There is a newer version of the record available.

Published February 10, 2026 | Version v2
Preprint Open

Fisher Information Geometry and the Thermodynamic Cost of Self-Referential Calibration

Authors/Creators

Description

Version 2 adds a new Section 8: Numerical Demonstration, a self-contained 20-dimensional linear-prediction simulation confirming the three core predictions (delusion-correction separation, detectable staleness, optimal calibration budget) under continuous drift.                                                                                                                                                                         

Changes from v1:

  - New Section 8 with three figures and reproducible Python code                  

  - Abstract: corrected ln 2 factoring to match Theorem 5

  - Detectability proof: fixed θ₀ linearisation

  - Additivity proof: added separable-DOF qualifier

  - Discussion: updated summary table to reference numerical results

  - Editorial: minor wording improvements throughout

Supplementary material: tdome_demo.py (self-contained, requires only NumPy + Matplotlib), three figure PDFs, and alpha_opt_curve.csv.

Files

T_DOME_paper_III_v2.pdf

Files (918.2 kB)

Name Size Download all
md5:564f70deed588da7e99dabd6164a3acc
425 Bytes Preview Download
md5:bf1722c0cc3aefc8e9bd4ac54f168507
22.0 kB Preview Download
md5:73aedeeeeeb2915cbb6f136da501cc95
146.1 kB Preview Download
md5:2e876ce4335b5b80887a29ebb715ddfc
36.2 kB Preview Download
md5:5ab9aa1f72b9feb0437d624b9b10a9e1
602.5 kB Preview Download
md5:fa7366c1b60158f15119ce3732a2123c
86.9 kB Download
md5:547efdedaf1a78418cdb9aa96aea915e
24.1 kB Download

Additional details

Dates

Updated
2026-02-10
v2: Added Numerical Demonstration section (Section 8) with reproducible simulation and three new figures
Created
2026-02-10
v1: Original publication