An optical spectrum would reveal redshift directly (if extragalactic) or stellar spectral type (if galactic). The current driving feature Γ = +1.9 ± 1.0 (53% relative error) is a photometric proxy — a spectrum is the direct measurement.
Currently 60 detections in 4 bands over 113.8 days. Color index z−y is unmeasured — insufficient epochs in one or both component bands. Additional detections spanning that gap would constrain spectral shape and reduce reliance on the single measured index.
A deep image in the primary band (i) at sub-arcsecond resolution would determine whether a faint host galaxy underlies the transient position — a distinction Rubin difference frames at their current depth cannot make.
Negative difference flux implies the source was brighter when the survey template co-add was constructed. Inspecting the co-add at this position would confirm whether the source pre-dates LSST coverage, which would strongly favour a persistent extragalactic nucleus over a new transient.
Before trusting any "no match" result, the crossmatch infrastructure was validated against three controls: (1) a bright Gaia star — returned a clean Gaia DR3 match with correct separation; (2) a known AGN from SIMBAD — matched with the correct object type; (3) a blank sky patch — all four catalogs returned empty cones, confirming the code does not fabricate matches on empty fields. Only after all three controls passed were any "no match" verdicts trusted.
Scores use a percentile_composite scorer (v1.0.0). Each feature column is ranked within its tier — sparse objects are compared only to other sparse-tier objects, so a 4-epoch candidate is not penalised against a 200-epoch one. The IsolationForest threshold is calibrated to the 99th percentile of the tier distribution.
This object sits in the rich tier with 15 scoring columns. Its anomaly score of 0.8300 means it landed in the bottom 99.4% of its tier on the driving feature.
Fink real/bogus classifier: mean 0.201, min 0.111. Minimum reliability is close to the 0.5 decision boundary — at least one detection may be borderline. Inspect image stamps before reporting.
Feature set v1 · gate v1 · model percentile_composite 1.0.0 · classified 2026-06-19 02:37 UTC.
Run ID: 0446da9b-077c-4b88-a253-7acc335d4e0d