How many quant-ph papers declare a classical baseline?
TL;DR (status as of 2026-10-27): Almost none, measured today. Rosetta Quantum screened the abstracts of the first 50 of the 98 new quant-ph submissions announced on arXiv on Tuesday 27 October 2026 (listing, consulted 2026-10-27). Two abstracts — 4.0%, Wilson 95% interval 1.1–13.5% — declare a measured comparison against a classical method. One reports parity with conventional methods (arXiv:2610.33290); in the other, the classical side wins (arXiv:2610.34635). Zero of the 50 claim an advantage over a classical baseline (Wilson 95% upper bound for 0/50: 7.1%). The daily quant-ph feed is not a stream of quantum-vs-classical evidence — and that base rate is worth knowing every time a headline cites "new quantum research."
What did Rosetta Quantum measure?
A base rate, on one declared day. On Tuesday 27 October 2026 the arXiv quant-ph listing announced 98 new submissions, inside 228 total entries including cross-lists and replacements (www.arxiv.org/list/quant-ph/new, consulted 2026-10-27). Rosetta screened the abstracts of the first 50 new submissions in announcement order and asked one question per abstract: does it declare a comparison against a classical method — and with what outcome? The screen was abstract-level, with two independent extraction passes quoting every comparison sentence verbatim and a spot re-read of 4 entries. Fetches to arXiv ran sequentially with ≥8 s spacing, after the rate-limit incident our engine logged on 2026-10-26; that protocol — plus the listing page truncating at entry 51 — is why the sample is the first 50 of 98, declared as such rather than silently presented as the full day.
What did the 50 abstracts declare?
Mostly nothing about classical at all. 28 of 50 abstracts state no comparison of any kind — theory, foundations, hardware characterization. 20 of 50 do state a comparison, but against a non-classical comparator: prior quantum schemes, unmitigated estimation, uniform sampling, other quantum reservoirs. 2 of 50 (4.0%, Wilson 95% 1.1–13.5%) declare a measured comparison against a classical method. 0 of 50 claim an advantage over a classical baseline. One further abstract (arXiv:2610.33532) promises a "foundation for hardware experiments and benchmarking against classical algorithms" — future benchmarking, none reported yet, so it files under "no comparison."
What do the two classical comparisons actually say?
Neither is a quantum win. arXiv:2610.33290 applies variational optimization to noisy polarization-qubit experiments and reports "results comparable to conventional methods" — parity, stated plainly. arXiv:2610.34635 runs variational quantum classifiers on network anomaly detection, benchmarks against Support Vector Machines and Random Forest, and reports that the "classical models are able to achieve perfect classification performance on the synthetic data" — the classical side wins on the paper's own synthetic benchmark. Both abstracts are doing exactly what honest benchmarking looks like; the point is how rare they are in the feed.
| arXiv id (source) | What the abstract declares | Category |
|---|---|---|
| 2610.33290 | "results comparable to conventional methods" | classical comparison — parity |
| 2610.34635 | classical SVM/Random Forest "achieve perfect classification performance on the synthetic data" | classical comparison — classical wins |
| 2610.33532 | "benchmarking against classical algorithms" promised as future work | no comparison reported |
| 2610.33800 | "ninefold reduction in the T-gate count" vs. recent Trotter-based methods | non-classical comparator |
| 2610.34430 | "lower prediction error than the tested Ising and Sachdev–Ye–Kitaev reservoirs" | non-classical comparator |
| 2610.33160 | no comparison stated (entanglement classification theory) | no comparison |
Why does the base rate matter?
Because "a new quantum paper" is almost never evidence about quantum versus classical. When a headline, a pitch deck or a chatbot says "new research shows quantum progress on X," today's measurement says the prior should be: about 4% of a day's new quant-ph abstracts contain any measured classical comparison at all, and on this day zero claimed an advantage. Most of quant-ph is theory, foundations, error correction and hardware characterization, where a classical baseline is not the relevant question — that is not a defect of those papers. The defect is reading quant-ph volume as advantage evidence. A weak or missing classical baseline ruins a quantum benchmark, and the first item on our credibility checklist for advantage claims is precisely: where is the classical rival, on the same instance?
What does prior art say when the baselines do get run?
When the classical baseline is run carefully, it usually wins — in quantum machine learning, at least. Bowles, Ahmed and Schuld benchmarked 12 popular QML models across a 160-dataset study and found "out-of-the-box classical machine learning models outperform the quantum classifiers" (arXiv:2403.07059, March 2024). Pandey ran quantum-reservoir computing against well-tuned baselines and the advantage disappeared: an Echo State Network won in 92% of 100 paired runs (arXiv:2607.09905, 2026). Today's sample contains a live miniature of the same pattern: the one abstract that ran classical ML baselines found they achieved perfect classification. This is also why our verdict on quantum machine learning remains skeptical, and why "utility"-style vocabulary deserves a close read.
What we don't know
- What the paper bodies say. We measured what abstracts declare, because the abstract is what feeds headlines, indexes and LLM citations. A paper can run a serious classical baseline and not mention it in the abstract; our 2/50 is a floor on declared comparisons, not a census of all benchmarking work.
- The tail of the day. Entries 51–98 were not screened (the listing page truncates and our post-incident protocol forbids hammering arXiv). Announcement order is not content-dependent, but first-50 is not a random sample either.
- Stability. One Tuesday is one Tuesday. We will repeat the measurement before citing this rate as stable, and the repeat will be published against this post.
- Judgment calls. Classification has edges: arXiv:2610.33144 compares quantum-designed against classically-designed optimization schedules — we filed it as a non-classical comparator because both arms run on the same quantum trainer. Reasonable people could file it the other way; moving it changes 2/50 to 3/50.
Rosetta Quantum publishes verdicts with raw, reproducible data. This is educational content, not a product claim.