檢索增強生成（RAG）讓大型語言模型（LLM）能即時引用外部證據，但只要檢索結果摻雜雜訊或對抗片段時，主流方法往往拋棄自身參數知識，直接複述錯誤內容並引發幻覺。既有工作雖提出各自策略解決雜訊問題，整體卻仍缺乏對如何系統性調度與對齊內外知識的深入探討；近期機制層研究更發現，面對強污染檢索，LLM 依賴內部記憶的程度急劇下降。
為此，我們提出 extbf{IDEAL-RAG}，一個三階段指令驅動式框架，讓模型先顯式喚醒自身記憶，再讓內、外兩源各自生成自洽論點，最後在鏈結模組對證據交叉驗證並產生可追蹤推理，全程無需改動檢索器或額外人工標註。此外，我們提供一套機制層面診斷方法：以自建的對抗敏感度指標（CSS）搭配現有的分層參數知識分數（PKS），觀察 RAG 在強污染環境下「知識-FFN」通道的動態，並驗證「預先提取並平衡殘差」的步驟確實能抑制內知失衡、降低幻覺風險。
在實驗中表明， IDEAL-RAG 在乾淨檢索下，與現有的強勁基準方法 InstructRAG 相當；於對抗式強污染測試集中，則準確率最高提升 +22.8 %，準確率衰減率減半。透過機制分析進一步證實，本框架可穩定調用內知並維持答案置信度。結果證實：在 LLM「已知」與「新讀」之間先立論點、再理性對照，是構築更可靠 RAG 系統的有效途徑。

Retrieval-augmented generation (RAG) lets large language models (LLMs) cite fresh external evidence on demand. Yet when the retrieved texts contain noise or adversarial edits, most RAG systems ignore their parametric knowledge and simply echo the wrong passages, producing hallucinations. Existing studies propose various denoising strategies, but few explore how to systematically balance and align internal and external knowledge. Recent probing work even shows that, under heavy corruption, an LLM’s reliance on its stored memory drops sharply.
We introduce extbf{IDEAL-RAG}, a three-stage, instruction-driven framework. The model (i) explicitly recalls its latent knowledge, (ii) lets the internal and retrieved sources each craft a self-contained standpoint, and (iii) cross-checks these standpoints in a linking module to yield a traceable rationale—all without changing the retriever or adding manual labels. To inspect what happens inside the network, we pair a new Counterfactual Sensitivity Score (CSS) with the existing layer-wise Parametric Knowledge Score (PKS), revealing how the “knowledge-FFN” pathway behaves under strong noise and showing that our “extract-then-balance” step reduces hallucination risk.
Experiments demonstrate that IDEAL-RAG matches the strong InstructRAG baseline on clean retrieval, while under adversarially corrupted contexts, it boosts exact-match accuracy by up to +22.8 % and cuts accuracy loss by half. CSS and PKS analyses confirm that the framework keeps answer confidence steady by drawing on internal facts when external evidence is unreliable. These findings indicate that first establishing separate standpoints and then reasoning over their agreement is an effective route to more dependable RAG systems.

[1] Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. Self-rag: Learning to retrieve, generate, and critique through self-reflection. In The Twelfth International Conference on Learning Representations, 2023.
[2] Bernd Bohnet, Vinh Q Tran, Pat Verga, Roee Aharoni, Daniel Andor, Livio BaldiniSoares, Massimiliano Ciaramita, Jacob Eisenstein, Kuzman Ganchev, Jonathan Herzig,et al. Attributed question answering: Evaluation and modeling for attributed large language models. arXiv preprint arXiv:2212.08037, 2022.
[3] Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020.
[4] Sébastien Bubeck, Varun Chadrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, et al. Sparks of artificial general intelligence: Early experiments with gpt-4, 2023.
[5] Danqi Chen, Adam Fisch, Jason Weston, and Antoine Bordes. Reading wikipedia to answer open-domain questions. In 55th Annual Meeting of the Association for Computational Linguistics, ACL 2017, pages 1870–1879. Association for Computational Linguistics (ACL), 2017.
