New neuro-symbolic system eliminates LLM requirement errors with 94.6% accuracy

Large Language Models (LLMs) offer natural-language flexibility for automated requirements elicitation but frequently produce structurally invalid or logically inconsistent outputs. Ahmed Ibrahim's latest paper presents a novel neuro-symbolic multi-agent architecture that addresses this by operationalizing the Object-Oriented Method for Requirements Authoring and Management (OOMRAM) lattice. In this system, the LLM acts as a non-deterministic heuristic for lattice traversal, while a deterministic symbolic validator enforces all structural constraints. Introduced is a three-valued uncertainty framework—Truth (T), Indeterminacy (I), Falsity (F)—to classify and score the LLM's requirement decisions both before and after validation, providing a formal measure of neural uncertainty.

Evaluated across 37 natural-language project visions in eleven application families, the system completely eliminated structural inconsistencies in 35 out of 37 cases (94.6%). The remaining two cases contained only 6 unresolved structural errors, representing 0.39% of all decisions, attributed solely to iteration limits. Three-valued analysis revealed that 24.7% of all decisions are indeterminate—structurally valid but discretionary choices not explicitly mandated by stakeholders. The study concludes that offloading structural integrity to a deterministic symbolic layer successfully guarantees conformance, while the uncertainty scoring facilitates safe LLM deployment in formal requirements engineering, a critical step for AI-assisted software development.