Aletheia: Quantifying Cognitive Conviction in Reasoning Models via Regularized Inverse Confusion Matrix

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Fu, Fanzhe
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914233077202944
author Fu, Fanzhe
author_facet Fu, Fanzhe
contents In the progressive journey toward Artificial General Intelligence (AGI), current evaluation paradigms face an epistemological crisis. Static benchmarks measure knowledge breadth but fail to quantify the depth of belief. While Simhi et al. (2025) defined the CHOKE phenomenon in standard QA, we extend this framework to quantify "Cognitive Conviction" in System 2 reasoning models. We propose Project Aletheia, a cognitive physics framework that employs Tikhonov Regularization to invert the judge's confusion matrix. To validate this methodology without relying on opaque private data, we implement a Synthetic Proxy Protocol. Our preliminary pilot study on 2025 baselines (e.g., DeepSeek-R1, OpenAI o1) suggests that while reasoning models act as a "cognitive buffer," they may exhibit "Defensive OverThinking" under adversarial pressure. Furthermore, we introduce the Aligned Conviction Score (S_aligned) to verify that conviction does not compromise safety. This work serves as a blueprint for measuring AI scientific integrity.
format Preprint
id arxiv_https___arxiv_org_abs_2601_01532
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Aletheia: Quantifying Cognitive Conviction in Reasoning Models via Regularized Inverse Confusion Matrix
Fu, Fanzhe
Artificial Intelligence
Computation and Language
Machine Learning
68T50, 65F22
I.2.7; I.2.6; I.2.m
In the progressive journey toward Artificial General Intelligence (AGI), current evaluation paradigms face an epistemological crisis. Static benchmarks measure knowledge breadth but fail to quantify the depth of belief. While Simhi et al. (2025) defined the CHOKE phenomenon in standard QA, we extend this framework to quantify "Cognitive Conviction" in System 2 reasoning models. We propose Project Aletheia, a cognitive physics framework that employs Tikhonov Regularization to invert the judge's confusion matrix. To validate this methodology without relying on opaque private data, we implement a Synthetic Proxy Protocol. Our preliminary pilot study on 2025 baselines (e.g., DeepSeek-R1, OpenAI o1) suggests that while reasoning models act as a "cognitive buffer," they may exhibit "Defensive OverThinking" under adversarial pressure. Furthermore, we introduce the Aligned Conviction Score (S_aligned) to verify that conviction does not compromise safety. This work serves as a blueprint for measuring AI scientific integrity.
title Aletheia: Quantifying Cognitive Conviction in Reasoning Models via Regularized Inverse Confusion Matrix
topic Artificial Intelligence
Computation and Language
Machine Learning
68T50, 65F22
I.2.7; I.2.6; I.2.m
url https://arxiv.org/abs/2601.01532