Neuro-Symbolic Architecture Validates LLM Requirements with Three-Valued Scoring
A recent publication on arXiv presents a neuro-symbolic multi-agent framework that employs a three-valued system (Truth, Indeterminacy, Falsity) to assess and verify requirements produced by Large Language Models. This method implements the Object-Oriented Method for Requirements Authoring and Management (OOMRAM) lattice, with the LLM serving as a non-deterministic heuristic for navigating the lattice and a deterministic symbolic validator that upholds structural constraints. The primary aim is to eradicate logical inconsistencies and ensure structural adherence in requirements generated by LLMs, while also measuring uncertainty in pre-validation decisions. The findings are detailed in arXiv:2607.26220.
Key facts
- arXiv:2607.26220
- Announce Type: cross
- Context: LLMs offer natural-language flexibility but generate structurally invalid requirements and logical inconsistencies
- Objectives: eliminate logical inconsistencies, enforce structural conformance, quantify pre-validation decision uncertainty
- Methods: neuro-symbolic multi-agent architecture operationalizing OOMRAM lattice
- LLM acts as non-deterministic heuristic for lattice traversal
- Deterministic symbolic validator enforces all structural constraints
- Three-valued (T, I, F) framework: Truth, Indeterminacy, Falsity
Entities
Institutions
- arXiv