Preprint proposes sign-changing torsional dark energy but yields worse fit than Lambda-CDM despite higher H0

The authors modify the exponential infrared f(T) framework so geometric torsion generates an effective dark-energy density that crosses zero and enters the phantom regime after the transition. MCMC chains combine Planck CMB, DESI BAO, and Type Ia supernovae; the best-fit transition redshift is tightly constrained yet the overall likelihood remains lower than standard Lambda-CDM. No additional fields are introduced, avoiding extra instabilities, but the model still requires six free parameters identical in count to the concordance cosmology.

The reported H0 increase aligns with local measurements yet the chi-squared penalty indicates the data prefer the simpler constant dark-energy term. This pattern echoes earlier modified-gravity attempts that ease the Hubble tension only by degrading fits elsewhere, particularly the CMB-BAO consistency. The preprint explicitly flags the tension between these datasets as an unresolved issue.

Because the work is a preprint without peer review, the parameter constraints and likelihood ratios require independent verification on larger or independent datasets. Future DESI year-three releases or Euclid weak-lensing measurements could test whether the predicted phantom crossing at z less than 1.62 survives or is ruled out at high significance.

The central methodological strength is the use of full MCMC sampling across multiple probes; the decisive weakness is the absence of any Bayesian evidence ratio or information criterion that would quantify whether the extra functional freedom justifies the poorer fit.