The question from the 2026-06-23 tomography call: including blended (FLAGS=2) objects produces B-modes under the current noise-fill, while discarding them is a non-metacalibratable selection bias. Does BLEND_HANDLING = uberseg (#770) remove those B-modes?
Setup: two P3 runs differing only in BLEND_HANDLING; FLAGS=2-inclusive sample; B-modes via the Paper II / sp_validation machinery. The new NGMIX_NEIGHBOUR_FLAG column enables a neighbour/no-neighbour split as a cross-check.
Metric: B-mode PTEs on the FLAGS=2-inclusive catalogue, noisefill vs uberseg.
Decider: Fabian signs off on the science; only then is a default flip discussed.
Open design: area required for a decisive PTE, patch choice, matched-object vs matched-area comparison.
— Claude, on behalf of Cail
The question from the 2026-06-23 tomography call: including blended (
FLAGS=2) objects produces B-modes under the current noise-fill, while discarding them is a non-metacalibratable selection bias. DoesBLEND_HANDLING = uberseg(#770) remove those B-modes?Setup: two P3 runs differing only in
BLEND_HANDLING;FLAGS=2-inclusive sample; B-modes via the Paper II / sp_validation machinery. The newNGMIX_NEIGHBOUR_FLAGcolumn enables a neighbour/no-neighbour split as a cross-check.Metric: B-mode PTEs on the FLAGS=2-inclusive catalogue, noisefill vs uberseg.
Decider: Fabian signs off on the science; only then is a default flip discussed.
Open design: area required for a decisive PTE, patch choice, matched-object vs matched-area comparison.
— Claude, on behalf of Cail