04:00
2026-07-15
arxiv.org
artificial-intelligence
Ontology-Amplified Distillation and Contextuality Auditing for Sovereign Enterprise Language Models: A Combined Proof-of-Mechanism and Negative-Results Method Study
A proof-of-mechanism study found that a Qwen3.6-27B student model distilled via ontology-amplified techniques matched GPT-5 on grounding Vietnamese financial tasks, but the result was underpowered to …