| paper | d4-conf-ijcai-2022-0516 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The numbered results establish mechanism properties and incompatibilities, but do not assert a computational result to continuize. The food-bank scenario is plausible, yet aggregating students with network roles would require new semantics for invitations and critical paths. That does not provide a continuous analogue of a computational result in this paper.
fails bit a — no named computational result to mirror
No computational result is mirrored; Theorems 1 and 8 concern axiomatic properties, as do Theorems 9 and 10.
The best positive intuition is a large, recurring food-bank matching program: students could represent people seeking aid, colleges could represent providers, and mass could represent the share of applicants with a given preference and network role. That is a plausible high-multiplicity setting if the number of such types stays small. But it is only a motivation, not a valid anchored case.
The paper has no named computational result of the required kind. Its numbered theorems concern which mechanism properties can coexist: for example, Theorem 8, proved here, shows that OSCO satisfies SP, FRN, and MB-C; Theorem 1 shows that no mechanism satisfies both PE and FRD. Neither classifies the complexity of a computational problem. Turning Theorem 8 into a continuous-population mechanism-design question would therefore introduce a new question, not mirror a computational result in this paper.
That gap matters especially here because invitations and diffusion depend on individual network paths: aggregating people into types may erase who can disconnect whom. So I cannot give a legitimate lead anchor or a prover-ready continuous problem per anchor. The paper’s results are about axiomatic mechanism design, rather than the computational complexity landscape this programme asks us to mirror.
The case against is stronger than the proponent’s: there is no computational result here to mirror. The numbered theorems establish compatibility or impossibility among axiomatic properties of mechanisms. The paper does not classify the complexity of finding a matching, deciding whether a mechanism with given properties exists, or solving an optimization problem. Executing OSCO, SeqDA, or SeqTTC is not the result the paper proves. Turning one of its claims into a computational question would create a new project, rather than continuize a computational result from this paper.
The closest candidate is Theorem 8: OSCO satisfies SP, FRN, and MB-C. But those properties rely on individual students’ positions in a particular network. The paper’s type is \((\succ_s,r_s)\), where \(r_s\) is a set of named neighbors; the DCT, safe students, and the effects of not inviting a neighbor depend on exact paths and critical parents. A food-bank population with repeated preferences might have small preference support, but that does not give it small support over these network roles. Keeping the named network preserves the relevant incentives while making the type count grow with the population. Aggregating to groups and links between groups instead loses the unilateral actions and path changes that SP and MB-C evaluate. Restoring them as group deviations changes the incentive question.
Theorem 1’s PE–FRD incompatibility has the same problem. Its finite counterexample uses particular students’ priorities, preferences, and distances. Replicating it as a distribution over types might preserve an allocation pattern, but the result remains an axiomatic incompatibility, not a computational classification. Theorem 9 and Theorem 10 are no better anchors: their claims about SeqDA and SeqTTC are also axiomatic, and their sequential steps use students’ individual distances from the moderator.
A more elaborate continuum network model could be interesting in its own right. But it would need to choose how individual invitations and critical paths work in a population of mass, or replace them with aggregate deviations. Either way, it adds the central mathematical object and question that this paper does not supply. A plausible high-multiplicity food-bank scenario supports studying matching with similar agents; it does not turn this paper’s property theorems into computational results. So the negative case is firm within ChoCo’s stated remit, though it is not a claim that continuous network matching could never be worthwhile as a separate research programme.
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.