[6] Jiawei Chen, Hongyu Lin, Xianpei Han, and Le Sun. Benchmarking large language models in retrieval-augmented generation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 17754–17762, 2024.
[7] Florin Cuconasu, Giovanni Trappolini, Federico Siciliano, Simone Filice, Cesare Campagnano, Yoelle Maarek, Nicola Tonellotto, and Fabrizio Silvestri. The power of noise: Redefining retrieval for rag systems. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, pages 719–729, 2024.
[8] Bhuwan Dhingra, Jeremy R Cole, Julian Martin Eisenschlos, Daniel Gillick, Jacob Eisenstein, and William W Cohen. Time-aware language models as temporal knowledge bases. Transactions of the Association for Computational Linguistics, 10:257–273, 2022.
[9] Feiteng Fang, Yuelin Bai, Shiwen Ni, Min Yang, Xiaojun Chen, and Ruifeng Xu. Enhancing noise robustness of retrieval-augmented language models with adaptive adversarial training. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 10028–10039, 2024.
[10] Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Qianyu Guo, Meng Wang, et al. Retrieval-augmented generation for large language models: A survey. CoRR, 2023.
[11] Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Mingwei Chang. Retrieval augmented language model pre-training. In International conference on machine learning, pages 3929–3938. PMLR, 2020.
[12] Zeyu Han, Chao Gao, Jinyang Liu, Jeff Zhang, and Sai Qian Zhang. Parameter-efficient fine-tuning for large models: A comprehensive survey. arXiv preprint arXiv:2403.14608, 2024.
[13] Xanh Ho, Anh-Khoa Duong Nguyen, Saku Sugawara, and Akiko Aizawa. Constructing a multi-hop qa dataset for comprehensive evaluation of reasoning steps. arXiv preprint arXiv:2011.01060, 2020.
[14] Or Honovich, Roee Aharoni, Jonathan Herzig, Hagai Taitelbaum, Doron Kukliansy, Vered Cohen, Thomas Scialom, Idan Szpektor, Avinatan Hassidim, and Yossi Matias. True: Re-evaluating factual consistency evaluation. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3905–3920, 2022.
[15] Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. arxiv 2021. arXiv preprint arXiv:2106.09685, 2021.
[16] Raisa Islam and Owana Marzia Moushi. Gpt-4o: The cutting-edge advancement in multimodal llm. Authorea Preprints, 2024.
[17] Gautier Izacard, Mathilde Caron, Lucas Hosseini, Sebastian Riedel, Piotr Bojanowski, Armand Joulin, and Edouard Grave. Unsupervised dense information retrieval with contrastive learning. Transactions on Machine Learning Research, 2023.
[18] Gautier Izacard, Patrick Lewis, Maria Lomeli, Lucas Hosseini, Fabio Petroni, Timo Schick, Jane Dwivedi-Yu, Armand Joulin, Sebastian Riedel, and Edouard Grave. Atlas: Few-shot learning with retrieval augmented language models. Journal of Machine Learning Research, 24(251):1–43, 2023.
[19] Soyeong Jeong, Jinheon Baek, Sukmin Cho, Sung Ju Hwang, and Jong Park. Adaptive-rag: Learning to adapt retrieval-augmented large language models through question complexity. In NAACL-HLT, 2024.
[20] Zhengbao Jiang, Frank F Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang, Jamie Callan, and Graham Neubig. Active retrieval augmented generation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 7969–7992, 2023.
[21] Mandar Joshi, Eunsol Choi, Daniel S Weld, and Luke Zettlemoyer. Triviaqa: A large scale distantly supervised challenge dataset for reading comprehension. arXiv preprint arXiv:1705.03551, 2017.
[22] Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick SH Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. Dense passage retrieval for open-domain question answering. In EMNLP (1), pages 6769–6781, 2020.
[23] Urvashi Khandelwal, Omer Levy, Dan Jurafsky, Luke Zettlemoyer, and Mike Lewis. Generalization through memorization: Nearest neighbor language models. In Proceedings of ICLR, 2020.
