A Multi-Survey Anomaly Search for Main-Belt and Inner Solar System Objects
An exploratory pipeline scored 4,266 objects using 940,861 accepted observations. It retained 300 ranked entries, 19 after hard filters, and 17 under its “strict cross-survey residual” label. Most retained objects still carry a single-survey caution. These are provisional screening results, with no established non-natural detection.
Method and data coverage
The pipeline fits photometric models using observation history, geometry, physical-property proxies, and survey information from Gaia, NEOWISE, ZTF, and associated catalogues. The reported stage-1 RMSE is 0.108 mag. That value alone does not establish generalization, absence of leakage, or calibration of outlier significance.
Accepted corpus: Gaia 940,283 rows (99.94%) · NEOWISE 470 (0.05%) · ZTF 108 (0.01%).
Run ID: 20260410T021500Z_noztf_neowise200clean_tranche10.
The corpus is overwhelmingly Gaia-based. “Strict cross-survey residual” is an internal filter label, not a guarantee of independent detections of the same physical discrepancy. Coverage and quality must be assessed for each object.
Selected follow-up targets
Rankings prioritize inspection within this pipeline. The legacy “artifact score” is a composite anomaly score, not a probability that an object is artificial. Stability across reruns of similar data does not establish physical reality.
005232: cross-rank 1, score 29.28, four flagged apparitions and 116 MOST matches. Its survey_only and live_query_no_rows flags leave independent optical and thermal confirmation incomplete.
016467: cross-rank 3, score 25.42 and gaia_slope_z = 2.11. Agreement between diagnostic scores motivates review, but those scores may share inputs.
000466: cross-rank 4, final rank 16, five flagged apparitions and 287 MOST matches after additional ZTF extraction. The reported C-type object is approximately 95.5 km across; added matches are not themselves independent confirmations.
026895: cross-rank 9, final rank 100, spectral score 1.96 and gaia_slope_z = 2.99. It leads this subset's spectral ranking.
000200 — natural explanation favoured
The report records 462 NEOWISE observations for 000200 and a subsequent move toward natural_likely after thermal review. This is an example of a candidate being downgraded. It does not prove the validity of the other rankings or completely resolve the object's physical properties.
Structural cautions
survey_only (11/17) · spike_glint_like (7/17) · absolute_offset_pattern (5/17) · band_inconsistency (5/17) · phase_resid_pattern (1/17)
These flags overlap and describe screening patterns, not diagnosed causes. Only two of the seventeen residuals have NEOWISE time-series data in the active run. Missing thermal rows provide neither confirmation nor a strong exclusion of a proposed signal.
Sensitivity and false positives
The reported injection tests recovered 100% of the tested thermal-excess controls and 37.5% of glint-like injections at level 3.000. They recovered none of the weak additive optical injections at level 0.100, or intermittent injections at 0.300 with duty fraction 0.100. These results apply to the tested controls and injection settings; they do not establish sensitivity to all signals in those families.
The report estimates stage-1 and additive-track false-positive rates of 0.158 and 0.000 from nineteen control objects. This small control set leaves substantial uncertainty; observing zero false positives does not establish a zero underlying rate. Likewise, a null search cannot constrain a signature the pipeline has not demonstrated it can recover.
Follow-up criteria
Thermal-optical comparisons, calibrated spectra, and independently measured optical residuals could clarify individual cases. Each requires natural-model and instrument checks. No single unusual feature automatically promotes an object to a technology claim.
The reported three-dimensional motif test gives q = 0.597 for stable_triangle_shape, which does not support significance under that test. Improving weak-signal recovery would justify rerunning the screen; it would not itself strengthen the evidence for any particular object.
The working console
The interactive console presents the reported shortlist, reason codes, and injection results. Its orbital view uses illustrative positions. The downloadable PDF preserves the earlier report; the qualifications on this page apply to its stronger conclusions.