| paper | d4-conf-wine-2017-0013 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
Theorem 1 is an eligible computational anchor because it gives an LP formulation that the paper uses in its optimization procedures. The proposed mass-Wardrop question replaces a finite player’s deviation, including entry onto an unused fixed-cost edge, with a zero-mass deviation that changes no aggregate load. The entry quote does not resolve that mismatch, so the stated question changes the anchored problem rather than providing its continuous analogue.
fails bit b-changes — the continuum limit changes the question
A finite first user can be charged the full fixed edge cost, whereas a Wardrop infinitesimal changes no load, and the proposed quote at zero flow cannot reproduce that incentive while balancing the edge cost.
fatal: True
The proposed model addresses only fixed-forest enforceability from Theorem 1; it leaves Theorem 2’s forbidden-subgraph characterization, Theorem 3’s graph classes, and the price-of-stability results untouched.
The paper has no named theorem that classifies a problem as NP-hard, in P, or by another complexity class. My strongest eligible anchor is Theorem 1, proved in this paper and used in its characterization proof. It is the formulation result the prompt allows: for a fixed Steiner forest \(F\), with each player \(i\)’s unique path \(P_i\) in \(F\), the paper’s LP maximizes \(\sum_i\sum_{e\in P_i}\xi_{i,e}\). It requires that the shares on each used edge sum to at most \(c(e)\); for every player and alternative path \(P_i'\), the shares on edges abandoned by that player sum to at most the full costs of the newly used edges; and all shares are nonnegative. Theorem 1 says that \(F\) is enforceable by a budget-balanced, separable cost-sharing protocol exactly when an optimal LP solution fully pays every edge of \(F\). This is an anchor about the underlying enforceability problem, not about the encoding.
The regime I would use is a large network serving recurring connection requests from a few origin–destination (OD) types: for example, many households or businesses needing service between the same pairs of locations. A type specifies the complete information relevant to routing and payment, including its OD pair and any cost-relevant parameters. Many individual requests share each type, so the number of users \(n\) is much larger than the number \(\tau\) of distinct OD types. This is a credible high-multiplicity regime: the paper’s players are connection requests, and a large pool of users with repeated requests is a natural setting for them.
My lead mirror is Mass-Wardrop Enforceability. An instance gives an undirected graph \(G=(V,E)\), rational edge costs \(c_e\), a finite type set \(T\) with rational masses \(\lambda_t>0\) summing to \(1\), and a specified feasible routing \(f\). Here \(f_{t,p}\) is the mass of type \(t\) assigned to path \(p\), with \(\sum_p f_{t,p}=\lambda_t\); the load of type \(t\) on edge \(e\) is \(x_{t,e}=\sum_{p\ni e}f_{t,p}\). In applications, \(f\) would be a candidate routing on a minimum-cost network, whose cost is \(\sum_e c_e\,\mathbf 1[\sum_t x_{t,e}>0]\).
The question is whether there are nonnegative, edge-local per-unit charges \(r_{t,e}\) that make \(f\) a Wardrop equilibrium. “Edge-local” means the charge depends only on the loads of the types using that edge, not on named individuals or their routes elsewhere. Budget balance requires \(\sum_t x_{t,e}r_{t,e}=c_e\) on every used edge. Every path carrying positive mass of type \(t\) must have minimum total charge among that type’s available paths. Charges quoted for entering an unused edge are part of the protocol too; they can deter a deviation just as the paper’s protocol can charge a deviating player for newly used edges. A solution is a yes/no answer and a rational charge certificate at the target loads, which can be extended edge by edge into a separable rule. The action variable is the distribution of each type’s mass over paths; the implementability question holds that candidate routing fixed, just as Theorem 1 holds \(F\) and its paths fixed.
The continuous analogue of the paper’s LP maximizes \(\sum_{e:\sum_t x_{t,e}>0}\sum_t x_{t,e}r_{t,e}\), subject to each used-edge budget being at most \(c_e\) and all Wardrop path inequalities. The target routing is enforceable exactly when the optimum reaches the sum of costs of its used edges, so every such budget is tight. There may be exponentially many path inequalities, but for fixed \(f\) they can be separated by shortest-path computations. I would expect this fixed-routing problem to be Class A, tractable, with running time polynomial in \(|V|,|E|,\tau\), the bit length \(L\), and the explicit description of \(f\). That tractability is my prediction for the mirror, not a result claimed by the paper.
This mirrors Theorem 1’s fixed-forest enforceability question, not the paper’s forbidden-subgraph characterization or its results about graph classes. The paper’s discussion of nonatomic routing games makes the population regime recognizable to its authors, while its analytic neighbors do not answer this computational implementability question.
The weakest point is the change from individual pure Nash deviations to Wardrop deviations by infinitesimal users. A single user’s move does not change aggregate type loads, and the entry charge at an unused edge needs a convention; there is no automatic theorem identifying this model with the limit of the finite-player games. That is the main dictionary question the programme would need to settle. The proposed problem still keeps the paper’s network, OD requests, edge costs, separability, budget balance, and route incentives; the co-modification is that edge payments are balanced across type mass. I would therefore defend it as a plausible continuous mirror of Theorem 1, while treating a limit identity with the finite game as an open part of the case.
Theorem 1 is the strongest anchor, but it characterizes enforceability by an LP; it does not give a complexity result. Theorem 2 classifies graph topologies, also without a complexity claim. That leaves a possible question to formulate, but makes the paper’s computational contribution a thinner anchor than the proponent suggests.
The repeated-OD scenario is plausible, and for ordinary routing costs a Wardrop mirror would be natural. The obstacle is that this paper shares fixed edge costs. In its finite-player game, a unilateral deviation can make a player the first user of an edge, triggering a finite cost that the protocol must allocate. In the proposed continuum, an infinitesimal user changes no aggregate load. If an edge carries mass \(x>0\), exact budget balance requires \(\sum_t x_{t,e}r_{t,e}=c_e\). As \(x\) approaches zero, the per-unit charges needed to recover \(c_e\) diverge. At \(x=0\), a finite “entry quote” can be assigned, but a zero-mass deviation does not create the positive flow needed to balance the edge’s cost. That quote is therefore an extra rule at an invisible state, not the limit of the finite game’s budget-balanced charge for a first entrant.
There are sensible ways to repair this, but each changes the question. Keep positive-mass cohorts that can deviate together, and the strategic agents are those finitely many cohorts; that is a weighted finite-player game. Allow coalitions of mass to move, and the solution concept is no longer unilateral Nash enforceability. Scale edge costs with traffic or replace fixed costs with per-unit operating costs, and the resulting Wardrop pricing problem may be tractable, but it no longer studies implementation of an optimal Steiner forest under the paper’s cost model. None defeats the mismatch by choosing a better type space: it comes from the discontinuity between a single entrant and zero mass.
So the proponent’s LP may define a coherent problem about mass routing with edge-local tariffs, and the paper’s mention of nonatomic routing makes that neighborhood recognizable. But its shortest-path pricing result would concern that new tariff model, not a continuous-population mirror of Theorem 1. The negative case is strongest on this fixed-cost discontinuity; it is not airtight, because authors could still value the adjacent nonatomic problem on its own.
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.