It is the same labels-are-expensive world, but now you hold the pen: you may choose which thirty of the thirty thousand points get sent to the expert for labeling, and you may choose them adaptively, each new label informing the next request. A point deep inside a cluster teaches you little; a point near the uncertain boundary can be worth dozens of random ones. Design the querying strategy and prove it beats labeling thirty points at random. The guarantee must be provable, not merely plausible. Get it wrong and you spend your tiny, precious label budget on redundant points and learn no faster than blind sampling — when every label costs an expert's hour, choosing the right questions is the entire game, and 'provably better than random' is the bar.
Clauset's toolkit — community detection, rigorous maximum-likelihood fitting of power laws, network sampling — comes from 2000s-2010s network science, where the analogous question is which nodes or edges to probe in a partially observed graph. That gives him genuine sensitivity to sampling design and to the ways naive sampling biases inference, and his 2009 power-law methodology paper shows he takes 'provably better than eyeballing' seriously. But his guarantees are statistical goodness-of-fit arguments on fixed data, not label-complexity bounds for adaptive querying. He has no version-space machinery, no disagreement-based analysis, and no PAC framework for showing thirty chosen labels beat thirty random ones. He could design a sensible cluster-aware sampling heuristic; the proof the problem demands lies outside his published methods.
Vaswani's contribution — the transformer architecture in 'Attention Is All You Need' (2017) — solved a representation and parallelization problem, not a data-selection problem. The transformer's historical role in the label economy is indirect: it enabled the pretraining regimes that made labels cheaper to exploit, but that story belongs to the self-supervision line, and it arrives a decade after this problem's 2000s setting. Nothing in his published work concerns adaptive querying, label-complexity bounds, or the version-space geometry that active learning theory turns on; attention weights select among tokens already in the input, which is a superficially similar phrase describing an unrelated mechanism. Asked to design a querying strategy and prove it beats random, he would have architecture skill and no relevant theory. His toolkit largely does not transfer.
Battle #118 · 8/10/2026, 11:38:45 AM · this result is deterministic: the same two personas on this problem always resolve the same way.