Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces RH-RAG, a multi-agent framework designed for secure long-form content generation in privacy-sensitive environments, addressing the shortcomings of existing retrieval-augmented generation (RAG) methods. By employing a Planner Agent for global document outlining, a Writer Agent for coherent section-wise content generation, and a Checker Agent for factual verification, RH-RAG enhances the reliability and coherence of outputs while maintaining data privacy. Evaluations across various domains reveal that RH-RAG significantly outperforms traditional RAG approaches in factual grounding and semantic coherence, rivaling proprietary systems without compromising security.
RH-RAG achieves competitive reliability with proprietary cloud systems while ensuring data privacy, transforming long-form content generation in sensitive contexts.
Generating long-form content from extensive internal reports remains challenging for organizations operating under strict privacy and security constraints, where proprietary cloud-based LLM APIs are often not viable. While locally deployed open-weight models offer a privacy-preserving alternative, existing retrieval-augmented generation (RAG) approaches on smaller models frequently lack effective global planning and accumulate factual inconsistencies over long outputs. To address these limitations, we present RH-RAG, a multi-agent framework for secure and trustworthy long form generation using local language models. RH-RAG decomposes generation into three coordinated stages: a Planner Agent that constructs a global document outline from high-level semantic summaries, a Writer Agent that incrementally generates coherent section-wise content using bounded coherence memory, and a Checker Agent that mitigates hallucinations through natural language inference-based factual verification and an attestation-driven revision loop. The framework further employs a dual-level retrieval index that supports efficient planning and fine-grained contextual generation on consumer-grade hardware. Evaluations across literary, financial, and legal domains demonstrate that RH-RAG consistently improves factual grounding, semantic coherence, and document-level alignment compared to standard and hierarchical RAG baselines, while achieving reliability competitive with proprietary cloud-based systems without compromising data privacy.