| paper | d4-conf-aaai-2010-0030 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | medium |
| authors would recognise it | yes |
other→P discrete result → its continuous mirror
Answer: Under the stated score-gap promise, the exact SR flow is polynomial-time computable in \(m,\tau,L\), and it always equals fluid APSD. · discrete side: other · adjudication: proved
full proof & adversarial review →
Theorem 2
statement extracted from the paper’s text layer
Given a rational distribution \(\mu\) over types \(\theta=(a,b,\phi)\), per-capita inventories \(q_j\), an event and tie-breaking order, and the score-gap condition from Theorem 2, compute the mass flow \(y_{\theta j}\) when infinitesimal agents arrive and best respond under SR as available inventory changes; determine whether this flow equals the fluid APSD flow.
A continuum of agents with type masses \(\mu_\theta\), where each type specifies arrival, departure, and a strict ranking. Resource class \(j\) has per-capita supply \(q_j\); its consensus score is \(S_j=\sum_\theta\mu_\theta r_\phi(j)\). Each arriving infinitesimal agent chooses a report that minimizes its reported rank minus \(S_j\) among classes with positive available mass, and assigned mass remains allocated until departure.
The fluid extension needs a specified order for simultaneous positive-mass arrivals; the paper's theorem does not itself define that convention.
fatal: False
The mirror covers Theorem 2's all-strategic outcome identity and, locally, Lemma 2's best-response procedure. It leaves the strategyproof-mechanism characterization and the finite-population welfare simulations outside the mirror.
The strongest defensible mirror is about the SR mechanism’s manipulation problem. The paper does not prove a named NP-hardness, P, or parameterized-complexity result, so this is a modest algorithmic case, not a claim that it opens a new hardness landscape.
A plausible high-multiplicity setting is a large carpool or community-resource allocation system. There are many commuters, but relatively few complete types: each type records an arrival and departure window and a strict ranking of the available resource classes. The population is a rational distribution \(\mu\) over those types, with support \(\tau\ll N\), where \(N\) is the number of commuters. Each person still receives one indivisible resource; the mass assigned to a resource class can be fractional. To preserve scarcity as the population grows, give class \(j\) a per-capita supply \(q_j\), or \(Nq_j\) copies for a population of size \(N\). This is a necessary co-modification because the paper’s fixed stock of items cannot serve an unbounded population.
Lead mirror: Continuous Population SR Best Response. The anchor is Lemma 2, proved in this paper. It says that, given the current available items and fixed item scores, the paper’s procedure returns an agent’s best-response preference report. It tries the agent’s true choices from most to least preferred; for a candidate item, it puts that item first, places unavailable items after it, and orders the remaining available items by ascending score. It returns the first report that makes SR assign the candidate.
The continuous instance has \(m\) resource classes and a finite support \(T\) of types \(\theta=(a,b,\phi)\), with rational masses \(\mu_\theta\) summing to \(1\). Here \(a,b\) are the arrival and departure times and \(\phi\) is a strict ranking. The population consensus score for class \(j\) is \(S_j=\sum_{\theta\in T}\mu_\theta r_\phi(j)\), matching the paper’s expected-rank score. The instance also gives the per-capita inventory and the residual available mass \(v_j\) at the focal agent’s arrival. The question is: for an infinitesimal agent of type \(\theta^\star\), find a report \(\psi\) and the resulting class \(j\) that minimize the agent’s true rank \(r_{\phi^\star}(j)\), subject to \(j\in\arg\min_{\ell:v_\ell>0}(r_\psi(\ell)-S_\ell)\). Fix a tie-breaking rule as part of the instance. A solution is the report and the item class it induces.
Lemma 2 gives a polynomial-time method for this mirror. It checks at most \(m\) candidate items and constructs and evaluates each report using a ranking of at most \(m\) items. Computing the scores from the support costs \(O(\tau m)\) rational operations; the remaining work is polynomial in \(m\) and the input bit length \(L\). The paper proves the best-response property; the polynomial bound is an inference from its stated procedure.
A second, more population-level question is supported by Theorem 2, also proved here. The theorem says that, when adjacent item scores differ by less than \(1\) and all agents optimally manipulate, SR produces the same allocation as APSD. Its continuous mirror is All-Strategic Fluid SR: given \(\mu\), per-capita inventories \(q_j\), arrival and departure times, and the theorem’s score-gap condition, compute the resulting mass flow from types to resource classes under best responses, and determine whether it equals the fluid APSD flow. A solution gives the mass \(y_{\theta j}\) of each type assigned to each class, with capacity respected while each assignment is held until departure, plus reports that induce the flow.
I expect this second problem to be tractable. Process arrivals and departures in order, applying the best-response rule as inventory changes. A cohort may split across classes as stock is depleted, but it can cross at most \(m\) such boundaries. The score-gap condition makes each infinitesimal agent’s best response select its most-preferred available class, so the resulting flow is precisely the fluid APSD flow. This is a proposed continuous extension, not a theorem stated by the paper.
The authors would likely recognize both mirrors: they retain the same arrival-priority setting, fixed population-consensus scores, strategic preference reports, and rank-based utility. The original paper already uses a distribution over preferences to calculate scores, but its assigned population is still a finite list of agents. Here the distribution describes the society itself, and assignments become mass flows. The continuous problems cover the SR manipulation results; they do not claim to mirror the characterization of strategyproof mechanisms or the paper’s finite-population simulation results.
The weakest point is that the lead question remains a focal agent’s best response in a state summarized by available inventory. Continuization does not make that calculation newly hard or reveal a hardness barrier; Lemma 2 already solves its discrete counterpart efficiently. The fluid outcome question is also likely a direct extension. Further work could ask for finite-\(N\) rounding guarantees, or for the fluid rank-efficiency curve as the strategic share varies. The positive case is that the paper’s population-consensus scores and dynamic assignment make a mass-flow model natural, and its named manipulation results give precise, tractable questions to carry into that model.
The negative case is that neither anchor identifies a new computational problem about a continuous society. Lemma 2 is the one genuinely algorithmic result: it finds a single agent’s best report from that agent’s ranking, the currently available items, and the score vector. In the proposed mirror, the population distribution \(\mu\) matters only insofar as it computes the \(m\) scores \(S_j\). The focal agent then consumes zero mass, so the question does not compute how a continuous population is assigned or how its reports affect future availability. Making the focal agent consume a positive amount would restore that effect, but would turn the question into a cohort or coalition deviation, which Lemma 2 does not characterize.
A better population mirror would specify mass flows through per-capita inventories as types arrive and depart. That is a coherent extension, and the assignment problem does admit plausible high-multiplicity regimes. But Theorem 2 gives the all-strategic flow almost pointwise: under its score-gap condition, each infinitesimal agent’s best response selects their most-preferred currently available class. Applying that rule across arrival and departure events yields the fluid APSD flow. A version with both truthful and strategic mass could give an exact fluid welfare curve, but its central task would be evaluating a specified dynamic rule, not solving a new optimization or pricing problem. A positive-mass type that coordinates on one report could affect inventory, but that changes the strategic model from individual best responses to cohort manipulation.
So the strongest case against is about fit and value, not whether the fluid model can be made coherent: the paper’s core results concern mechanism properties and welfare simulations, while its best-response procedure already has a direct finite description. Its natural population extension looks more like fluid evaluation of that mechanism than a new computational-complexity landscape. This case is not airtight. An exact fluid welfare analysis and finite-population rounding results could be worthwhile; the paper’s dynamic resource setting is a real vulnerability in the negative argument.
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.