| paper | d4-conf-ecai-2025-0528 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | yes |
The paper’s named Propositions 1 and 2 establish semantic properties of its forecaster QBAFs, not a computational result. Its accuracy findings are empirical and do not provide a named computational guarantee. The proposed population filtering procedure may be recognizable to the authors, but it is unanchored and its typewise computation adds no computational result from this paper.
fails bit a — no named computational result to mirror
No named computational result is mirrored; Propositions 1 and 2 establish semantic properties, while the forecasting evaluations are empirical.
I don’t find a qualifying named computational result in the paper. The numbered results are Proposition 1, proved here, which says forecaster QBAFs satisfy Neutral Forecasting, and Proposition 2, also proved here, which says they satisfy Bias-based Forecasting. They establish semantic properties; neither classifies the complexity of a problem or gives an algorithmic result of the kind your anchor rule requires. The forecasting improvements in Sections 5 and 6 are empirical findings, not numbered results. So I can’t honestly attach an anchor-specific mirror to this paper.
The strongest positive possibility is an unanchored population version of the paper’s coherence filtering. Fix a debate framework \(\langle X,R\rangle\), thresholds \(\xi_1,\xi_2\), and a finite forecast grid \(G\). A type \(\theta\) is a complete vote vector \(V_\theta\) over the debate’s non-forecasting arguments, together with predictions \(P_\theta(f)\in G\) for each forecasting argument. Given a rational distribution \(\mu\) over the listed types, apply the paper’s forecaster-QBAF construction and Definition 2 to each type. Return the coherent mass \(W=\sum_{\theta\ \mathrm{coherent}}\mu_\theta\) and, when \(W>0\), each coherent-population forecast
\[
\bar P(f)=\frac{\sum_{\theta\ \mathrm{coherent}}\mu_\theta P_\theta(f)}{W}.
\]
A plausible regime is a large forecasting platform using shared debates and discretized forecasts: many participants, with repeated vote-and-prediction types. This directly extends the paper’s filtering and averaging procedure, and the authors could plausibly recognize it as their problem at population scale. Evaluating the listed types and aggregating their masses appears tractable.
But this is a proposed extension of the empirical procedure, not a mirror of a named computational result in the paper. Its weakest point is that its tractability comes from evaluating each supplied type and summing; it does not expose the kind of complexity boundary the ChoCo programme seeks to study.
There is no named computational result here to anchor a mirror to. Propositions 1 and 2 establish semantic properties of the forecaster QBAF; the accuracy findings are empirical. Neither gives a complexity classification or algorithmic result. The proposed population extension therefore has no result in the paper for it to mirror.
The strongest version of that extension is still a sensible population model: imagine a large platform where users assess the same debate, record votes and predictions, and many share the same response type. Give each complete vote-and-prediction type a mass, filter types by Definition 2, and average the predictions of the coherent mass. That respects high multiplicity; individual identity is not the objection.
The problem is that this computation just applies the existing filter type by type. For each supported type, construct and evaluate its forecaster QBAF, check the thresholds, then sum the masses and predictions. On the paper’s tree-shaped debate structures, this takes time polynomial in the support size and debate size. The support is supplied with the instance, so there is no hidden search over all possible types. The continuous model turns counts into masses, but it adds no computational question of the sort ChoCo is meant to chart.
A more ambitious mirror might ask how sensitive the aggregate forecast is to shifts in type mass, or how to choose coherence thresholds. Those questions require an added perturbation model, intervention costs, or an accuracy objective; the paper supplies none of these. With coherence fixed, reweighting types is an optimization over already-labelled masses. Threshold selection is an empirical forecasting question. Either could become a separate project, but neither is an algorithmic result of this paper.
So the case against a worthwhile ChoCo computational mirror is strong: there is no qualifying anchor, and the natural population extension is computationally routine. The honest limitation is that a mass-based version of the authors’ filtering procedure could still be useful for describing large forecasting platforms. That makes “no useful continuous extension whatsoever” too broad; it does not make this paper a promising source of computational mirrors.
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.