{"slug": "evident-building-trustworthy-enterprise-assistants-through-evidence-groundedness", "title": "EvidenT: Building Trustworthy Enterprise Assistants through Evidence Groundedness and Traceability", "summary": "EvidenT, a lightweight pipeline that verifies extracted evidence against retrieved documents before answer generation without model retraining, improved gold-source hit rate by an average of 29% over prompting baselines across approximately 500 real enterprise queries, according to an arXiv paper (arXiv:2609.22537v1). The pipeline combines structured passage extraction with deterministic lexical alignment to filter unsupported content, correct citation drift, and preserve source-span traceability, and it produced no citations to nonretrieved URLs while achieving near-saturated answer-to-source lexical coverage. The work targets citation drift, unsupported content, and weak source traceability in retrieval augmented generation over heterogeneous enterprise data.", "body_md": "arXiv:2609.22537v1 Announce Type: new \nAbstract: Enterprise AI assistants must produce responses that are verifiable and traceable to source evidence. However, retrieval augmented generation (RAG) over heterogeneous enterprise data can suffer from citation drift, unsupported content, and weak source traceability. We present EvidenT (T = Trust + Transparency + Traceability), a lightweight pipeline that verifies extracted evidence against retrieved documents before answer generation, without model retraining. EvidenT combines structured passage extraction with deterministic lexical alignment to filter unsupported content, correct citation drift, and preserve source-span traceability. On approximately 500 real enterprise queries, EvidenT improves gold-source hit rate by an average of 29% over prompting baselines, produces no citations to nonretrieved urls, and achieves near-saturated answer-to-source lexical coverage.", "url": "https://wpnews.pro/news/evident-building-trustworthy-enterprise-assistants-through-evidence-groundedness", "canonical_source": "https://arxiv.org/abs/2609.22537", "published_at": "2026-09-23 04:00:00+00:00", "updated_at": "2026-09-23 04:26:41.307474+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "ai-safety"], "entities": ["EvidenT", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/evident-building-trustworthy-enterprise-assistants-through-evidence-groundedness", "markdown": "https://wpnews.pro/news/evident-building-trustworthy-enterprise-assistants-through-evidence-groundedness.md", "text": "https://wpnews.pro/news/evident-building-trustworthy-enterprise-assistants-through-evidence-groundedness.txt", "jsonld": "https://wpnews.pro/news/evident-building-trustworthy-enterprise-assistants-through-evidence-groundedness.jsonld"}}