From Document Silos to Process Intelligence: A Multi-Layer Knowledge Graph for CMC Process Development
3.60T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2609.11493.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryPresents a modular agentic-AI platform that converts heterogeneous chemistry, manufacturing, and controls (CMC) documents into a dual-layer knowledge graph (lexical + ontology-aligned intelligence), with LLM agents routing queries between layers. Evaluated on 505 questions from 38 Sanofi reports using a novel three-tier protocol: 95% Tier-1 multiple-choice accuracy and 85% Tier-2 LLM-judge pass rate.
Why it mattersConcrete dual-layer knowledge-graph architecture and a reusable three-tier evaluation protocol, demonstrated against proprietary pharmaceutical data, with a failure taxonomy that standard accuracy metrics miss.
Cited by
No citations on record.