English

RING: Retrieval-Internalized Generation for Continual Large-Scale Knowledge Injection

Computation and Language 2026-08-03 v1 Artificial Intelligence

Abstract

Retrieval-augmented generation (RAG) improves factuality but adds latency and engineering overhead at serving time. We propose RING (Retrieval-Internalized Generation), a holistic paradigm spanning both architecture and training that injects large-scale external knowledge into a \textit{Mixture-of-Memory Experts} and learns parametric search over this internal memory via reinforcement learning, removing the external retriever entirely. Training proceeds in three stages: continued pre-training injects new corpora into a Knowledge Expert via our novel \textit{Dual Causal Attention}; supervised fine-tuning teaches a ``search-then-answer'' pattern; and reinforcement learning with hierarchical rewards optimizes the routing-and-search policy over the parametric memory. Unlike prior parametric injection methods that pair internal memory with a fixed or rule-based retriever, RING {learns} its retrieval policy directly from task signals. We further frame RING theoretically as a search-free approximation to the classical RAG objective. To evaluate large-scale injection of genuinely {new} knowledge without test-time leakage, we further construct News-2025, a benchmark built from news strictly post-dating the base LLM's pretraining cutoff. RING matches or surpasses both search-based RAG and parametric injection baselines in accuracy and efficiency.

Cite

@article{arxiv.2608.01630,
  title  = {RING: Retrieval-Internalized Generation for Continual Large-Scale Knowledge Injection},
  author = {Shicheng Xu and Liang Pang and Liyi Chen and Zihao Wei and Jingcheng Deng and Yan Gao and Yi Wu and Yao Hu and Huawei Shen and Xueqi Cheng},
  journal= {arXiv preprint arXiv:2608.01630},
  year   = {2026}
}

Comments

16 pages