RAPID: Efficient Retrieval-Augmented Long Text Generation with Writing Planning and Information Discovery

2025-03-02

AI Summary

Large language models (LLMs) face three key challenges in generating knowledge-intensive long-form text (e.g., encyclopedia articles): factual hallucination, topical incoherence, and high end-to-end latency. To address these, we propose a retrieval-augmented writing planning and information discovery framework featuring a novel three-module architecture: (1) retrieval-augmented structured outline generation, (2) attribute-constrained factual retrieval, and (3) plan-guided coherent decoding. Our method integrates retrieval-augmented generation (RAG), attribute-driven search, and structured outline modeling. We further introduce FreshWiki-2024, a new benchmark for evaluating long-form knowledge-grounded generation. Experiments demonstrate that our approach outperforms state-of-the-art methods across all core metrics—factuality, discourse coherence, and end-to-end latency—achieving significant improvements in long-text quality and accelerating generation by 2.3×.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Generating knowledge-intensive and comprehensive long texts, such as encyclopedia articles, remains significant challenges for Large Language Models. It requires not only the precise integration of facts but also the maintenance of thematic coherence throughout the article. Existing methods, such as direct generation and multi-agent discussion, often struggle with issues like hallucinations, topic incoherence, and significant latency. To address these challenges, we propose RAPID, an efficient retrieval-augmented long text generation framework. RAPID consists of three main modules: (1) Retrieval-augmented preliminary outline generation to reduce hallucinations, (2) Attribute-constrained search for efficient information discovery, (3) Plan-guided article generation for enhanced coherence. Extensive experiments on our newly compiled benchmark dataset, FreshWiki-2024, demonstrate that RAPID significantly outperforms state-of-the-art methods across a wide range of evaluation metrics (e.g. long-text generation, outline quality, latency, etc). Our work provides a robust and efficient solution to the challenges of automated long-text generation.

Problem

• Addresses challenges in generating knowledge-intensive long texts.
• Reduces hallucinations and improves thematic coherence in text generation.
• Enhances efficiency and coherence in automated long-text generation.

Innovation

• Retrieval-augmented preliminary outline generation
• Attribute-constrained search for information discovery
• Plan-guided article generation for coherence