Artificial intelligence and policy-governed systemsPublic institute preprint; arXiv:2608.30757 in cs.AI, cross-listed cs.CY
A theory and research-program preprint on routing proposed actions to the policy regimes that warrant authoritative review. It reports no empirical efficacy result and does not make permission, prohibition, legality, breach, compliance, certification, or enforcement determinations.
Lineage: policy routing, semantic representation, information retrieval, hierarchical classification, selective prediction, human review.
Read preprint and arXiv recordInformation retrieval and natural language processingPublic institute preprint; arXiv:2608.18448 in cs.IR, cross-listed cs.CL
A public information-retrieval preprint showing that, under a frozen HotpotQA FullWiki design and matched budget, relational context allocation recovered complete supporting-evidence sets more often than flat top-k retrieval. The outcome is evidence recovery, not answer accuracy.
Lineage: top-k retrieval, multi-hop question answering, graph-based retrieval, context-budget allocation, HotpotQA evaluation.
Read preprint and arXiv recordElectrical engineering and systems sciencePublic institute preprint; arXiv:2607.21771 in eess.SY, cross-listed cs.SY
A systems-and-control preprint on goal governance: a fast loop regulates under the current reference while a slow loop decides whether that reference should move, tested in a preregistered synthetic experiment that produced a negative result.
Lineage: W. Ross Ashby's ultrastability, physiological allostasis, bio-inspired allostatic control, adaptive control, supervisory control, reference governors, model predictive control.
Read preprint and arXiv record