| paper | d4-conf-aaai-2014-0233 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | no |
Theorem 3 is a named welfare approximation guarantee for the stated mechanism, so it passes the computational-result screen. But in an atomless population, an individual's bid has no effect on the allocation and the individual's VCG externality payment vanishes, so the finite-player ℓ-strong equilibrium result does not transfer. A model with positive-mass coalitions or cohort-level charges would require a different strategic game, leaving no continuous mirror of these anchors that the authors would recognize as their problem.
fails bit b-changes — the continuum limit changes the question
The attempted mirrors concern Theorems 3 and 4; they leave Theorems 1, 2, 5, and 6 and Corollary 1 without a surviving population version.
I can make a population-based analogue of the paper’s equilibrium analysis, but I can’t give it a valid anchor under your rule: none of the paper’s numbered results establishes a computational complexity or algorithmic result.
The nearest are Theorem 3, proved here: for XOS valuations, every \(\ell\)-strong equilibrium of the item-bidding mechanism has social welfare at least \(\mathrm{OPT}/(1+\lceil n/\ell\rceil)\). Theorem 4, also proved here: for cL valuations, the worst-case welfare loss at an \(\ell\)-strong equilibrium is at least a factor \(\max\{2,n/\ell\}\). These are price-of-anarchy bounds, not complexity classifications. The CPPP’s NP-hardness for unit-demand valuations is attributed to Buchfuhrer, Schapira, and Singer; it is not a numbered result proved in this paper.
The tempting mirror would take valuation functions as types, let \(\mu_t\) be the fraction of agents with valuation \(v_t\), and choose \(k\) items from aggregate bids. One could ask how welfare loss at equilibria changes when coalitions may coordinate up to an \(\alpha\) fraction of the population. That is a plausible population version of the paper’s economic question. But it does not supply a qualifying anchor or a ChoCo-style computational question. There is also a substantive modeling snag: an individual’s VCG payment vanishes in an atomless population. Preserving meaningful payments for coordinating masses would require specifying a coalition-level payment rule, which the paper does not study.
So the strongest honest positive case is a nearby equilibrium-analysis mirror, not a continuous computational mirror anchored in this paper’s results. The paper’s named results do not support a prover-ready problem per the requested standard.
The proponent has identified the central problem: none of this paper’s numbered results is a computational result. Theorem 1 and Theorem 2 establish equilibrium existence; Theorems 3–6 bound welfare at equilibria. The CPPP hardness results are cited from other papers. There is no complexity classification or algorithm here to carry into a continuous population.
The strongest plausible mirror is the one the proponent sketches: a municipality with many residents grouped by complete valuation type, and a question about welfare when coalitions controlling an \(\alpha\)-fraction of the population can coordinate. That is a credible high-multiplicity scenario. It does not, however, preserve the game behind Theorems 3 and 4. In an atomless society, one agent—or any fixed finite coalition—has zero mass and cannot change the aggregate item scores. Removing one agent also leaves the outcome unchanged, so the individual VCG externality payment is zero. Each agent’s bid therefore has no effect on their allocation or payment: every bid profile is a Nash equilibrium. The finite-player equilibrium notion that supports those theorems has degenerated.
Allowing deviations by positive-mass coalitions avoids that problem only by changing the mechanism’s strategic model. The paper specifies VCG payments for individuals, not a payment rule for a coalition with positive mass. A coalition-level externality charge and its allocation among members would need new choices; different choices can define different games. Any resulting welfare or algorithmic question might be interesting, but it would be about a newly specified collective-action mechanism, not a continuous mirror of Theorem 3 or Theorem 4.
So the negative case is strong on the paper’s contribution, but I would not claim it proves that no worthwhile continuous model could ever be built. A large public-project population is a sensible regime. The narrower and better-supported conclusion is that this paper supplies no computational result to mirror, and its most tempting equilibrium results lose their strategic meaning under a faithful atomless limit.
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.