We test a model-independent dynamical constraint on late-time cosmic expansion formulated as an averaged relaxation bound on the deceleration parameter. The hypothesis states that once the Universe enters the accelerated regime, the coarse-grained evolution of the expansion history does not revert toward net deceleration. This condition is expressed as an inequality on a global relaxation functional constructed from the reconstructed Hubble rate H(z). We combine cosmic chronometers, Pantheon+ supernovae, and the most recent baryon acoustic oscillation measurements from DESI DR1 and DR2. Using spline reconstruction and bootstrapweighted likelihood sampling, we evaluate the posterior probability of the relaxation condition without assuming any specific dark energy model. Both DESI generations yield saturated support for the relaxation bound, with posterior probability P(I ≤ 0) ≈ 1. The result is stable under independent BAO updates and resolves ambiguities present in legacy datasets. We conclude that the accelerated expansion of the Universe is observationally consistent with a relaxation process at the background level. This provides a new model-independent characterization of cosmic acceleration and motivates further tests in growth and lensing sectors to determine whether the relaxation principle extends beyond background dynamics. This manuscript will first be released as a Zenodo preprint to solicit community feedback prior to journal submission.

This record corresponds to the first public Zenodo release of the manuscript “Is Cosmic Acceleration a Relaxation Phenomenon? A Model-Independent Bound on Late-Time Cosmic Dynamics”. This Zenodo preprint is released to solicit community feedback prior to journal submission. This preprint belongs to the LERAC/HERACLES program: a suite of model-independent consistency tests intended to assess cosmic acceleration through independent closure conditions across background, gravity (matter–light closure), and growth sectors. The present paper focuses exclusively on the background relaxation bound and does not assume any growth model or modified gravity parameterization. This manuscript was prepared with limited assistance from a generative AI tool (LLM) for drafting support, structural editing, and internal consistency checks. The author independently designed the methodology, curated the datasets, performed the analysis, verified the results, and takes full responsibility for all scientific claims.