What question did this study set out to answer?

The aim is to develop a bidirectional neural-symbolic architecture for generating and validating integer fraction derivations across physics contexts.

April 14, 2026Open Access

Integer LLM with Prolog Knowledge Base: Bidirectional Neural-Symbolic Architecture for Provenanced Generation

Key Points

The aim is to develop a bidirectional neural-symbolic architecture for generating and validating integer fraction derivations across physics contexts.
Leverage Prolog knowledge base for structured reasoning
Utilize exact fraction arithmetic for deriving values
Conduct comparisons against published experimental measurements
Document both successful and failed comparisons in the research archive
Derived values are rigorously tested against independent published data
Failure rates of derived values are tracked and published alongside successes
The archive serves as a comprehensive resource for integer fraction physics across various domains

Abstract

Integer LLM with Prolog Knowledge Base: Bidirectional Neural-Symbolic Architecture for Provenanced Generation This paper is part of the HOWL research archive—a collection of physics papers exploring integer fraction derivations across multiple domains using exact arithmetic and automated comparison. Falsification Criteria All papers in this archive are subject to falsification through direct comparison to published experimental measurements. Each derived value is tested against independent data with explicit PASS/FAIL criteria. Any derived value that fails its comparison is documented and published alongside the successes. Research Context This archive documents an ongoing research program in integer fraction physics. The methodology is: derive values from gauge group integers using exact fraction arithmetic, compare to published measurements, and document all results including failures. The archive spans multiple physics domains connected through the soliton boundary framework described in the constituent papers. Package Contents manuscript.md: The complete derivation and supporting analysis. README.md: Navigation, dependencies, and citation (Registry: HOWL-LLM-1-2026). Dependencies: HOWL-INFO-1-2026, HOWL-INFO-2-2026, HOWL-INFO-3-2026, HOWL-INFO-4-2026, HOWL-MATH-2-2026, HOWL-SOPH-2-2026 Motto: Derive. Compare. Publish.Status: Architecture Specification

Integer LLM with Prolog Knowledge Base: Bidirectional Neural-Symbolic Architecture for Provenanced Generation

Key Points

Abstract

Cite This Study