| paper | d4-conf-wine-2009-0011 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
Answer: Exactly polynomial-time computable in \(m,\tau,L\). All limits \(a_t\) exist as finite rational numbers; moreover, \(N\nu_N(t)\) is constant throughout the bridge-free tail. · discrete side: polynomial · adjudication: proved
full proof & adversarial review →
Theorem 5
statement extracted from the paper’s text layer
Given fixed sites \(V\), a connected set of corridor types \(T\subseteq {V\choose 2}\), and positive rational masses \(\mu_t\) summing to \(1\), form \(G_N\) with \(N\mu_t\) parallel edges of type \(t\), for integers \(N\) making all multiplicities integral. In the spanning-connectivity game on \(G_N\), let \(\nu_N(e)\) be the nucleolus allocation to an edge of type \(t\). Do the limits \(a_t=\lim_{N\to\infty}N\nu_N(e)\) exist for every type, and can their exact values be computed in time polynomial in \(|V|,\tau,L\), where \(\tau=|T|\) and \(L\) is the mass encoding length?
A high-multiplicity spanning-connectivity game: each type \(t\) is a corridor between two fixed sites and has mass \(\mu_t\); the finite approximants contain \(N\mu_t\) interchangeable edge-players of that type. The decision object is the per-edge nucleolus allocation scaled by \(N\), and the output is its limiting type density \(a_t\), with \(\sum_t\mu_t a_t=1\).
The mirror covers Theorem 5’s nucleolus allocation and its interpretation for a population of links, with Proposition 6 connecting that allocation to the spanning-connectivity cooperative game. It leaves the wiretap value, the full maxmin-polytope characterization, and other equilibrium refinements outside its scope.
The best case I can stand behind is a high-multiplicity spanning-connectivity nucleolus. It uses links as the population: a regional network has a fixed set of exchange sites, but very many interchangeable circuits between them. This is a plausible regime when the number \(N\) of circuits is much larger than the number \(\tau\) of distinct corridor types.
The anchor is Theorem 5, proved in this paper. For a connected graph with \(\mathrm{opt}<1\), form layers \(L_1,\ldots,L_\ell\) by repeatedly removing sinks from the paper’s parent-child relation on the prime-partition blocks. With \(\kappa=1/\sum_{i=1}^{\ell} i|L_i|\), Theorem 5 says the nucleolus assigns \(\nu(e)=i\kappa\) to each edge in \(L_i\), and zero to edges outside the layers. The paper also proves that this strategy is the nucleolus of its spanning-connectivity cooperative game (Proposition 6). The paper gives no named hardness result to anchor a continuous hardness question, so this algorithmic result is the strongest fit.
Here is the mirror. Fix at least three sites \(V\), a connected support graph with edge types \(T\), and rational masses \(\mu_t>0\) summing to \(1\). Type \(t\) is a corridor between two sites; all its circuits have the same role in the game. For each \(N\) divisible by the denominators of the \(\mu_t\), build \(G_N\) with \(N\mu_t\) parallel circuits of type \(t\). Treat each circuit as a player in the paper’s spanning-connectivity game: a coalition is worth \(1\) exactly when its circuits connect all sites. Its allocation is the nucleolus, which lexicographically maximizes the ordered vector of coalition excesses.
High-Multiplicity Spanning-Connectivity Nucleolus asks: given \((V,T,\mu)\), does the per-circuit nucleolus allocation have a limit, and what is its exact type-level density? A complete solution outputs \(a_t=\lim_{N\to\infty}N\nu_N(e)\) for each type \(t\), where \(e\) is any circuit of type \(t\). Uniqueness of the finite nucleolus and symmetry among parallel circuits make that coordinate well defined. The density is normalized by \(\sum_t\mu_t a_t=1\). The mass \(\mu_t\) is the share of circuits of type \(t\); \(a_t\) is the limiting award per circuit after scaling away the \(1/N\) size of an individual award.
The finite theorem suggests a concrete answer: if the edge layers stabilize into type layers \(\mathcal L_1,\ldots,\mathcal L_\ell\), then \(a_t=i/\sum_j j\mu(\mathcal L_j)\) for \(t\in\mathcal L_i\), and \(a_t=0\) for types corresponding to the degenerate set. The continuous problem is to prove that this weighted layering is well defined and stable, and compute it exactly in time polynomial in \(|V|,\tau,L\), where \(L\) is the bit length of the masses. I expect tractability: the paper’s layered formula reduces the target to a weighted version of its graph decomposition, and the number of types can stay small while the circuit population grows.
The mirror covers the paper’s cooperative-game nucleolus result and its associated link-population allocation. It does not claim to continuize every equilibrium refinement in the paper. A key weakness is that links-as-agents is a narrower population story than voters-as-agents, and endpoint pairs are valid types only when circuits with the same endpoints really are indistinguishable for the game. Also, a naïve continuum cooperative game can behave badly: a connected coalition could use arbitrarily small mass from each needed type. This question therefore defines the continuous object as the high-multiplicity limit of the paper’s discrete games, with per-circuit awards scaled by \(N\). That scaling preserves a meaningful allocation density, but proving that the layers stabilize is part of the problem, not something Theorem 5 already establishes.
The only serious objection I see is to a literal atomless nucleolus. Represent each corridor type \(t\) by mass \(\mu_t\), give it an allocation density \(x_t\), and let a coalition choose any submass \(q_t\in[0,\mu_t]\). If a connected set of corridor types spans the sites, it can form a winning coalition using arbitrarily small positive mass from each. Its allocation tends to zero while its worth remains \(1\), so its excess tends to \(-1\). The least-core value is therefore \(1\) for every normalized allocation, and the finite game’s least-core selection loses its force. Connectivity jumps when mass changes from zero to positive.
That defeats a direct continuum version of Proposition 6’s nucleolus, but it does not defeat the stronger mirror the proponent proposes. Their scaled limit \(a_t=\lim_{N\to\infty}N\nu_N(e)\) deliberately retains the finite games’ \(1/N\) granularity before taking the limit. That is consistent with the programme’s high-multiplicity bridge, and parallel circuits of the same endpoints are a plausible interchangeable type. Theorem 5 gives a named computational result to mirror; the fact that the limiting density may follow from a weighted layering does not make the question invalid.
So the negative case is weak where it matters: it shows that the atomless cooperative game is a poor formulation, but does not show that the high-multiplicity limit is unworthy. I cannot honestly rule out this mirror—or claim that no scenario makes it worthwhile.
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.