aprimora-ai/hugo-framework: HUGO Series Paper I — v1.0.0

HUGO Paper I v2 — Homeostatic Temperature and Primary Emotions as Topological Signatures in Neural Information Flow. Version 2 introduces HomeostaticField v3.0 with three architectural corrections: (1) set-point model where each vector oscillates around its initial condition, encoding agent identity in set points and experience in oscillations; (2) orthogonal projections via Gram-Schmidt creating 5 independent feedback pathways; (3) ensemble-calibrated decay rates derived from perturbation statistics during a 30-step burn-in period (CALM analog), eliminating decay_rate as a hyperparameter. v1 suffered from saturation instability — all H vectors drifted to [1,1,1,1,1] within 40 steps regardless of initial conditions. v2 maintains 4 distinct emotional labels and tau spread > 0.23 at 40, 200, and 500 steps. All three experiments re-executed: Exp 1.5 (5 configs, KW p=0, H=107–239), Expansion 2 (8 configs, FEAR gradient Spearman r=±1.0), Expansion 3 (6 configs, 4/6 trajectory metrics significant — convergence NOT significant, a genuine negative correcting a spurious v1 result). 33 unit tests passing. Part of the HUGO AGI Framework. Extended by ECHO (Paper II, DOI: 10.5281/zenodo.19043115). Repository: https://github.com/aprimora-ai/hugo-framework