| paper | d4-journals-mss-2002-0001 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→P discrete result → its continuous mirror
Answer: \(\mathrm P\): the exact minimum is \(\sum_j\mu_j r_j(1-r_j)\), where \(r_j=a_j-\lfloor a_j\rfloor\), and the decision takes \(O(\tau L^2)\) bit operations. · discrete side: other · adjudication: proved
full proof & adversarial review →
Theorem 1
Given demand types \(d_1,\ldots,d_\tau\in\mathbb{N}\) with atomless masses \(\mu_j>0\), \(\sum_j\mu_j=1\), and rational per-capita supply \(b\le D:=\sum_j\mu_jd_j\), find a symmetric law over measurable allocations that gives each type-\(j\) individual an integer bundle in \(\{0,\ldots,d_j\}\), allocates exactly \(b\) units per capita in every realization, and gives each type-\(j\) individual expected allocation \(a_j:=bd_j/D\). Minimize the population-weighted sum of individual variances over such laws.
Represent decisions by masses \(z_{jk}\) of type \(j\) receiving \(k\) units, with \(\sum_{k=0}^{d_j}z_{jk}=\mu_j\) and \(\sum_{k=0}^{d_j}kz_{jk}=\mu_ja_j\). Atomless type populations can be partitioned into these bundle groups and symmetrically randomized so every individual has the required marginal lottery while each realization uses exactly \(b\) units per capita. Minimize \(\sum_j\sum_{k=0}^{d_j}z_{jk}(k-a_j)^2\); the minimum puts each type’s mass on \(\lfloor a_j\rfloor\) and \(\lceil a_j\rceil\).
The atomless model removes the finite lottery-coupling feasibility problem addressed by the paper’s simplex construction, so the mirror mainly carries over its marginal conclusion; the proponent acknowledges the missing polynomial bound but does not fully address this loss of computational content.
fatal: False
The mirror covers Theorem 1’s minimum-variance construction for proportional rationing. It leaves the theorem’s cases for other classical rationing rules and the axiomatic characterizations in Propositions 2 and 3 aside.
The strongest honest positive case is narrow: the paper has no named theorem asserting a complexity class such as \(\mathrm{P}\) or NP-hardness. Its best computational anchor is Theorem 1, proved in this paper. It says that for a given classical rationing method \(r\) and discrete rationing problem, a probabilistic method with the same expected allocation as \(r\) is minimum-variance exactly when each individual’s realized allocation differs from their expected allocation by less than one. Thus, if \(r_i=\lfloor r_i\rfloor+u_i\), the individual’s marginal lottery gives \(\lfloor r_i\rfloor\) units with probability \(1-u_i\) and \(\lfloor r_i\rfloor+1\) units with probability \(u_i\). The theorem also requires a joint lottery that respects the total supply. Its proof constructs that lottery through a linear feasibility system over integral allocations and says it can be found by simplex. That makes it a constructive algorithmic anchor under the paper’s formulation exception, but it is not a polynomial-time result: the paper gives no such bound.
Lead mirror: Proportional Fair-Remainder Rationing on a Continuum. An instance has finitely many demand types \(d_1,\ldots,d_\tau\in\mathbb N\), rational masses \(\mu_j\ge 0\) summing to one, and a rational supply \(b\) per unit of population, with \(b\le D:=\sum_j\mu_jd_j\). Everyone of type \(j\) has the same demand \(d_j\) and is otherwise indistinguishable for the rationing problem. The proportional expected allocation per type-\(j\) individual is \(a_j=bd_j/D\).
The task is to give each individual an integral allocation between zero and \(d_j\), while allocating exactly \(b\) units per unit of population in every realized allocation, preserving expected share \(a_j\) for each type, and minimizing each type’s individual variance. A solution can be specified by masses \(z_{jk}\): the mass of type \(j\) assigned \(k\) units. They must satisfy \(\sum_{k=0}^{d_j}z_{jk}=\mu_j\) and \(\sum_{k=0}^{d_j}kz_{jk}=\mu_ja_j\). Minimize \(\sum_j\mu_j\operatorname{Var}_j\), where \(\operatorname{Var}_j\) is the allocation variance for a representative type-\(j\) individual. Because the population is atomless, each type’s mass can be split into the required outcome groups; randomizing that split symmetrically gives each individual the specified marginal lottery while preserving the exact aggregate supply in every realization.
