Fresh frozen screening finds candidates and exposes single-band loss

How does the frozen mixed-family detector transfer to unfiltered fresh sky patches?

Four fixed new WISE rectangles and separate unfiltered primary spectra; unchanged model, mapping and threshold, class-free scoring freeze, all-source burden and explicit reference coverage.

Eight bounded requests; independent DOM/Decimal checks 318,917 primitives and 62,012 numeric nulls. Scalar Gaussian replay computes 63,369 family quantities, compares 37,829 supported exported scalars and checks both frozen methods and all 7,041 status/flag records; Cartesian review reconstructs 896,236 geometry decisions, all references, candidates and counts.

7,041 new identities, zero development overlap; 2,059 candidates, 3,439 supported and 3,602 unknown. Qualified spectral proxies: 43/55 stars selected and 0/37 others, versus E98 37/55 and 4/37. Candidate density 1,894–4,263 per square degree. 259 spectra, 92 qualified associations (1.31% source coverage). Downloads 10,428,169 bytes; stage about 17.4 MiB.

Next: Develop calibrated sparse-view/one-band inference with uncertainty and independent validation; assess starless/protostellar cloud context compatibly with stellar candidates, then native astrometric/temporal refinement. Preserve the original frozen fresh result and every missing target.

Finite deliberately selected regions and target-selected spectra; 100% conditional precision proxy does not label the unobserved majority. Ten qualified stars have only usable W1 profile evidence and remain unknown. All raw STAR labels: 58/100 have flagged alternatives, including low-quality labels. Static proximity does not certify physical ownership, motion or class. Cloud and temporal birth evidence remain incomplete.