arXiv:2604.11284v5 Announce Type: replace-cross Abstract: We present THEIA, a 2.75M-parameter modular neural architecture that learns the complete Kleene three-valued logic (K3) truth table from task data without external symbolic inference or hand-encoded K3 gate primitives. Across 5 seeds it passes all 39 K3 rules at >99% per-rule accuracy. K3 learnability is not the central finding: Transformer baselines also pass all 39 rules, and flat MLPs match THEIA on Phase-1 accuracy within 0.04pp. The contributions are two properties of the learned system. (1) Uncertainty-verdict asymmetric propagati
This research addresses fundamental questions about how neural networks can learn foundational logical structures, which is critical for advancing AI capabilities.
Understanding how AI can inherently learn and represent complex logical systems like Kleene three-valued logic without explicit programming offers pathways to more robust and explainable AI.
The ability of neural architectures to learn K3 logic without explicit symbolic inference suggests potential advancements in AI's reasoning capabilities beyond mere pattern recognition.
- · AI researchers
- · Machine learning developers
- · Logic-based AI systems
- · Traditional symbolic AI approaches
- · AI systems lacking robust logical foundations
Further research will explore the generalization of such modular neural architectures to other complex logical systems.
This foundational work could contribute to more explainable, auditable, and reliable AI systems in critical applications.
The inherent learning of logic might eventually lead to AI capable of more sophisticated reasoning and problem-solving, impacting areas like scientific discovery or complex systems design.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI