Bistro Research
astrophysics
archive · read-only

Hidden protostars

Can archival infrared surveys catch a star being born?

Maintained by Igor Moiseev · source repository

Imported archive: 163 experiments run in the source repository between September and October 2026. Read-only here; its claims re-execute in the source repo.

Articles

Research graph

163 experiments · 5 tracks
literature & feasibil…turn-on searchburst & YSO calibrati…contaminationembeddingsexperiment order, earliest on the left →
confirmedpartialrefutedpendingfinding motivated a later hypothesislater experiment controlled an earlier caveatlater experiment superseded an earlier oneEach dot is one experiment, coloured by its finding's verdict; click one to open it.

What runs next

Ranked by value ÷ cost; donations pay for the top one first
  1. 01

    Complete E12 on a subsample (clusters, stability, pseudotime)

    The full-catalogue E12 run failed after 15.5 h (ARPACK no convergence). On a stratified 20k subsample, are the unsupervised clusters stable, not a distance map, and does pseudotime order stages?

    Pre-registered: As pre-registered for E12: E12a distance AUC < 0.6 per cluster; E12c median ARI > 0.7; pseudotime stage AUC reported; E12d expected null (E12b already showed no far-IR signature).

    score 2.0value 2 · cost 1
  2. 02

    Far-IR / sub-mm two-epoch test of prestellar clumps

    The literature puts the first observable sign of a birth in the far-IR/sub-mm (internal luminosity rising at 70-850 um), not at 3-5 um. Do any Hi-GAL prestellar clumps brighten between Herschel (2010-12) and a later far-IR/sub-mm epoch?

    Pre-registered: A matched two-epoch comparison (Hi-GAL vs a later survey at overlapping wavelength and resolution) finds a brightening rate above the calibration-noise tail estimated from protostellar and field sources; the data-availability check comes first (which later surveys overlap, at what depth).

    score 1.3value 5 · cost 4
  3. 03

    Absolute lifetimes from corrected shares

    What absolute prestellar and protostellar clump lifetimes follow from the E17/E19 corrected shares with the Elia+2021 clock, and how do they compare with Battersby+2017?

    Pre-registered: The corrected prestellar lifetime is at least 15% shorter than the uncorrected one in every bin within 8 kpc.

    score 1.0value 1 · cost 1
  4. 04

    Physics VAE v3 (warm slope, censored likelihood, extinction)

    With 8/12 um warm-slope bands, a censored 22/24 um likelihood and an extinction term, does the physical latent beat the nine-feature baseline on held-out bands?

    Pre-registered: Held-out 70 um calibration >= 90% under the limit for dark clumps and probe AUC >= 0.95.

    score 1.0value 3 · cost 3
  5. 05

    Trimodal embedding: far-IR images + mid-IR images + 70um-free SED

    Does adding a 70um-free SED encoder as a third aligned view (AstroM3-style) raise the hidden-protostar signal?

    Pre-registered: 24um-bright AUC gain over two-view CLIP > 0.02 with paired CI above 0.

    score 0.3value 1 · cost 3

Findings

Latest 60