| paper | d4-conf-aaai-2004-0157 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The continuous Kemeny question is a direct, recognisable mirror, but it cannot supply the required anchor: the NP-hardness is cited as prior work, and the paper's own algorithms and experiments have no numbered computational theorem. Bit (a) therefore fails even though the proposed mass formulation is well-posed.
fails bit a — no named computational result to mirror
The mirror captures the Kemeny optimization task addressed by the paper's exact procedure and heuristic; it does not turn their unnumbered algorithm descriptions or empirical findings into a numbered computational result.
There is no qualifying anchor in this paper: it contains no numbered Theorem, Lemma, Corollary, or Proposition asserting a computational result. Its NP-hardness statement is unnumbered and explicitly attributed to Cohen, Schapire, and Singer (1999) and Dwork et al. (2001). The branch-and-bound and heuristic procedures are described in prose, with experiments, rather than stated as numbered results. So I cannot give the required anchor-based case without misrepresenting the paper.
The strongest positive mirror I see is still quite direct. Take a large electorate in which many voters share the same strict ranking. A type is a ranking \(t\) of the alternatives; its mass \(\mu_t\) is the fraction of voters with that ranking. Given the \(\tau\) supported rankings and rational masses summing to one, ask for a total order \(r\) minimizing
\[
\sum_t \mu_t\,\Delta(r,t),
\]
where \(\Delta\) is Kendall tau distance. A solution is a minimizing consensus ranking. This is the Kemeny aggregation problem for a distribution over voter types, and it preserves the paper’s objective rather than replacing it with a tractable surrogate.
That is a plausible high-multiplicity regime: for example, a large user base supplying rankings drawn from a relatively small set of recurring preference patterns. The paper’s central problem is precisely to find a Kemeny order from voters’ rankings, so its authors would likely recognize this as the continuous-population version. Moreover, any finite profile embeds by assigning each ranking mass equal to its frequency divided by the number of voters. Thus the known discrete NP-hardness transfers; this is a Class B candidate, with further questions about approximation and parameterization by \(m\), \(\tau\), and the mass bit length \(L\).
The weakness is the anchor requirement: the paper’s unnumbered NP-hardness claim is the natural hardness anchor, but it does not meet your stated rule. The continuous formulation is a credible mirror, but under that rule it cannot serve as the requested numbered-result case.
The strongest negative point is procedural: the paper has no numbered theorem, lemma, or corollary establishing a computational result. Its NP-hardness statement is attributed to earlier work; its own algorithmic contributions are described in prose and evaluated experimentally. So there is no named result here to anchor a case against, and the proponent was right not to invent one.
That does not defeat the proposed Kemeny mirror. A large population with recurring preference rankings gives a natural type space, and minimizing \(\sum_t \mu_t\Delta(r,t)\) preserves the paper’s objective. The objective does not depend on voter identities, and normalized Kendall distance does not vanish in the limit. Treating frequencies as rational masses may resemble a change of units, but the support-size representation creates the high-multiplicity computational question the programme is meant to study. The paper supplies no existing continuous computational treatment that answers it.
So my negative case is weak beyond the lack of a named result to anchor. The substantive mirror survives: I cannot honestly rule out a worthwhile continuous-population study of Kemeny aggregation.
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.