| paper | d4-conf-aaai-2026-4556 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→other discrete result → its continuous mirror
Answer: The sharp uniform factor is \(\epsilon^*(m,\tau,Q)=1\) for \(\tau=1\), and \(\epsilon^*(m,\tau,Q)=0\) for \(\tau\ge2\), for every allowed \(m,Q\). · discrete side: other · adjudication: proved_with_minor_gaps
full proof & adversarial review →
Theorem 2
statement extracted from the paper’s text layer
Given a finite-support distribution \(\mu\) over buyer valuation schedules and referral roles, a measurable block or graphon referral relation, and a specified supply scaling, determine the largest \(\epsilon\) achievable by a measurable allocation-and-transfer mechanism that is individually incentive compatible and individually rational, respects diffusion feasibility and supply, and gives each truthful buyer utility between \(\epsilon\phi_i\) and \(\phi_i\), where \(\phi_i\) is the buyer’s expected marginal welfare contribution under a random-priority ordering of the population.
A continuum buyer population with mass assigned to complete valuation-and-referral types, a network relation preserving individual referral effects, and mass-scaled supply. The mechanism assigns each buyer an integral bundle and transfer, with type-level averages allowed; its objective is to maximize the guaranteed fraction \(\epsilon\) of each buyer’s Shapley contribution subject to incentive compatibility, individual rationality, and diffusion feasibility.
The proposal does not specify how an individual buyer’s unilateral referral report makes positive mass reachable while preserving individual incentive compatibility; aggregating referral roles could instead turn an individual action into a cohort action.
fatal: False
The mirror covers the Shapley-fairness guarantee for homogeneous-item PDA and could extend to CPDA with an appropriate supply and fairness scaling. It leaves aside the standalone IC and IR result, the revenue and non-deficit discussion, and comparisons with earlier mechanisms.
The strongest positive case is a plausible high-multiplicity regime. In a large online marketplace, many buyers might share one of a small number of valuation schedules and referral roles. Define a type as a buyer’s complete valuation schedule and role in the referral network; let \(\mu_t\) be the fraction of buyers of type \(t\). Then \(n\) could be very large while the support size \(\tau\) remains moderate. A continuous version would assign each type an allocation and transfer under valuation and referral reports, and ask how much Shapley fairness can be guaranteed subject to incentive compatibility, individual rationality, supply, and diffusion feasibility. The authors could recognize the motivation: their own model measures buyers’ contributions through the welfare enabled by their participation. Modeling referral roles in aggregate, and scaling supply if needed, would be substantial co-modifications.
But I cannot give that candidate a qualifying anchor under your rule. The closest numbered results are Theorem 1, proved here: PDA is individually rational and incentive compatible; Theorem 2, proved here: PDA is \(1/(k+1)\)-Shapley fair when selling \(k\) homogeneous items; and Theorem 3, proved here: CPDA is individually rational, incentive compatible, and \(1/n\)-Shapley fair. None asserts a computational complexity classification such as membership in P or NP-hardness. The paper does say in the text that PDA can be run in polynomial time, using one random order and greedy allocation, but that is not a numbered complexity result of the required kind.
So there are no qualifying anchors and, consequently, no continuous problem per anchor to hand to a prover. The proposed mass-mechanism question is a research direction, not a faithful computational mirror of one of those numbered results.
The main weakness is the network process itself. A buyer’s invitation can make other buyers feasible, and PDA relies on a random order of discrete agents. Replacing individuals with mass requires specifying how referral roles and pivotal invitations behave in the limit; otherwise the model may erase the incentive that the paper studies. My positive case is therefore about the plausibility of a high-multiplicity *scenario*, not a demonstrated continuous computational question.
The strongest negative is that the paper has no named computational result to mirror. Theorem 1 proves individual rationality and incentive compatibility; Theorem 2 gives PDA’s \(1/(k+1)\)-Shapley-fairness guarantee; Theorem 3 gives CPDA’s \(1/n\) guarantee alongside those properties. None classifies the complexity of a computational problem. The paper’s unnumbered observation that PDA runs in polynomial time is the nearest candidate, but it describes one implementation using a random order and greedy allocation. A continuous version of that observation would not supply the complexity question this programme is looking for.
The proponent’s best repair is to ask for the largest Shapley-fairness guarantee achievable by a continuous-population mechanism subject to incentive compatibility, individual rationality, and diffusion constraints. That could become an interesting computational problem, but it is a new mechanism-synthesis problem, not a mirror of any of these theorems. It also needs a new account of network structure. In the paper, an individual’s report can make particular neighbors feasible, and a single buyer may be pivotal to reaching others. Aggregating buyers by valuation and referral role does not by itself preserve those relations. In a sparse network, individual buyers have vanishing mass; in a block model where a whole cohort invites another cohort, the strategic action has become a mass-level action. Either route needs substantial new modeling before there is a computational question to classify.
That network objection is not airtight: a carefully chosen high-multiplicity block-network regime could be plausible. So I would not claim the paper’s topic admits no worthwhile continuous research. The firmer conclusion is narrower: Theorems 1–3 do not provide the required computational anchors, and turning them into continuous mechanism-optimization questions would create new research problems rather than continuize the paper’s computational results.
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.