| paper | d4-conf-ecai-2014-0223 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
NP-hard→NP-hard discrete result → its continuous mirror
Answer: \(\mathrm{NP}\text{-complete}\) · discrete side: hardness · adjudication: proved
full proof & adversarial review →
Proposition 6
statement extracted from the paper’s text layer
Given a finite category set \(O\), a policy word \(\Pi\) over types \(\{A,B\}\), masses \(\mu_A=\mu_B=1/2\), supply \(q_o=1/2\) for each \(o\in O\), rational additive valuations \(v_A,v_B\), and a feasible baseline mass flow with bundle distributions \(\beta_A,\beta_B\), does there exist a feasible alternative flow under the same policy and supplies with distributions \(\alpha_A,\alpha_B\) such that, for each type \(\theta\), there is a coupling \(\kappa_\theta\) of \(\beta_\theta\) and \(\alpha_\theta\) supported on pairs \((S,S')\) satisfying \(v_\theta(S')\ge v_\theta(S)\), with positive mass on strictly improving pairs? Each unit of mass receives an integral bundle \(S\subseteq O\); at a policy position, the called cohort's mass picks one available category, with mass tracked by bundle history and cumulative category use bounded by supply.
Two anonymous cohorts with masses \(\mu_A=\mu_B=1/2\) pick integral bundles from a finite set of course categories according to a shared policy word. Each category has mass supply \(1/2\), and each cohort has one additive valuation; feasible mass flows describe picks by history, while a coupling checks whether the final cohort distributions Pareto-improve on the baseline.
The case does not show that every permitted mass split at a cohort turn is a limit of a serial finite replication of the paper's picking process.
fatal: False
The mirror covers Proposition 6's no-transfer, no-side-payment Pareto-manipulation decision problem; it leaves single-agent manipulation, the transfer variants, and the price-of-manipulation bound untreated.
The strongest mirror I see is a high-multiplicity version of coalitional picking-sequence manipulation in a large course-allocation market. Imagine two large cohorts of students, \(A\) and \(B\), with the same additive course valuations within each cohort. The published policy is a finite word \(\Pi\) over \(A\) and \(B\); each occurrence gives that cohort its next picking round. Both cohorts coordinate to improve their members’ bundles, without trades or side-payments.
The paper’s anchor is Proposition 6, asserted in this paper; the proceedings version omits its proof for space and points to the authors’ long version for it. It states that deciding whether a successful manipulation exists without transfers or side-payments is NP-complete, even with two manipulators, additive preferences, and no non-manipulator. Here, success means that the coalition’s joint strategy Pareto-dominates sincere picking: every coalition member is weakly better off and at least one is strictly better off.
Call the mirror Continuous-Cohort CM-SIMPLE. An instance has a finite set \(O\) of course categories, the policy word \(\Pi\), two types with masses \(\mu_A=\mu_B=1/2\), and rational additive valuations \(v_A,v_B\). Give each category supply \(1/2\): this is the normalized supply obtained by scaling each original item to \(N\) copies for \(N\) students of each type. Each student still receives an integral bundle, a subset of \(O\); only cohort totals and averages can be fractional. Include a feasible baseline mass strategy \(\sigma^0\), with sincere cohort play under a fixed rationing rule as the canonical case.
A proposed strategy is a sequence of mass flows. At stage \(s\), if \(\Pi_s=\theta\), variables \(x_{s,S,o}\) specify how much mass of type-\(\theta\) students holding bundle \(S\) picks \(o\). Require \(o\notin S\), conserve mass across histories, and ensure cumulative use of each category never exceeds its supply. These flows yield a distribution \(\alpha_\theta\) over integral bundles. The strategy succeeds if, for each type, the baseline distribution \(\beta_\theta\) and \(\alpha_\theta\) can be coupled so that every matched student weakly improves according to \(v_\theta\), with positive mass strictly improving. A solution is the feasible flow and such a coupling. This preserves the paper’s question—whether coordinated picks can Pareto-improve on the baseline—while making the population and its supplies continuous.
The regime is a large market with many students but few preference-and-role types: for example, two large degree-track cohorts choosing from a shared elective catalogue under a repeated cohort schedule. The support is two, regardless of cohort size. This is a population continuization, with mass-valued supplies as the necessary co-modification; it is not merely a move to fractional outcomes.
I would provisionally expect this mirror to be Class A, with an LP or configuration-LP approach: the aggregate feasibility and Pareto-coupling constraints are linear, and each individual’s bundle remains a finite choice. The central algorithmic question is whether pricing over feasible pick histories can be done efficiently. The paper’s hardness may dissolve when a cohort can divide its mass across different integral strategies, but that is not automatic. I mirror only Proposition 6; I make no claim about the paper’s other results.
The weak point is the cohort lift. The paper has two named manipulators; this version has two large cohorts whose members can follow different strategies. That creates fractional mixtures the original problem does not have, and so the new problem is not simply the old one with larger voter counts. A referee could reject the lift if the cohort-level picking stage and rationing rule seem too far from a picking sequence. I think the case survives because the mechanism, policy word, additive preferences, coalition, and Pareto criterion all remain recognizable, and the change is exactly the high-multiplicity question: whether the same manipulation problem behaves differently when each role is occupied by a mass of interchangeable agents.
The strongest objection is that Proposition 6 concerns two identifiable agents taking turns in a serial process. Each agent’s choice depends on the remaining objects, and Pareto improvement is judged against the bundle that same agent receives under sincere play. Replacing each agent with a cohort changes how turns work: if a whole cohort picks at once, the mechanism now needs a rationing rule; if cohort members pick serially, the lifted policy must specify their order. The proposed flow model also needs to preserve who is being compared with whom. A coupling between baseline and final bundle distributions can otherwise permit reassignment that the original manipulation does not allow.
That objection points to real modeling work, but it does not defeat the best mirror. Course allocation is an unusually natural high-multiplicity setting. One can specify a large population with repeated preference types, a replicated turn schedule, per-capita supplies, and mass tracked by type and picking history. The mass model can preserve individual bundles and require weak improvement for every represented agent, with strict improvement for positive mass. Those choices define a recognizable computational problem, and the necessary changes to supply and turn structure fit the programme’s scope.
Proposition 6 is a named computational result, and the paper itself cites course allocation as an application. The paper supplies no continuous computational treatment of this manipulation problem. Any failure of the continuous version to retain the original complexity would be a result to discover, not a reason to reject the question. I can identify a lift that needs careful definition, especially around turn order and Pareto comparison, but I cannot make a convincing case that every plausible lift is unworthy of study. The proponent’s anchor survives.
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.