Elamur Research
☰
home / papers / 2023 / Benchmarking Large Language Models in Retrieval-Augmented Generation

Benchmarking Large Language Models in Retrieval-Augmented Generation

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive license to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/2309.01431

Summary

Retrieval-augmented generation (RAG) is widely promoted as a way to reduce LLM hallucination and outdated knowledge. However, there has been little systematic evidence about where it fails for different models. The authors build RGB (Retrieval-Augmented Generation Benchmark), an English/Chinese corpus divided into four testbeds. Each testbed probes one skill:

  • coping with related but unhelpful documents;
  • declining to answer when no document holds the answer;
  • combining facts spread across several documents;
  • detecting false statements in retrieved text.

To keep the models' memorized knowledge from skewing results, questions are built from recent news. ChatGPT drafts question–answer pairs, a search API gathers web pages, a dense retriever reranks passages, and humans filter the output. Six LLMs are tested, including ChatGPT, ChatGLM2-6B and Vicuna-7b. The models tolerate some noise but often mix up similar facts. They seldom refuse when evidence is missing, handle multi-document questions poorly, and usually believe retrieved misinformation over their own knowledge, even when warned. RGB offers a diagnostic tool and shows that RAG pipelines need careful design.

Key points

  • Four target abilities: withstanding noisy documents, refusing to answer when evidence is absent, merging information from multiple sources, and resisting factually wrong retrieved content. Each has its own testbed in RGB. Code and data are at github.com/chen700564/RGB.
  • Data pipeline: news articles → event/question/answer extraction with gpt-3.5-turbo → Google Search API → dense-retrieval reranking of chunks (top-30) → human adjustment and filtering.
  • Models evaluated: ChatGPT, ChatGLM-6B, ChatGLM2-6B, Vicuna-7b, Qwen-7B-Chat and BELLE-7B.
  • Integration errors (ChatGLM2-6B): merging sub-answers (28% of errors), ignoring one sub-question (28%) and misaligning documents with sub-questions (6%). The authors suggest chain-of-thought decomposition but note it slows inference.
  • Counterfactual robustness: only models with over 70% closed-book accuracy were tested. Their accuracy of 77–91% without documents fell to 9–17% when given counterfactual documents. Error correction rates ranged from 25% to about 57%.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Other

Cited by

191 papers of our library cite it (of the 140673 whose bibliographies we read); 8 of them have a page here, 183 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2309.01431 · arXiv 2309.01431

@misc{chen2023benchmarking,
  title = {{Benchmarking Large Language Models in Retrieval-Augmented Generation}},
  author = {Chen, Jiawei and Lin, Hongyu and Han, Xianpei and Sun, Le},
  year = {2023},
  doi = {10.48550/arxiv.2309.01431},
  eprint = {2309.01431},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2309.01431}
}

