Complete molecular context refines missed clouds; ranking gain depends on review budget
5 Oct 2026 · embeddings
Does cached independent-modality line activity refine the real cloud collection and support the frozen label-free rank beyond temperature?
Preserve all3556 native records/3246 pointings, all94604 clouds, every fixed3/15/30arcsec alternative and60arcsec latitude control. Require both-direction unique15arcsec links and one catalogue record for primary refinement; no refit or threshold tuning.
Independent14224 line flags,94604 cloud states,614169168 latitude-strip decisions,4797483 haversine decisions,3394 pairs; maximum angular difference1.70e-10arcsec. Separate Decimal/rational review of10 fixed parent budgets and162710 memberships.103-entry native view checks1854 values;303 legacy entries reviewed.
2331 unique links,2156 one-record links,433 reliably line-active; family selects324 versus temperature299 at different burdens. At equal1%/5% review family retrieves92/171 versus temperature55/189; no universal gain.103 missed/no-cached-W2 refinement candidates comprise58 positive-stage and45 negative-stage references. Legacy11 multirecord entries change no old activity labels. No downloads.
Next: Improve broad native star-family learning without class-balanced sampling and retain exact native errors/IDs, raw baselines and whole requested denominators. Obtain pending upload approval for cloud source recovery; broaden cold views and resolve native velocity/spatial ownership, independent stellar/spectral/image/temporal evidence. Full robust goal stays active.
Targeted massive-clump survey, consumed development observations and sensitivity to mass/distance. Line activity is not stellar truth; nondetection is not starless. One catalogue record may merge profiles; beam/velocity ownership remains unverified. Fixed-budget amendment registered after initial operating counts; no robust survey gate or physical birth certified. IRSA upload remains pending explicit approval; no request ran.