| paper | d4-conf-aaai-1999-0088 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The paper’s numbered items are Definitions 1–6, and its contract comparisons are experimental findings without a numbered computational guarantee. The opponent proposes a plausible continuous study, but that cannot supply an eligible result from this paper to mirror.
fails bit a — no named computational result to mirror
The paper experimentally compares O-, C-, S-, and M-contracts on small multiagent TSP instances; none of those findings is an eligible anchored result. Its global-optimum statement is attributed to earlier work.
This paper contains no numbered Theorem, Lemma, Corollary, or Proposition asserting a computational complexity or algorithmic result, so it has no eligible anchor for the requested case. Its numbered items are Definitions 1–6. The paper’s own results are experimental comparisons of contract types on small multiagent TSP instances.
The statement that OCSM contracts reach a global optimum is attributed to Sandholm’s earlier work, and the TSP’s NP-completeness is background. Neither is a numbered result proved here. A continuous task reallocation question could be proposed, but it would not be anchored in a qualifying result from this paper.
The proponent raises no result-specific anchors to defeat; its only claim is that the paper has no numbered complexity theorem. That fact is true, but it does not rule out a worthwhile mirror. The paper’s central findings are computational experiments: how O-, C-, S-, and M-contracts compare in welfare, contracts tried, and time to reach or verify a local optimum.
A stronger mirror would group agents by their *complete* cost type, including their cost function and any relevant location or capabilities. Let \(\mu_k\) be the population share of type \(k\), and scale task supplies with the population. If \(z_{k,S}\) is the mass of type-\(k\) agents assigned bundle \(S\), the aggregate allocation has constraints \(\sum_S z_{k,S}=\mu_k\) and task-balance constraints, with objective \(\sum_{k,S} c_k(S)z_{k,S}\). This preserves integral bundles for each agent while allowing fractional type-level averages. It also gives a natural continuous question tied to the paper: how do the contract families’ local optima and welfare compare as a function of type support, task supply, and a normalized negotiation budget? Reallocative markets with large cohorts of agents sharing cost profiles provide a plausible regime for that question.
There is a real limitation: the paper’s anytime tradeoff is measured in discrete contract proposals and CPU time. With divisible mass, a contract might move an arbitrarily small or large amount; the paper supplies no canonical normalization for that, and different choices could change the comparison. A direct mirror of its timing curves would therefore need additional modelling decisions. But that weakens the time-and-contract-count mirror; it does not erase the aggregate welfare and local-optimality questions. Nor does the absence of a numbered theorem mean the paper has no computational findings to mirror. The strongest negative case is that this paper offers a less direct, less theorem-shaped target than the bribery examples. The claim that *no* scenario yields a worthwhile mirror is considerably weaker.
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.