[24] Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, et al. Natural questions: a benchmark for question answering research. Transactions of the Association for Computational Linguistics, 7:453–466, 2019.
[25] Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph Gonzalez, Hao Zhang, and Ion Stoica. Efficient memory management for large language model serving with pagedattention. In Proceedings of the 29th Symposium on Operating Systems Principles, pages 611–626, 2023.
[26] Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, et al. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in neural information processing systems, 33:9459–9474, 2020.
[27] Daliang Li, Ankit Singh Rawat, Manzil Zaheer, Xin Wang, Michal Lukasik, Andreas Veit, Felix Yu, and Sanjiv Kumar. Large language models with controllable working memory. In Findings of the Association for Computational Linguistics: ACL 2023, pages 1774–1793, 2023.
[28] Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017.
[29] Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Hannaneh Hajishirzi, and Daniel Khashabi. When not to trust language models: Investigating effectiveness and limitations of parametric and non-parametric memories. arXiv preprint, 2022.
[30] Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Daniel Khashabi, and Hannaneh Hajishirzi. When not to trust language models: Investigating effectiveness of parametric and non-parametric memories. In ACL, 2023.
[31] Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35:27730–27744, 2022.
[32] Samyam Rajbhandari, Jeff Rasley, Olatunji Ruwase, and Yuxiong He. Zero: Memory optimizations toward training trillion parameter models. In SC20: International Conference for High Performance Computing, Networking, Storage and Analysis, pages 1–16. IEEE, 2020.
[33] Ori Ram, Yoav Levine, Itay Dalmedigos, Dor Muhlgay, Amnon Shashua, KevinLeyton-Brown, and Yoav Shoham. In-context retrieval-augmented language models. Transactions of the Association for Computational Linguistics, 11:1316–1331, 2023.
[34] Adam Roberts, Colin Raffel, and Noam Shazeer. How much knowledge can you pack into the parameters of a language model? In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5418–5426, 2020.
[35] Stephen E Robertson and Steve Walker. Some simple effective approximations to the 2-poisson model for probabilistic weighted retrieval. In SIGIR＇94: Proceedings of the Seventeenth Annual International ACM-SIGIR Conference on Research and Development in Information Retrieval, organised by Dublin City University, pages 232–241. Springer, 1994.
[36] John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017.
[37] Freda Shi, Xinyun Chen, Kanishka Misra, Nathan Scales, David Dohan, Ed H Chi, Nathanael Schärli, and Denny Zhou. Large language models can be easily distracted by irrelevant context. In International Conference on Machine Learning, pages 31210–31227. PMLR, 2023.
[38] ZhongXiang Sun, Xiaoxue Zang, Kai Zheng, Jun Xu, Xiao Zhang, Weijie Yu, Yang Song, and Han Li. Redeep: Detecting hallucination in retrieval-augmented generation via mechanistic interpretability. In The Thirteenth International Conference on Learning Representations, 2025.
[39] Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M Dai, Anja Hauth, Katie Millican, et al. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805, 2023.
[40] Nandan Thakur, Luiz Bonifacio, Xinyu Zhang, Odunayo Ogundepo, Ehsan Kamalloo, David Alfonso-Hermelo, Xiaoguang Li, Qun Liu, Boxing Chen, Mehdi Rezagholizadeh, et al. Nomiracl: Knowing when you don’t know for robust multilingual retrieval-augmented generation. CoRR, 2023.
[41] Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023.
[42] Hitesh Wadhwa, Rahul Seetharaman, Somyaa Aggarwal, Reshmi Ghosh, Samyadeep Basu, Soundararajan Srinivasan, Wenlong Zhao, Shreyas Chaudhari, and Ehsan Aghazadeh. From rags to rich parameters: Probing how language models utilize external knowledge over parametric information for factual queries. CoRR, 2024.
[43] Fei Wang, Xingchen Wan, Ruoxi Sun, Jiefeng Chen, and Sercan Ö Arık. Astute rag: Overcoming imperfect retrieval augmentation and knowledge conflicts for large language models. arXiv preprint arXiv:2410.07176, 2024.
