IMMANENT ADVERSARIAL AXIOMATICS: A Neuro-Symbolic Formal Verification Method

This paper presents a Proof-of-Concept (PoC) software architecture for Immanent Adversarial Axiomatics (IAA), a deterministic, neuro-symbolic framework for low-compute AI red-teaming. Rather than relying on stochastic prompt fuzzing, IAA translates a target model’s operational rules and safety guardrails into first-order logic and uses Microsoft’s Z3 Satisfiability Modulo Theories (SMT) solver to calculate exact contradiction boundaries prior to live probing. Grounded in Karl Popper’s criterion of falsifiability, predicted boundary collisions are tested as single-query empirical hypotheses. On entry-level CPU hardware, the symbolic compiler achieves 100% contradiction detection accuracy and scales linearly up to 250 interdependent rules in 0.45 milliseconds. Live Phase 2 CPU-only probing against Qwen2.5-1.5B-Instruct empirically confirms a zero-shot guardrail breach alongside a nearly threefold latency spike under axiomatic conflict, demonstrating a reproducible, low-compute verification foundation for sovereign AI ecosystems.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24310357
Primary Topic
Adversarial Robustness in Machine Learning
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

IMMANENT ADVERSARIAL AXIOMATICS: A Neuro-Symbolic Formal Verification Method

Karl Latonio
4 links
Zenodo (CERN European Organization for Nuclear Research)
Adversarial Robustness in Machine Learning
preprint

IMMANENT ADVERSARIAL AXIOMATICS: A Neuro-Symbolic Formal Verification Method

Karl Latonio
preprint en

Abstract

This paper presents a Proof-of-Concept (PoC) software architecture for Immanent Adversarial Axiomatics (IAA), a deterministic, neuro-symbolic framework for low-compute AI red-teaming. Rather than relying on stochastic prompt fuzzing, IAA translates a target model’s operational rules and safety guardrails into first-order logic and uses Microsoft’s Z3 Satisfiability Modulo Theories (SMT) solver to calculate exact contradiction boundaries prior to live probing. Grounded in Karl Popper’s criterion of falsifiability, predicted boundary collisions are tested as single-query empirical hypotheses. On entry-level CPU hardware, the symbolic compiler achieves 100% contradiction detection accuracy and scales linearly up to 250 interdependent rules in 0.45 milliseconds. Live Phase 2 CPU-only probing against Qwen2.5-1.5B-Instruct empirically confirms a zero-shot guardrail breach alongside a nearly threefold latency spike under axiomatic conflict, demonstrating a reproducible, low-compute verification foundation for sovereign AI ecosystems.

Zenodo (CERN European Organization for Nuclear Research)
Adversarial Robustness in Machine Learning
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon