PHILIA Crucible: A Human-Centered AI Output Verification System

Read full text Summarise with VIRAT
Authors: GritMan_ Shin
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

PHILIA Crucible: A Human-Centered AI Output Verification SystemGritMan_D.S | Trinity AI | 2026.030∞1∞0.5∞ — Description, not Proof AbstractThis paper presents PHILIA Crucible, a post-generation AI output verification system developed under the Trinity AI framework. During PHILIA Engine validation (v8–v13), AI teammates (Grok, Gemini, ChatGPT, Claude) produced outputs containing fabricated or unsupported numerical values. This led to a core verification principle: only values directly executed by GritMan_D.S on a local machine count as ground truth. Crucible operationalizes this principle as a keyword-based purity scoring engine. Three hallucination pattern types were identified and classified through empirical testing: (1) empirical claim fabrication, (2) estimation presented as measurement, and (3) simulation-to-application overreach. Crucible v2 Real correctly assigned F (0.0000) to two fabricated papers (DOI: 10.5281/zenodo.19084834, DOI: 10.5281/zenodo.19098650) and A (0.75) to three verified papers (v8, v9 Real, v10).

Fields of Study & Subjects

Ethics and Social Impacts of AI Misinformation and Its Impacts Academic integrity and plagiarism

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
21 March 2026
Language
EN