[44] Zihao Wang, Anji Liu, Haowei Lin, Jiaqi Li, Xiaojian Ma, and Yitao Liang. Rat: Retrieval augmented thoughts elicit context-aware reasoning in long-horizon generation. CoRR, 2024.
[45] Zhepei Wei, Wei-Lin Chen, and Yu Meng. Instructrag: Instructing retrieval-augmented generation via self-synthesized rationales. In The Thirteenth International Conference on Learning Representations, 2025.
[46] Siye Wu, Jian Xie, Jiangjie Chen, Tinghui Zhu, Kai Zhang, and Yanghua Xiao. How easily do irrelevant inputs skew the responses of large language models? CoRR, 2024.
[47] Yuanhao Wu, Juno Zhu, Siliang Xu, Kashun Shum, Cheng Niu, Randy Zhong, Juntong Song, and Tong Zhang. Ragtruth: A hallucination corpus for developing trustworthy retrieval-augmented language models. CoRR, 2024.
[48] Chong Xiang, Tong Wu, Zexuan Zhong, David Wagner, Danqi Chen, and Prateek Mittal. Certifiably robust rag against retrieval corruption. In ICML 2024 Next Generation of AI Safety Workshop, 2024.
[49] Peng Xu, Wei Ping, Xianchao Wu, Lawrence McAfee, Chen Zhu, Zihan Liu, Sandeep Subramanian, Evelina Bakhturina, Mohammad Shoeybi, and Bryan Catanzaro. Retrieval meets long context large language models. In The Twelfth International Conference on Learning Representations, 2023.
[50] Rongwu Xu, Zehan Qi, Zhijiang Guo, Cunxiang Wang, Hongru Wang, Yue Zhang, and Wei Xu. Knowledge conflicts for llms: A survey. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 8541–8565, 2024.
[51] Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W Cohen, Ruslan Salakhutdinov, and Christopher D Manning. Hotpotqa: A dataset for diverse, explainable multi-hop question answering. arXiv preprint arXiv:1809.09600, 2018.
[52] Ori Yoran, Tomer Wolfson, Ori Ram, and Jonathan Berant. Making retrieval-augmented language models robust to irrelevant context. In ICLR 2024 Workshop on Large Language Model (LLM) Agents, 2024.
[53] Wenhao Yu, Dan Iter, Shuohang Wang, Yichong Xu, Mingxuan Ju, S Sanyal, Chenguang Zhu, Michael Zeng, and Meng Jiang. Generate rather than retrieve: Large language models are strong context generators. In International Conference on Learning Representations, 2023.
[54] Wenhao Yu, Hongming Zhang, Xiaoman Pan, Peixin Cao, Kaixin Ma, Jian Li, Hongwei Wang, and Dong Yu. Chain-of-note: Enhancing robustness in retrieval-augmented language models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 14672–14685, 2024.
[55] Wenhao Yu, Zhihan Zhang, Zhenwen Liang, Meng Jiang, and Ashish Sabharwal. Improving language models via plug-and-play retrieval feedback. arXiv preprint arXiv:2305.14002, 2023.
[56] Tianjun Zhang, Shishir G Patil, Naman Jain, Sheng Shen, Matei Zaharia, Ion Stoica, and Joseph E Gonzalez. Raft: Adapting language model to domain specific rag. In First Conference on Language Modeling, 2024.
[57] Ruochen Zhao, Xingxuan Li, Shafiq Joty, Chengwei Qin, and Lidong Bing. Verify-and-edit: A knowledge-enhanced chain-of-thought framework. In The 61st Annual Meeting Of The Association For Computational Linguistics, 2023.
[58] Yutao Zhu, Huaying Yuan, Shuting Wang, Jiongnan Liu, Wenhan Liu, Chenlong Deng, Haonan Chen, Zheng Liu, Zhicheng Dou, and Ji-Rong Wen. Large language models for information retrieval: A survey. arXiv preprint arXiv:2308.07107, 2023.