Metraipolis › archive › dc_c7ca73a130c8
Knowledge-Graph-Gated Defactualization for Style-Controllable and Fact-Preserving Generation in Agentic Conversational AI — arXiv:2608.20393
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.20393
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T10:18:16.499Z
- SHA-256 of the archived text
8bb6c6de678f9981ce0167525e740fe7406c983ed2a50dcf58ba56bf06ed3a7b- Stable URL of this page
https://metraipolis.com/documento/dc_c7ca73a130c8
How to cite
Knowledge-Graph-Gated Defactualization for Style-Controllable and Fact-Preserving Generation in Agentic Conversational AI — arXiv:2608.20393 — arXiv (en), https://arxiv.org/abs/2608.20393 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c7ca73a130c8 (Metraipolis archive dc_c7ca73a130c8, sha256 8bb6c6de678f9981ce0167525e740fe7406c983ed2a50dcf58ba56bf06ed3a7b)
BibTeX
@misc{dcc7ca73a130c8,
title = {Knowledge-Graph-Gated Defactualization for Style-Controllable and Fact-Preserving Generation in Agentic Conversational AI — arXiv:2608.20393},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.20393},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c7ca73a130c8; sha256 8bb6c6de678f9981ce0167525e740fe7406c983ed2a50dcf58ba56bf06ed3a7b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.20393. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Agentic large language models (LLMs) deployed in fact-sensitive applications such as customer support must simultaneously preserve factual correctness and generate responses in a controllable stylistic register. Activation steering enables fine-tuning-free style control by perturbing hidden representations, but it lacks an explicit mechanism for distinguishing verifiable facts from stylistic content, leading to semantic leakage. We address this challenge through \emph{Defactualize-Steer-Rehydrate} (DSR), a knowledge-engineering framework that integrates a typed, salience-weighted knowledge graph (KG) with activation steering. DSR extracts salient entities using a layered regex or NER or lexical-classifier pipeline, replaces them with typed placeholders prior to steering, and deterministically restores verified values through salience-guided rehydration after generation. DSR is evaluated across six LLaMA-family models (1B--13B parameters) on 600 A2A-generated customer-support cases (1,200 generations), with a dedicated KG ablation study. DSR significantly increases verified-entity recovery relative to a steering-only baseline (Cohen's $d=0.225$, $p_{\text{Bonf}}=1.0\times10^{-4}$), though the absolute recovery rate remains modest, while preserving effective style control across diverse model families. Layer-wise separability and steering-strength diagnostics further show previously unexplored interactions between representation-level steering and factual grounding. hese results demonstrate that explicit knowledge engineering can systematically enhance trustworthy, controllable, and reproducible generative AI without requiring model fine-tuning. Code, cached steering vectors, and evaluation scripts are publicly released to support reproducibility.\footnote{ this https URL }
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-24 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.20393 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-18T10:17:37.569Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-08-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).