| paper | d4-conf-atal-2013-0274 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | no |
The paper’s named results characterize incentive and mechanism properties; none states an algorithmic guarantee or computational hardness result to anchor a computational mirror. The opponent also shows why Theorem 5’s individual false-name deviation does not carry over to type masses: a single account has zero mass, while positive mass changes the deviation into a coalition.
fails bit a — no named computational result to mirror
No named result supplies a computational anchor; the paper’s results concern strategy-proofness, false-name-proofness, and related mechanism properties.
A plausible high multiplicity setting is a large school or online course intake: a type is a student’s complete preference ranking and characteristic vector, and \(\mu_\theta\) is the fraction of applicants of that type. Scale each school’s or course’s places with the intake. Individuals still receive at most one indivisible place, while the assignment averaged over a type can be fractional. That is a sensible continuous matching model for this paper’s highly anonymous motivation.
The closest named result is Theorem 5, proved here: in directed environments—where characteristics form a DAG, agents may report weaker characteristics, and priorities respect that DAG—the TTC mechanism is false-name-proof. An agent cannot benefit by adding a fake account with an allowed report, whether it claims the place assigned to its real account or to the fake one. This is a mechanism-property theorem, not a result about a problem being in P, NP-hard, FPT, or otherwise computationally classified. The same is true of the paper’s other named results. So, under the requested anchor rule, this paper has no named computational result to anchor a continuous computational question on.
There is also a substantive obstacle to turning Theorem 5 into one. In an atomless society, a single fake account has zero mass and cannot affect the allocation. Giving the fake account positive mass would instead model a coordinated group or coalition, which changes the manipulation in the theorem. Thus there is a credible high multiplicity *setting*, but I cannot honestly supply the required continuous problem per anchor while preserving the paper’s question. Theorem 5 is the strongest near-miss, not a valid computational anchor.
Theorem 5 is the strongest candidate, but it establishes a strategic property: TTC satisfies false-name-proofness in directed environments. It does not classify the complexity of a computational problem. Theorem 1 and Theorem 6 likewise characterize mechanism properties; none supplies the computational result the programme is meant to continuize.
A large school or course intake can be a sensible high-multiplicity setting, especially if capacity scales with intake. But Theorem 5’s key question concerns one identified agent adding one fake account and comparing the object assigned to that agent or its fake account. In a population represented only by type masses, one account has zero mass and cannot affect the aggregate allocation. Giving the deviation positive mass instead models a coordinated group, not an individual false name. Keeping a named agent alongside a continuous background could preserve the deviation, but that is a hybrid model and a new computational question, not a mirror of the theorem.
One could define a mass-coalition version, or ask for algorithms to compute matching outcomes from type distributions. Either might be worth studying. But each requires choices about coalition utility, account splitting, and how DA or TTC operates over masses—choices the paper’s individual incentive theorem does not settle. The negative case is strongest under the requested anchor rule: no computational result has been anchored, and its strongest strategic result loses its central deviation when represented by masses. That does not show that every computational question adjacent to continuous matching is uninteresting; it shows that this paper provides no surviving computational anchor for one.
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.