Retrieve-Plan-Generation: An Iterative Planning and Answering Framework for Knowledge-Intensive LLM Generation
Yuanjie Lyu, Zihan Niu, Zheyong Xie, Chao Zhang, Tong Xu, Yang Wang, Enhong Chen
Abstract
Despite the significant progress of large language models (LLMs) in various tasks, they often produce factual errors due to their limited internal knowledge. Retrieval-Augmented Generation (RAG), which enhances LLMs with external knowledge sources, offers a promising solution. However, these methods can be misled by irrelevant paragraphs in retrieved documents. Due to the inherent uncertainty in LLM generation, inputting the entire document may introduce off-topic information, causing the model to deviate from the central topic and affecting the relevance of the generated content. To address these issues, we propose the Retrieve-Plan-Generation (RPG) framework. RPG generates plan tokens to guide subsequent generation in the plan stage. In the answer stage, the model selects relevant fine-grained paragraphs based on the plan and uses them for further answer generation. This plan-answer process is repeated iteratively until completion, enhancing generation relevance by focusing on specific topics. To implement this framework efficiently, we utilize a simple but effective multi-task prompt-tuning method, enabling the existing LLMs to handle both planning and answering. We comprehensively compare RPG with baselines across 5 knowledge-intensive generation tasks, demonstrating the effectiveness of our approach. 1 * Corresponding author. † Equal Contribution. 1 Our code and models will be publicly available at https://github.com/haruhi-sudo/RPG . Question: How do jellyfish function without brains or nervous systems? Retrieved Documents: Title: Jellyfish, Simple yet Mysterious Ocean Creatures. Jellyfish lack a brain but possess a rudimentary nervous system and neural cells capable of sensing external stimuli such as light and touch[...] They likely coordinate these functions through a nerve ring [...] Jellyfish are invertebrates [...] Some species have remarkably brief lifespans, lasting merely hours. Yet their vitality is remarkably resilient[...] Most strikingly, jellyfish possess an astounding capacity for regeneration [...] Normal RAG: Despite their seemingly lifeless appearance, jellyfish are among the most wondrous creatures on Earth! Though lacking a brain, jellyfish possess a rudimentary nervous system and neural[...] Their bodies also exhibit a peculiar self-healing ability, rapidly regenerating even when cut into pieces and reassembled. Furthermore, jellyfish can bioluminescent, illuminating the dark [...] Plan Before Generate: [Plan: Body composition and structure of jellyfish] <p>Jellyfish…Their bodies are almost entirely made up of gelatinous hydrogel tissue, with a water content of more than 98%...</p>Jellyfish are indeed very simple creatures, almost more than 98% of which are made up of water. [Plan: Nervous system of jellyfish] <p>Jellyfish barely have nervous systems…They have primitive nerve cells that help them</p>However, although jellyfish do not have brains, they do have a rough nervous system that can sense external stimuli such as light and touch[...] [Plan: Nerve ring of jellyfish] <p>..</p>Jellyfish coordinate through nerve rings [...]
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext dfd51b3d-3c0c-4fca-a63d-3840b69205b5Cited by top-tier papers3
- RAS: Retrieval-And-Structuring for Knowledge-Intensive LLM GenerationPengcheng Jiang, Lang Cao, Ruike Zhu, Minhao Jiang et al.ICLR 2026 · 20 citations
- Beyond Prompting: An Efficient Embedding Framework for Open-Domain Question AnsweringZhanghao Hu, Hanqi Yan, Qinglin Zhu, Zhenyi Shen et al.ACL 2025
- s3: You Don't Need That Much Data to Train a Search Agent via RLPengcheng Jiang, Xueqiang Xu, Jiacheng Lin, Jinfeng Xiao et al.EMNLP 2025
Builds on16
- Retrieval-Augmented Generation for Knowledge-Intensive NLP TasksPatrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni et al.NeurIPS 2020 · 19,162 citations
- LoRA: Low-Rank Adaptation of Large Language ModelsEdward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu et al.ICLR 2022 · 18,833 citations
- Retrieval Augmented Language Model Pre-TrainingKelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat et al.ICML 2020 · 2,937 citations
- Self-RAG: Learning to Retrieve, Generate, and Critique through Self-ReflectionAkari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil et al.ICLR 2024 · 1,798 citations
- MAUVE: Measuring the Gap Between Neural Text and Human Text using Divergence FrontiersKrishna Pillutla, Swabha Swayamdipta, Rowan Zellers, John Thickstun et al.NeurIPS 2021 · 606 citations
Related papers
- SAGE: A Framework of Precise Retrieval for RAGJintao Zhang, Guoliang Li, Jinyang SuICDE 2025 · 9 citations
- DeepRAG: Thinking to Retrieve Step by Step for Large Language ModelsXinyan Guan, Jiali Zeng, Fandong Meng, Chunlei Xin et al.ICLR 2026 · 30 citations
- Separate the Wheat from the Chaff: Winnowing Down Divergent Views in Retrieval Augmented GenerationSong Wang, Zihan Chen, Peng Wang, Zhepei Wei et al.EMNLP 2025 · 1 citation
- Disco-RAG: Discourse-Aware Retrieval-Augmented GenerationDongqi Liu, Hang Ding, Qiming Feng, Xurong Xie et al.ACL 2026
- REAP: Enhancing RAG with Recursive Evaluation and Adaptive Planning for Multi-Hop Question AnsweringYijie Zhu, Haojie Zhou, Wanting Hong, Tailin Liu et al.AAAI 2026
