| paper | d4-conf-sagt-2024-0011 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Neither cited theorem asserts that an object can be computed or that computing it is hard. The mass-deviation proposal is a possible new multi-resource coalition problem, but it changes the paper’s unilateral multiresource incentive question and is not anchored in a named computational result.
fails bit a — no named computational result to mirror
The proposed mirror covers none of the paper’s named results: Theorem 4.4 concerns multiresource individual incentives, while Theorem 3.5 concerns single-resource incentives. The paper’s fairness guarantees, single-resource coalition analysis, and repeated-manipulation result are also not computational results to mirror.
The paper has no numbered theorem, lemma, corollary, or proposition asserting a computational complexity result of the kind required for an anchor. Its named results concern incentive guarantees and fairness properties. The closest is Theorem 4.4, proved here: when every user has positive ratios for every resource, no user can increase cumulative utility through demand or ratio misreporting by more than \(1+\rho_i\), where \(\rho_i=\max_{k\ne i,q,t}\frac{w_i a_{iq}^t}{w_k a_{kq}^t}\). Theorem 3.5, also proved here, gives the \(3/2\) incentive bound for single-resource Dynamic MMF. Neither says that a problem is tractable, hard, or otherwise classifies its computational complexity. So there are no qualifying anchors, and I can’t honestly present a per-anchor complexity question.
The strongest plausible population mirror I see is a mass-deviation Dynamic DRF problem for a large cloud system. Fix a finite horizon \(H\) and a finite set of workload trajectories. A type \(\theta\) specifies a priority \(w_\theta\) and, for each epoch, the user’s true demand \(d_\theta^t\) and resource ratios \(a_\theta^t\). The society is a rational distribution \(\mu\) over the supported trajectories. Mass represents the fraction of tenants with the same full trajectory; the number of tenants is large while the number of supported trajectories \(\tau\) is moderate. Resource capacity scales with population, so \(\mathcal R_q^t\) is capacity per unit mass. Time remains in discrete epochs.
The mechanism allocates per-capita resources by Dynamic DRF over the reported trajectory groups, applying its weighted leximin rule to cumulative allocations. A deviation is a mass-transfer array \(x_{\theta,\hat\theta}\): mass of true type \(\theta\) reports trajectory \(\hat\theta\), with each type’s outgoing mass at most \(\mu_\theta\). The question is to find a feasible \(x\) maximizing the ratio of the deviators’ total true utility under the resulting allocation to their total utility under truthful reports. A solution consists of the transfer array and the resulting per-type allocations. This is a precise population-level question about the paper’s dynamic allocation and manipulation setting; the paper’s coalition analysis for single-resource weighted MMF gives some conceptual support for studying group deviations.
The weak point is substantial: the main multiresource result concerns one user’s deviation. A single user has vanishing mass in this mirror and cannot affect the allocation; a positive-mass deviation is a coalition problem, whose multiresource version the paper does not analyze. Theorem 4.4 therefore does not establish a bound for this question. I would treat it as a plausible new extension, with its complexity open, rather than a valid anchored case for the continuization programme.
The proponent raised no qualifying anchor. Theorem 4.4’s \(1+\rho_i\) bound and Theorem 3.5’s \(3/2\) bound describe how much a user can gain by misreporting; they do not classify the complexity of computing an allocation or finding a manipulation. The coalition result in Section 5.1 is also a guarantee about incentive strength, not an algorithmic result. Under the programme’s stated gate, there is no computational result here to continuize.
The cloud example is plausible as a high-multiplicity regime: tenants with the same complete workload trajectory can be grouped, with capacity scaled per capita. So I would not argue that population mass is inherently meaningless for this paper. The problem is what happens to the paper’s strategic question. A single user has zero mass in the continuum and cannot change aggregate allocations. Giving the deviators positive mass preserves their effect, but changes the question into coalition manipulation. The multiresource theorem studies a single user; the paper’s coalition bound is for single-resource weighted MMF. Neither result establishes a multiresource coalition bound.
One could make the proposed mass-transfer optimization precise, with a finite report universe and a fixed horizon. That would be a new computational problem about coalition strategy, not a continuous version of a computational question the paper posed or resolved. It may be worthwhile as a separate project, especially for large cloud systems, but it does not supply an anchor for this paper under ChoCo’s criteria.
So the strongest negative is a scope judgment, not a claim that no plausible continuous population model exists. The proponent’s candidate survives as a possible new direction; it does not survive as a mirror of a named computational result in this paper.
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.