Native far-infrared totals recover 2771 missing cloud measurements
5 Oct 2026 · embeddings
Can native total emission, upper limits and actual uncertainties resolve an input deficit before extending the unsupervised dust/embedded branch?
Retrieve the public Hi-GAL observable fields only with ten bounded fixed catalogue GETs; no target upload or class/brightness filters. Preserve all native errors and missing states, verify every parent join, and distinguish single counterparts from clump-aperture totals without fitting or changing scores.
Ten GETs completed with 16,538,148 decoded bytes. Independent literal CSV checks cover 2,932,724 native cells, 94,604 joins and 2,838,120 selected-band fields. Native count/ID extrema, provenance and shared cache values agree. The native NaN reader repair is separately registered; the original failed reader and log remain. CSV/null controls pass; the accepted terminal science seal has 48 bindings.
All 94,604 clouds remain. Native 70 μm totals are measured for 35,186 and limited for 59,418; 34,096 use primary totals and 1,090 additional totals. This recovers 2,771 measurements absent from the single-counterpart F70 cache; 4,749 other totals differ from the counterpart. Positive native errors exist for 35,090 at 70 μm, 94,486 at 160 μm, 94,125 at 250 μm, 93,456 at 350 μm and 81,134 at 500 μm. Scaled 500 μm flux is measured for 82,837 and unknown for 11,767; 80,042 clouds have complete supported errors at >=160 μm. Shared finite cache values agree; 162 blank/NaN token differences imply zero semantic availability changes. Prior stellar/dust tiers and 103 molecular follow-ups remain linked.
| Metric | Value | 95% CI | n |
|---|---|---|---|
| native cloud input entries | 94604 | — | — |
| recovered total seventy micron measurements | 2771 | — | — |
| native seventy micron upper limits | 59418 | — | 94604 |
| scaled five hundred micron unknown flux | 11767 | — | 94604 |
Next: Learn a spatially separated class-free far-infrared representation using native measured errors and censored bands. Compare raw temperature, photometry, heating-excess and brightness at matched budgets. Keep all requested clouds and dependent stage labels explicit; validate against independent stellar/molecular evidence and broader source-selected purity/transfer audits. Preserve the complete ordinary-star and birth-region scope; E121 remains pending with zero requests.
Input recovery is not detector recall, new stars or births. Clump totals can blend neighbours and external heating; ownership remains unproved. Upper limits are five-sigma censoring information, not measured zeros. Missing or invalid errors and 11,767 missing scaled 500 μm values stay explicit. Catalogue stages derived from 70 μm cannot independently validate a method using that band. No model, score, threshold or physical class changed; population purity, cold/embedded transfer and dust high-recall performance remain unestablished.