| paper | d4-conf-ijcai-2003-0098 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The paper claims polynomial algorithms for optimal and \(a\)-strong controllability, but gives no numbered theorem, lemma, corollary, or proposition asserting a computational result. The proposed workload-share question is a coherent extension that repeats those checks across templates, but it cannot satisfy the required anchor criterion.
fails bit a — no named computational result to mirror
The proposed mirror covers only the optimal and \(a\)-strong controllability claims in §3, by checking each template independently; it leaves weak controllability and the proposed future study of dynamic controllability aside.
There is a strict limit on the positive case: the paper contains no numbered Theorem, Lemma, Corollary, or Proposition asserting a computational result. It does make computational claims, but under your anchor rule none can serve as a formal anchor.
The best substantive candidate is the unnumbered claim in §3. The paper says that checking optimal or \(a\)-strong controllability is in P. Its proposed method applies Path-solver, converts the result to an STPU \(P_{\mathrm{opt}}\), then checks strong controllability; it gives a running time of \(O(n^3 R \ell)\), where \(n\) is the problem size, \(R\) the largest interval range, and \(\ell\) the number of preference levels. For \(a\)-strong controllability, it gives an \(O(n^3R)\) algorithm. The paper says the correctness results are proved but omits them for space. This is a claim from the paper, but it is unnumbered, so it cannot meet the requested anchor criterion.
A plausible population setting would be a large operations desk handling many recurring temporal jobs—such as repeated observation packages—with only a modest number of distinct templates. A type would be a complete STPPU job template: its timepoints, requirement and contingent partitions, temporal constraints, uncertainty intervals, and preferences. The mass \(\mu_t\) is the fraction of jobs of template \(t\); the binary-rational masses sum to one. This is a genuine high-multiplicity regime when there are many jobs but few templates.
Here is a precise candidate mirror, based on that unnumbered \(a\)-strong-controllability claim. Given finite support \(T\), masses \(\mu\), a preference threshold \(a\), and a target share \(q\), find a maximum-mass subset of types \(S\subseteq T\) such that for every \(t\in S\) there is a fixed execution strategy \(\pi_t\) that satisfies the job’s temporal constraints and reaches preference at least \(a\) for every realization of its contingent durations. Equivalently, decide whether such strategies can be supplied for types of total mass at least \(q\). A solution consists of \(S\) and one valid strategy for each included type.
The mirror’s computational question is tractable on this model: test each supported template using the paper’s \(a\)-strong-controllability algorithm, then sum the masses of the feasible types. The discrete version counts jobs with guaranteed execution; the continuous version reports the guaranteed fraction. This looks like a Class A case, assuming the paper’s stated algorithm applies to each template as given.
My weakest point is that the population layer adds no interaction between jobs. It turns a single-network controllability check into a weighted service-coverage question, rather than making controllability itself depend on a society-wide state. That makes the high-multiplicity scenario plausible, but leaves room for the authors or a referee to say that it is an application of their result rather than a continuous form of their problem. And because the paper has no numbered computational result, this candidate still does not satisfy the requested anchor requirement.
On the requested anchor standard, the proponent’s only anchor fails at the outset: the paper has no numbered theorem, lemma, corollary, or proposition asserting a computational result. It does make unnumbered complexity claims, including that checking \(a\)-strong controllability is in P, so there is something one could try to mirror. But the proponent acknowledges that this is not a named result under the stated standard.
Even taking that claim as an anchor, the proposed mirror adds no population-level computation. Let \(P_a(t)\) mean that type \(t\)’s STPPU is \(a\)-strongly controllable. For independent jobs, the maximum guaranteed share is simply
\[
\sum_{t\in T}\mu_t\,\mathbf{1}[P_a(t)].
\]
To find it, run the paper’s checker once for each supported type and sum the masses. Asking for the largest \(a\) that guarantees share \(q\) is a weighted quantile of the types’ individual thresholds, computed the same way. The mass does not affect any controllability constraint. This is a sensible workload statistic, but its computational content is repeated use of the paper’s single-instance algorithm, followed by aggregation—not a new continuous controllability problem.
A repeated-template setting is plausible: an operations desk might handle many jobs with the same temporal network. But a type must include the full network, its contingent intervals, and its preferences. Differences in windows or uncertainty bands create different types; if exact templates do repeat, their controllability results are identical across copies. Either way, population size adds no new algorithmic structure. The paper’s uncertainty is about nature’s choices within a network, and its continuous time intervals and preference values are not themselves a continuous population. Nor is prior continuous treatment the objection here.
The strongest improved mirror would couple jobs through shared capacity or a common execution policy. That could produce a substantive population-level optimization question. But the paper’s results concern controllability of one temporal network; shared resources and interactions between jobs would add a different model, with new semantics for joint execution and uncertainty. It may be worthwhile work, but it does not rescue the proposed mirror of this paper’s result. The negative case is strongest against the anchors actually raised; the universal claim remains hard to make airtight if one permits adding those new interactions.
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.