References

47 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Adlakha, V.; BehnamGhader, P.; Lu, X. H.; Meade, N.; and Reddy, S. 2023. Evaluating Correctness and Faithfulness of Instruction-Following Models for Question Answering. arXiv:2307.16877. find
  2. Bang, Y.; Cahyawijaya, S.; Lee, N.; Dai, W.; Su, D.; Wilie, B.; Lovenia, H.; Ji, Z.; Yu, T.; Chung, W.; Do, Q. V.; Xu, Y.; and Fung, P. 2023. A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivity. arXiv:2302.04023. find our card
  3. Bian, N.; Liu, P.; Han, X.; Lin, H.; Lu, Y.; He, B.; and Sun, L. 2023. A Drop of Ink Makes a Million Think: The Spread of False Information in Large Language Models. arXiv:2305.04812. — [arXiv:2305.04812](https://arxiv.org/abs/2305.04812) find
  4. Borgeaud, S.; Mensch, A.; Hoffmann, J.; Cai, T.; Rutherford, E.; Millican, K.; van den Driessche, G.; Lespiau, J.-B.; Damoc, B.; Clark, A.; de Las Casas, D.; Guy, A.; Menick, J.; Ring, R.; Hennigan, T.; Huang, S.; Maggiore, L.; Jones, C.; Cassirer, A.; Brock, A.; Paganini, M.; Irving, G.; Vinyals, O… find
  5. Cai, D.; Wang, Y.; Bi, W.; Tu, Z.; Liu, X.; Lam, W.; and Shi, S. 2019a. Skeleton-to-Response: Dialogue Generation Guided by Retrieval Memory. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (L… find
  6. Cai, D.; Wang, Y.; Bi, W.; Tu, Z.; Liu, X.; and Shi, S. 2019b. Retrieval-guided Dialogue Response Generation via a Matching-to-Generation Framework. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Languag… find
  7. Cao, M.; Dong, Y.; Wu, J.; and Cheung, J. C. K. 2020. Factual Error Correction for Abstractive Summarization Models. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), 6251– 6258. Online: Association for Computational Linguistics. find
  8. Chang, Y.; Wang, X.; Wang, J.; Wu, Y.; Yang, L.; Zhu, K.; Chen, H.; Yi, X.; Wang, C.; Wang, Y.; Ye, W.; Zhang, Y.; Chang, Y.; Yu, P. S.; Yang, Q.; and Xie, X. 2023. A Survey on Evaluation of Large Language Models. arXiv:2307.03109. find our card
  9. Chiang, W.-L.; Li, Z.; Lin, Z.; Sheng, Y.; Wu, Z.; Zhang, H.; Zheng, L.; Zhuang, S.; Zhuang, Y.; Gonzalez, J. E.; Stoica, I.; and Xing, E. P. 2023. Vicuna: An Open-Source Chatbot Impressing GPT-4 with 90%* ChatGPT Quality. find
  10. Cui, J.; Li, Z.; Yan, Y.; Chen, B.; and Yuan, L. 2023. Chat-Law: Open-Source Legal Large Language Model with Integrated External Knowledge Bases. arXiv:2306.16092. find our card
  11. Drozdov, A.; Schärli, N.; Akyürek, E.; Scales, N.; Song, X.; Chen, X.; Bousquet, O.; and Zhou, D. 2023. Compositional Semantic Parsing with Large Language Models. In The Eleventh International Conference on Learning Representations. find
  12. Edward Beeching, N. H. S. H. N. L. N. R. O. S. L. T. T. W., Clémentine Fourrier. 2023. Open LLM Leaderboard. https://huggingface.co/spaces/HuggingFaceH4/open llm leaderboard. ↗
  13. Guo, B.; Zhang, X.; Wang, Z.; Jiang, M.; Nie, J.; Ding, Y.; Yue, J.; and Wu, Y. 2023. How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection. arXiv:2301.07597. find our card
  14. Guu, K.; Lee, K.; Tung, Z.; Pasupat, P.; and Chang, M.-W. 2020. REALM: Retrieval-Augmented Language Model Pre-Training. In Proceedings of the 37th International Conference on Machine Learning, ICML’20. JMLR.org. find
  15. He, H.; Zhang, H.; and Roth, D. 2022. Rethinking with Retrieval: Faithful Large Language Model Inference. arXiv:2301.00303. find
  16. Hendrycks, D.; Burns, C.; Basart, S.; Zou, A.; Mazeika, M.; Song, D.; and Steinhardt, J. 2021. Measuring Massive Multitask Language Understanding. In International Conference on Learning Representations. find our card
  17. Huang, Y.; Bai, Y.; Zhu, Z.; Zhang, J.; Zhang, J.; Su, T.; Liu, J.; Lv, C.; Zhang, Y.; Lei, J.; Fu, Y.; Sun, M.; and He, J. 2023. C-Eval: A Multi-Level Multi-Discipline Chinese Evaluation Suite for Foundation Models. arXiv preprint arXiv:2305.08322. find our card
  18. Izacard, G.; and Grave, E. 2021. Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, 874–880. Online: Association for Computational Linguis… find our card
