Abstract
Revision 1.2 of a defensive publication disclosing deterministic, model-independent grounding-integrity mechanisms for retrieval-augmented and agentic LLM systems. The mechanisms make citation resolvability an invariant of every output mode routed through a cross-tool evidence registry — detect → bounce → strip with compound-citation surgery, an ordering rule, and a strip-event notice — and make degradations of grounding (a removed citation, a deleted source, an unverified item, a suspect "not found") visible to the reviewer rather than silent. Revision 1.2 adds Technique G (decompose-and-verify: per-unit grounding, byte-identical scoped regeneration, de-correlated verification, citation tombstones, honest terminal states), a precise trust model and guarantee taxonomy including the removal-authority principle, and a substantially expanded, primary-source-verified related-work and patent-literature account. Nothing disclosed in v1.0 (27 June 2026) is retracted; the additions are disclosed as of this version's date.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wenzel, Roberto A., "Deterministic, Model-Independent Grounding-Integrity Mechanisms for Retrieval-Augmented and Agentic Language-Model Systems", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11357