| paper | d4-conf-atal-2021-0080 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The proposed finite-support \(\mu\)-question maps exactly to the paper’s integer-count model, and denominator clearing gives the reverse mapping; common scaling preserves every party’s best response. The theorem’s hardness would therefore carry over, but the paper already studies this population model in an equivalent encoding. Under the stated collision rule, that makes the mirror red.
fails bit b-already — the paper is already native to the continuum
The exact conversion between integer histograms and rational masses shows that the stated mirror adds no distinct population model; the proponent noticed the normalization concern but did not resolve this equivalence.
fatal: True
The proposed question covers Theorem 3.14’s multi-party pure-NE existence result; it leaves Theorem 3.7’s two-party linear-time result and the structural existence results untouched.
The strongest mirror is Continuous-Mass Nominee-Game NE Existence, anchored in Theorem 3.14, proved in this paper. The theorem says that, given a line \([0,k]\), a voter distribution \(f:[0,k]\to\mathbb{N}_0\), and parties with finite nominee sets, deciding whether the resulting game has a pure Nash equilibrium is NP-complete. It also says the problem remains NP-hard when each party’s nominee set is bounded by any fixed constant of at least five. A yes-instance has a nominee profile in which no party can improve its utility by switching nominees unilaterally.
For the continuous-mass version, an instance has the same line and nominee sets \(P_1,\ldots,P_n\), plus a rational probability distribution \(\mu\) over voter locations. Its support \(S\subseteq\{0,\ldots,k\}\) lists the \(\tau\) occupied locations and their masses \(\mu_x\), with \(\sum_{x\in S}\mu_x=1\). Parties still choose one nominee each. Given a profile \(c\), let \(N_x(c)\) be the parties whose nominees are closest to location \(x\). Party \(i\)’s utility is
\[
u_i^\mu(c)=\sum_{x\in S}\mu_x\frac{\mathbf{1}[i\in N_x(c)]}{|N_x(c)|}.
\]
The question is whether there is a profile \(c\) such that, for every party \(i\) and every alternative nominee \(a\in P_i\), \(u_i^\mu(c)\ge u_i^\mu(a,c_{-i})\). The instance is encoded by the nominee sets and the binary rational masses; \(n,k,\tau\), and the mass bit length describe its size.
A plausible regime is a large electorate whose members fall into a modest number of ideological location types: for example, millions of voters represented by fractions at a few dozen or hundred positions on a political spectrum. Voters at the same location are interchangeable for this game. Parties remain individual strategic players with their own finite candidate pools; it is the voter population that becomes mass. This preserves the paper’s nominee-selection question and its nearest-candidate utilities. The expected complexity result is hardness transfers (Class B): given rational masses, clearing denominators produces integer counts \(f\), and multiplying all utilities by the same positive number leaves every best response unchanged.
That equivalence is also the weakest point. The paper already represents voters by an aggregate integer histogram \(f\), so this mirror may look like a normalization of its existing model rather than a distinct continuous problem. Moreover, Theorem 3.14’s 3SAT reduction puts its combinatorics in the number of parties and their nominee sets; it does not rely on a very large voter population. The result makes a credible case that continuous voter mass is a sensible formulation of the authors’ game, but it is a modest case for a new computational landscape. It covers only multi-party pure-NE existence, and it suggests follow-up questions about fixed-parameter tractability in the number of parties, restrictions on the voter-mass distribution, and extensions to continuously distributed voter locations.
Theorem 3.14 is the proponent’s strongest anchor, but its proposed continuous-mass version is already the paper’s game in different units. The paper gives voters by a count \(f(x)\) at each line position, and a party’s utility is a sum of those counts times the relevant voter shares. Set \(\mu_x=f(x)/V(f)\), where \(V(f)=\sum_x f(x)\). Then \(u_i^\mu(c)=u_i^f(c)/V(f)\), so every profile has exactly the same best responses and is a Nash equilibrium in one formulation exactly when it is in the other.
The converse also holds: given rational masses \(\mu_x\), clear their denominators to obtain integer counts \(f(x)\). This takes only polynomially many bits in the mass encoding, and scaling all utilities by the resulting common factor leaves equilibria unchanged. Thus the proposed question has the same instances and equilibrium predicate, up to polynomially equivalent encodings. The 3SAT construction in Theorem 3.14 already uses integer counts, so it translates directly. This is not an objection that the continuous answer would be the same; it is a collision because the paper already aggregates voters by type and studies the resulting computational problem. Its histogram is already a population distribution, with integer counts rather than normalized rational masses.
A more ambitious variant could use an atomless distribution of voter locations over a real interval. That would change the type space and require a finite input format for the density; the paper itself points toward real intervals. I cannot honestly rule out that broader computational question as worthwhile. But it is a distinct extension, not a stronger version of the proponent’s finite-support mirror of Theorem 3.14. The negative case is strong against that anchor as stated, and weak as a universal claim about every spatially continuous model.
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.