective but leaves the paper's edge-level likelihood model and heuristics without guarantees.
The supplied paper contains no numbered Theorem, Lemma, Corollary, or Proposition asserting a computational result. Equation (6) formulates DRIMUX, while Algorithms 1 and 2 propose blocking heuristics; the experiments compare them, but the paper gives no complexity classification or algorithmic guarantee. Those formulations and experiments are not eligible anchors under your rule. So there is no honest per-anchor question to state, and no result here to classify as tractable, hardness-transferring, or continuum-specifically hard.
There is a plausible *candidate* continuous version of the paper’s central optimization. Consider a very large platform with a small number \(\tau\) of recurring user roles. A type \(a\) would describe users with the same rumor response, blocking tolerance, and interaction profile with every other role. Its mass \(\mu_a\) is the fraction of users in that role. The instance could also give initial rumor mass by type, a type-to-type transmission kernel over time, a blocking budget \(\kappa\) as a fraction of the population, a horizon, and the user-experience threshold. A solution would specify how much mass of each type to block over time; the objective would be to minimize rumor-infected mass at the horizon, subject to the budget and an aggregate blocking-time utility constraint. This mirrors the underlying blocking problem, rather than treating its inference formulation as the problem.
A large platform with repeated user segments and roughly stable interaction patterns makes that regime conceivable: the number of users could be enormous while the number of behavior and interaction types remains modest. But this version requires a block-structured or mean-field network model. The paper’s propagation probabilities and blocking decisions depend on individual edges and evolving cascades. Replacing those with type-level interactions may change the optimization problem; the paper provides no result showing that its algorithms or conclusions survive that change. The weakest point is therefore whether its authors would recognize the type-level model as their DRIMUX problem, rather than as a new approximation.
That is the strongest positive modeling lead I see, but it is not anchored in a named computational result from this paper. The paper is promising material for proposing a mirror question, not evidence—by the requested standard—that one of its stated results has a continuous counterpart.
There is no named computational result here to carry into the trichotomy. Equation (6) states an optimization problem, and Algorithms 1 and 2 propose blocking heuristics; the paper gives no complexity classification, approximation guarantee, or optimality theorem for them. A continuous solution to the proposed mass-control model could be interesting, but it would classify a new problem rather than supply a continuous counterpart to a computational result established in this paper.
The proponent’s recurring-user-segment model is the strongest candidate. Its obstacle is the paper’s edge-level input. In equations (1)–(4) and (13), propagation depends on the directed edges, in-degrees, and evolving cascade state. Blocking a particular vertex matters because of its position in that network. To make users of one type interchangeable for this process, a type must capture their interaction roles across the network, as well as their rumor response and blocking tolerance. In a general observed graph, those roles can distinguish nearly every vertex. Grouping users by broad behavior instead leaves out information that determines the effect of blocking them.
A type-to-type transmission kernel, or a graphon with a finite block description, can supply a sensible population model. It does so by replacing the observed network and its vertex-selection problem with a block-structured network and mass-level control. That may be a worthwhile adjacent problem; the paper does not show that its likelihood calculations or greedy choices carry over to it. An unrestricted graphon preserves more network detail, but then a succinct finite-type input—and the intended high-multiplicity advantage—needs further assumptions.
So the negative case is strong against treating this paper as evidence for a result-anchored continuous mirror: none of its stated results makes a computational claim of that kind, and the best population model changes the network structure the algorithms operate on. The case is weaker against the broader claim that no useful continuous DRIMUX variant exists. A finite-block model could be worth studying on its own, and the authors might recognize it as a natural extension. What the paper supplies, though, is little basis for calling that extension a mirror of one of its computational 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.