Training the Knowledge Base through Evidence Distillation and Write-Back Enrichment

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Lu, Yuxing, Zhao, Xukai, Wu, Wei, Wang, Jinzhuo
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912984017666048
author Lu, Yuxing
Zhao, Xukai
Wu, Wei
Wang, Jinzhuo
author_facet Lu, Yuxing
Zhao, Xukai
Wu, Wei
Wang, Jinzhuo
contents The knowledge base in a retrieval-augmented generation (RAG) system is typically assembled once and never revised, even though the facts a query requires are often fragmented across documents and buried in irrelevant content. We argue that the knowledge base should be treated as a trainable component and propose WriteBack-RAG, a framework that uses labeled examples to identify where retrieval succeeds, isolate the relevant documents, and distill them into compact knowledge units that are indexed alongside the original corpus. Because the method modifies only the corpus, it can be applied once as an offline preprocessing step and combined with any RAG pipeline. Across four RAG methods, six benchmarks, and two LLM backbones, WriteBack-RAG improves every evaluated setting, with gains averaging +2.14%. Cross-method transfer experiments further show that the distilled knowledge benefits RAG pipelines other than the one used to produce it, confirming that the improvement resides in the corpus itself.
format Preprint
id arxiv_https___arxiv_org_abs_2603_25737
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Training the Knowledge Base through Evidence Distillation and Write-Back Enrichment
Lu, Yuxing
Zhao, Xukai
Wu, Wei
Wang, Jinzhuo
Artificial Intelligence
Computation and Language
Information Retrieval
The knowledge base in a retrieval-augmented generation (RAG) system is typically assembled once and never revised, even though the facts a query requires are often fragmented across documents and buried in irrelevant content. We argue that the knowledge base should be treated as a trainable component and propose WriteBack-RAG, a framework that uses labeled examples to identify where retrieval succeeds, isolate the relevant documents, and distill them into compact knowledge units that are indexed alongside the original corpus. Because the method modifies only the corpus, it can be applied once as an offline preprocessing step and combined with any RAG pipeline. Across four RAG methods, six benchmarks, and two LLM backbones, WriteBack-RAG improves every evaluated setting, with gains averaging +2.14%. Cross-method transfer experiments further show that the distilled knowledge benefits RAG pipelines other than the one used to produce it, confirming that the improvement resides in the corpus itself.
title Training the Knowledge Base through Evidence Distillation and Write-Back Enrichment
topic Artificial Intelligence
Computation and Language
Information Retrieval
url https://arxiv.org/abs/2603.25737