Vyom
pipeline: operational·archive: JWST public·38 spectra / 15 worlds·pre-registered: OSF
standard of evidence

We didn't invent the bar. We hold to it.

A verdict is only as good as the standard behind it. Vyom adopts the community's own consensus — the Confidence-of-Life ladder and the Standards-of-Evidence checks — and implements them as a reproducible, audited pipeline. Same bar for every world.

The Confidence-of-Life ladder

Green et al. 2021 · Nature 598, 575
1
Signal detected

A candidate spectral feature is present in the data above instrument noise.

2
Artifacts ruled out

The feature survives data-reduction, instrument-systematic, and stellar-contamination checks — it is real, not an artifact.

3
Biological attribution

The species is plausibly produced by life, in a habitable context for this planet.

4
Abiotic sources rejected

Every known non-biological pathway for the species, on this planet class at this temperature, is ruled out.

5
Independent detection

The signal is confirmed in a second instrument, wavelength, or epoch — not one visit's quirk.

6
Corroborating evidence

Independent lines of evidence (companion molecules, disequilibrium pairs) point the same way.

7
No plausible alternative

Follow-up leaves no credible non-biological explanation. A claim the community can keep.

the seven checks

What a claim must pass.

Distilled from Green et al. 2021, the NASA Ladder of Life Detection, and the Meadows et al. 2022 standards workshop. Vyom scores every verdict against all seven — and shows you which ones fail.

S1

Signal reality

The feature is in the data — not a reduction artifact, not stellar contamination.

S2

Retrieval calibration

The posterior is calibrated (coverage ≈ nominal), not a prior read back to you.

S3

Statistical significance

Multiple-testing-corrected significance, not a raw n-sigma headline.

S4

Model adequacy

A posterior draw actually fits the data, and no better model is being ignored.

S5

Abiotic null ruled out

Every published abiotic route to the species, at this temperature, cannot reach the retrieved abundance.

S6

Biotic alternative preferred

Bayesian model comparison favours a biotic-plus-abiotic model over abiotic-only.

S7

Reproducibility

The verdict reproduces from released code, data, and the pre-registration.

See all seven scored live, and watch them change as you toggle the corrections, in the verdict console →

the claims ledger

What we claim. What we refuse to.

Claimed — supported today
  • A uniform, pre-registered, ML-accelerated Bayesian retrieval applied identically across the public JWST archive.
  • Calibrated posteriors in seconds per target, with simulation-based calibration and a nested-sampling cross-check on every critical claim.
  • A verdict placed on a published standard of evidence — the Confidence-of-Life ladder and the seven Standards-of-Evidence checks.
  • Reproducibility by construction: released code, released data, and an OSF pre-registration that fixes the analysis before the verdict.
Not claimed — and we'll say it first
  • A life-detection claim — the pipeline does not yet clear the community bar (the K2-18 b-style stellar-contamination model, S1, is still being completed).
  • Re-reduction of raw JWST data — Vyom consumes published, provenance-tracked stage-3 spectra; it does not reduce pixels.
  • Discovery of new planets, or proprietary-data analysis — public archive only, proprietary periods respected.
  • Beating large institutional teams on single-target depth — Vyom's edge is breadth, uniformity, and neutrality, not exotic per-target chemistry.