| paper | d4-conf-aaai-2006-0018 |
| 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 supplier-cohort clearing problem is a coherent extension of the paper's allocation task. But the paper gives no numbered theorem, lemma, corollary, or proposition asserting the computational result that could anchor this mirror; its closest claim is an unnumbered assertion that cited hardness carries over.
fails bit a — no named computational result to mirror
The proposed mirror concerns market clearing over supplier cohorts; it leaves the paper's algorithm engineering, implementation, scenario navigation, and reported deployments unmirrored.
There is no numbered Theorem, Lemma, Corollary, or Proposition in this paper that asserts a computational complexity result. So there is no qualifying anchor under your rules, and I can’t honestly present a fully compliant positive case.
The closest claim is in the “Tree Search to Enable Expressive Commerce” section: the paper says that vanilla combinatorial reverse auction clearing is NP-complete and worst-case inapproximable in polynomial time, citing Sandholm et al. (2002), then says those properties carry over to the richer expressive commerce problem. That claim is not numbered; the underlying vanilla result is cited from elsewhere, while the transfer to expressive commerce is asserted here. I would not relabel it as a theorem.
As a provisional mirror of that claim, consider Continuous Expressive Clearing. Let \(J\) be a finite set of recurring item or service types, with rational demand masses \(d_j\), and let \(S\) be supplier types with rational masses \(\nu_s\), where \(\sum_s \nu_s=1\). A supplier type completely specifies a supplier’s offer language, costs, capacity, and operational constraints. For each type \(s\), let \(\Omega_s\) be the set of feasible configurations of one supplier’s bids, including any permitted partial acceptance. A configuration \(q\) provides \(a_{sjq}\) units of service type \(j\), costs \(c_{sq}\), and contributes \(h_{sq}\) to the buyer’s side constraints. The decision variable \(x_{sq}\geq0\) is the mass of suppliers of type \(s\) run in configuration \(q\).
The question is to minimize \(\sum_{s,q}c_{sq}x_{sq}\), subject to \(\sum_q x_{sq}\leq\nu_s\), meeting every demand \(\sum_{s,q}a_{sjq}x_{sq}\geq d_j\), and satisfying the buyer’s side constraints on the aggregate \(h\)-contributions. A solution is an optimal mass allocation \(x\), or an exact decision whether one exists below a given cost bound. This keeps package bids, supplier efficiencies, and buyer constraints in the clearing problem; it does not turn the task into outcome-space fractionalization alone.
A plausible regime is a large procurement event with many recurring freight or service requirements and many suppliers grouped by identical offer menus, costs, capacities, and operational attributes. Then the number of suppliers and shipments can be large while the number of distinct demand and supplier types is comparatively small. The type definition is demanding, but it is the right high-multiplicity definition: suppliers with different offers or capacities belong to different types. A sourcing author could recognize this as their same allocation question over supplier cohorts and demand shares.
For an explicitly listed set of configurations, the mirror is an LP and is tractable. With compact expressive bid languages, however, the configurations may be implicit and the central issue becomes pricing over feasible supplier configurations. My cautious expectation is that the unrestricted version remains hard, plausibly through that pricing problem; the paper itself does not establish this, and the discrete hardness does not automatically transfer because continuous supplier mass can split across configurations.
The soft spot is substantial: the paper’s hardness is driven by the combinatorics of items, bids, and side constraints, not by a stated high-multiplicity regime. The proposed cohort model may therefore retain the hard combinatorics rather than dissolve them. It is a concrete, plausible mirror of the paper’s unnumbered clearing claim, but it cannot serve as the required named-result anchor.
The strongest objection is that the proposed anchor is not a named result of this paper. Its NP-completeness and inapproximability claim concerns vanilla combinatorial reverse-auction clearing, cited from Sandholm et al. (2002). The paper then says those results carry over to expressive commerce, but gives no named or numbered result establishing that transfer. So the proposed continuous model is a new problem inspired by the paper, not a mirror of one of its named results.
The proposed model is still the right one to test. Treating suppliers as types defined by their full offer menus, costs, capacities, and constraints, then assigning mass of each type to feasible configurations, preserves package bids and buyer-side constraints. In a market with many interchangeable providers, it could be a sensible continuous clearing problem, and the authors might recognize it as a generalization of their allocation task. I cannot dismiss it because it uses fractional mass, or because the discrete problem might remain hard.
The best case against is that the paper gives little reason to expect the necessary supplier multiplicity. Its examples emphasize firm-specific routes, production efficiencies, creative offers, and contractual commitments. Under the programme’s complete-type definition, those differences can make nearly every supplier its own type; then \(\tau\) grows with the supplier count and the high-multiplicity regime loses its advantage. The paper’s reported scale comes from items, bids, and side constraints, not from large cohorts of identical suppliers. Recurring freight or service providers could supply such cohorts, but that is a plausible new regime rather than one the paper establishes.
That objection weakens the mirror’s connection to this paper, but it does not defeat it in every scenario. The proposed continuous clearing question could be worthwhile in the right sourcing market. The honest negative case is therefore narrow: this paper offers no named computational result to anchor the case, and its description gives limited support for a high-multiplicity supplier regime. I cannot honestly support the stronger claim that no worthwhile continuous mirror exists under any scenario.
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.