Director-Class AI: Coherence Engine v1.4.1

Zenodo (CERN European Organization for Nuclear Research) 2026 Open access

Read full text
Authors: Miroslav Sotek
DOI: 10.5281/zenodo.18822166
Cited by: 0

Abstract

Real-time LLM hallucination guardrail. Dual-entropy scoring (NLI contradiction via FactCG-DeBERTa-v3-Large plus RAG fact-checking) with token-level streaming halt. 75.8 percent balanced accuracy on LLM-AggreFact (29K samples). v1.4.1 adds ORT_ENABLE_ALL graph optimization, ONNX GPU batch inference at 14.6 ms per pair, and MiniCheck evaluation harness. Integrates with LangChain, LlamaIndex, LangGraph, Haystack, and CrewAI.

Fields of Study & Subjects

Topic Modeling Natural Language Processing Techniques Misinformation and Its Impacts

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
1 March 2026
Language
EN