Papers · Preprint · v2
Six Birds for Neutrino Mass: Lens-Swap Stability and Packaging Audits
A neutrino mass bound that moves when you swap one dataset.
In plain words
Replace one release of a cosmic microwave background likelihood with another, keep everything else fixed, and an outside analysis reports a tighter upper bound on the summed neutrino mass. Before reading anything into that, the paper asks whether the two releases agree with each other. They do not: the newer release fits far worse at the older one's best parameters than the other way round.
The paper treats this as a packaging stability problem. It runs directional cross lens audits, splits the main part of the mismatch exactly across spectra and multipoles, and checks that it survives the obvious explanations: matching the sky coverage, dropping one prior, raising numerical precision.
The message is about method. When a bound moves under a lens swap, report the bound together with the swap audits. That turns a vague tension into a diagnosable disagreement. The first version also reported how far the bound moved; those figures came from chains that had not converged, and the second version withdraws them and reports no tightening magnitude.
What it shows
- A strongly asymmetric cross fit: testing the newer release at the older fit costs about 488 in χ² with DESI BAO, the reverse about 79; with Planck data added, about 63 and 23.
- An exact split of the main part of the mismatch across spectra and scales, which shifts with baseline and direction.
- An audit first, reproducible diagnosis.
What it does not claim
No root cause for the pipeline difference, no definitive neutrino mass, and no tightening magnitude.
Cite
Tsiokos, I. (2026). Six Birds for Neutrino Mass: Lens-Swap Stability and Packaging Audits. Zenodo. https://doi.org/10.5281/zenodo.23161613