The solution puts type \(j\)’s mass only on \(\lfloor a_j\rfloor\) and \(\lceil a_j\rceil\): the fraction \(\{a_j\}\) gets the latter and the rest the former. It therefore achieves the least possible variance for each type simultaneously. For fixed demand types, the formula is computable using rational arithmetic in time polynomial in \(\tau\) and the input bit length. I expect this mirror to be Class A, though that tractability prediction comes from the mirror’s rounding structure, not a complexity theorem in the paper.
The high-multiplicity regime is concrete: a large rationing program—such as distributing scarce water among households—has many claimants but a modest number of standard demand levels. A type is a demand level, with any other rationing-relevant features added to it if necessary. The population size \(n\) is much larger than \(\tau\); along instances where \(n\mu_j\) and \(nb\) are integers, this is the normalized version of the finite problem. Individual bundles remain integral. The fractional quantities are masses of people receiving each bundle and the supply per person, as required by scaling the rationing problem with the population.
The mirror covers Theorem 1’s minimum-variance construction, specifically its proportional-rationing case. It does not claim to continuize Propositions 2 or 3: those are axiomatic characterizations, not computational results. Further questions include whether analogous efficient constructions hold for other classical rationing rules, how exact finite-population rounding behaves when the mass proportions cannot be realized exactly, and how priorities or heterogeneous claims should be included in the type description.
The weak point is that the paper’s underlying rounding result is already easy to compute once expected shares are known, and its stated simplex construction has no polynomial-time guarantee. Also, the paper’s “classical” model already allows divisible individual allocations, whereas this mirror makes the population continuous and retains integral bundles for every individual. That distinction is defensible because the type masses, per-capita supply, and ex-post allocation over an atomless population are the substance of the mirror. Still, the result offers a modest tractable mirror, not a new hardness boundary or a strong computational-complexity anchor.
The strongest case against starts with the paper’s results. None asserts the complexity of a computational problem. Theorem 1 characterizes minimum-variance lotteries and gives a simplex construction for a fixed finite rationing problem; it supplies no running-time bound or complexity classification, and it takes the classical rationing rule as given. Propositions 2 and 3 are axiomatic characterizations. That leaves no computational result in the paper for a continuous mirror to carry over.
The proposed high-multiplicity setting is nevertheless sensible. Many claimants can share a complete demand type, and integral bundles alongside fractional per-capita supply are a natural co-modification. So multiplicity, type aggregation, and the model’s fit with the paper are poor grounds for rejecting the mirror.
The difficulty is what happens to the computational object. In the finite paper, a probabilistic rationing method is a distribution over complete integral allocations, whose coordinates must sum to the fixed supply in every outcome. The proposed continuum model instead chooses masses \(z_{jk}\) of type \(j\) receiving \(k\) units. An atomless population can be partitioned to realize those masses exactly; randomizing the partition can give each individual the desired marginal lottery. For proportional rationing, the minimum-variance marginals are already specified by Theorem 1’s floor-and-ceiling rule. The continuous formulation therefore has no remaining joint-feasibility problem of the kind handled by the paper’s simplex construction. It is a reasonable new mass-allocation model, but it does not mirror a computational result in this paper.
A more faithful finite high-multiplicity question could ask for a compact lottery over allocations when many people share each type. That would preserve the coupling issue, but it would remain a discrete high-multiplicity algorithm problem. Passing to atomless type masses removes the finite-count constraint that creates that issue. This gives the negative case its best argument: preserving the paper’s computational substance keeps the population finite; making the population continuous leaves no complexity result from the paper to study.
That argument is limited. Theorem 1 is constructive, and a compact high-multiplicity representation of its lotteries could be worth studying. More importantly, the proposed continuum question is coherent and the rationing setting admits a convincing large-population regime. I cannot honestly establish that no worthwhile mirror exists in any scenario; the case against this paper rests mainly on the absence of a complexity result to anchor one, not on a flaw in the proposed model.
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.