First-use RMS references support young-star retrieval and expose ownership limits
6 Oct 2026 · embeddings
Do unchanged unsupervised cloud rankings retrieve follow-up-informed young-star references, and do source/gas velocities support their positional associations?
Freeze E135 models, seven orders and operating fractions before downloading the complete public RMS snapshot once. Match all 94,604 clouds to all 2,798 sources at fixed 3/15/30 arcsec with a +60 arcsec Galactic-latitude control; primary 15 arcsec links must be unique both ways. Preserve unknown/evolved classes, all alternatives and full requested denominators. No label fitting or quota tuning.
One 306,888-byte public GET with no target upload. Independent fixed-byte Decimal reader checks 47,566 native cells and coordinates (maximum 5.69e-14 degree difference). Separate spherical matching checks 111,038 pair distances, all 1,816 alternatives and every radius decision; 46,230 review cells, 350 budget rows and 1,138,970 selected memberships agree. Accepted workers terminal exit0; two original checker failures are preserved with registered corrections; producer science unchanged. Science seal has 40 bindings.
There are 1,005 unique RMS/cloud positional links: 368 YSO references, 467 H II/mixed proxies, 116 evolved contexts and 54 unknowns. At 4,730 reviews, unchanged learned warmth retrieves 312 YSO references versus brightness 284, gaining 51 identities and losing 23; exact common support retains 312 versus 296. At 945 reviews warmth retrieves 130 YSOs versus brightness 97. Primary bright/warm dual retrieves 295 YSOs and 425 region proxies at 5%, with a native source-specific 1,005-row review export. Of 206 fit-qualified RMS/MALT90 velocity pairs, 195/198/203 agree within 3/5/10 km/s. Nine original finite-field comparisons abstain: eight have no fitted lines and one lacks a reliable consensus feeder. All prior stellar/dust/molecular tiers remain.
| Metric | Value | 95% CI | n |
|---|---|---|---|
| full requested cloud entries | 94604 | — | — |
| complete RMS reference rows | 2798 | — | — |
| primary unique RMS cloud links | 1005 | — | — |
| primary five percent YSO references | 295 | — | 805 |
| primary five percent region proxies | 425 | — | 960 |
| fit-qualified velocity pairs within three km per second | 195 | — | 206 |
Next: Audit source-specific crowding, ownership and the three fit-qualified velocity conflicts; obtain source-selected candidate truth and fresh faint/crowded transfer before changing models. RMS is now consumed development for any future choice. Keep the useful unsupervised warmth embedding, brightness comparison, ordinary-star/missing/low-motion tiers and compatible dust/embedded/birth-region scope. Full goal active; E121 pending with zero requests.
Warmth retrieves fewer H II proxies (387 versus brightness 419 at 5%) and more evolved contexts (29 versus nine), so no uniform winner or population-purity gain is established. The full RMS catalogue contains 805 YSOs and 960 region proxies: warmth 312/805=38.76% requested YSO recovery is narrower than any all-sky claim, while linked recovery is 312/368=84.78%. 3,966 primary 5% reviews lack a unique RMS reference; 21 linked types are unknown. RMS is a selected 2013 follow-up-informed snapshot with some shared infrared/kinematic context, not spectra of every source. Offset has 70 unique 15 arcsec links. Three fit-qualified H II-region velocity differences exceed 10 km/s; the original ten raw differences included unsupported catalogue placeholders. Missing velocity errors and beam confusion prevent formal membership. Unexpected native classes abstain. No new star, physical cloud union, birth or robust survey purity is established.