  19. Izacard, G.; Lewis, P.; Lomeli, M.; Hosseini, L.; Petroni, F.; Schick, T.; Dwivedi-Yu, J.; Joulin, A.; Riedel, S.; and Grave, E. 2022. Atlas: Few-shot Learning with Retrieval Augmented Language Models. arXiv:2208.03299. find
  20. Ji, Z.; Lee, N.; Frieske, R.; Yu, T.; Su, D.; Xu, Y.; Ishii, E.; Bang, Y. J.; Madotto, A.; and Fung, P. 2023. Survey of Hallucination in Natural Language Generation. ACM Comput. find our card
  21. Surv., 55(12). Lewis, P.; Perez, E.; Piktus, A.; Petroni, F.; Karpukhin, V.; Goyal, N.; Küttler, H.; Lewis, M.; Yih, W.-t.; Rocktäschel, T.; Riedel, S.; and Kiela, D. 2020. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. In Proceedings of the 34th International Conference on Neur… find
  22. Li, D.; Rawat, A. S.; Zaheer, M.; Wang, X.; Lukasik, M.; Veit, A.; Yu, F.; and Kumar, S. 2023a. Large Language Models with Controllable Working Memory. In Findings of the Association for Computational Linguistics: ACL 2023, 1774–1793. Toronto, Canada: Association for Computational Linguistics. find our card
  23. Li, X.; Zhang, T.; Dubois, Y.; Taori, R.; Gulrajani, I.; Guestrin, C.; Liang, P.; and Hashimoto, T. B. 2023b. AlpacaEval: An Automatic Evaluator of Instruction-following Models. https://github.com/tatsu-lab/alpaca eval. ↗
  24. Li, X.; Zhu, X.; Ma, Z.; Liu, X.; and Shah, S. 2023c. Are ChatGPT and GPT-4 General-Purpose Solvers for Financial Text Analytics? An Examination on Several Typical Tasks. arXiv:2305.05862. — [arXiv:2305.05862](https://arxiv.org/abs/2305.05862) find
  25. Liu, N. F.; Zhang, T.; and Liang, P. 2023. Evaluating Verifiability in Generative Search Engines. arXiv:2304.09848. find our card
  26. Maynez, J.; Narayan, S.; Bohnet, B.; and McDonald, R. 2020. On Faithfulness and Factuality in Abstractive Summarization. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, 1906–1919. find
  27. OpenAI. 2022. Chatgpt: Optimizing language models for dialogue. https://openai.com/blog/chatgpt. ↗
  28. Peng, B.; Galley, M.; He, P.; Cheng, H.; Xie, Y.; Hu, Y.; Huang, Q.; Liden, L.; Yu, Z.; Chen, W.; and Gao, J. 2023. find
  29. Check Your Facts and Try Again: Improving Large Language Models with External Knowledge and Automated Feedback. arXiv:2302.12813. Qin, Y.; Liang, S.; Ye, Y.; Zhu, K.; Yan, L.; Lu, Y.; Lin, Y.; Cong, X.; Tang, X.; Qian, B.; Zhao, S.; Tian, R.; Xie, R.; Zhou, J.; Gerstein, M.; Li, D.; Liu, Z.; and Sun… find our card
  30. ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs. arXiv:2307.16789. find our card
  31. QwenLM. 2023. Qwen-7B. https://github.com/QwenLM/ ↗
  32. Raunak, V.; Menezes, A.; and Junczys-Dowmunt, M. 2021. find
  33. The Curious Case of Hallucinations in Neural Machine Translation. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 1172– 1183. Online: Association for Computational Linguistics. find our card
  34. Ren, R.; Wang, Y.; Qu, Y.; Zhao, W. X.; Liu, J.; Tian, H.; Wu, H.; Wen, J.-R.; and Wang, H. 2023. Investigating the Factual Knowledge Boundary of Large Language Models with Retrieval Augmentation. arXiv:2307.11019. find
  35. Shen, X.; Chen, Z.; Backes, M.; and Zhang, Y. 2023. In ChatGPT We Trust? Measuring and Characterizing the Reliability of ChatGPT. arXiv:2304.08979. — [arXiv:2304.08979](https://arxiv.org/abs/2304.08979) find
  36. Shi, W.; Min, S.; Yasunaga, M.; Seo, M.; James, R.; Lewis, M.; Zettlemoyer, L.; and tau Yih, W. 2023. RE-PLUG: Retrieval-Augmented Black-Box Language Models. arXiv:2301.12652. find our card
  37. THUDM. 2023a. ChatGLM-6B. https://github.com/ ↗
  38. THUDM. 2023b. ChatGLM2-6B. https://github.com/ ↗
  39. Trivedi, H.; Balasubramanian, N.; Khot, T.; and Sabharwal, A. 2023. Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 10014–10037. To… find our card
  40. Wang, A.; Pruksachatkun, Y.; Nangia, N.; Singh, A.; Michael, J.; Hill, F.; Levy, O.; and Bowman, S. R. 2019a. SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems. find our card
  41. Red Hook, NY, USA: Curran Associates Inc. Wang, A.; Singh, A.; Michael, J.; Hill, F.; Levy, O.; and Bowman, S. R. 2019b. GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding. In International Conference on Learning Representations. find our card
  42. Xu, G.; Liu, J.; Yan, M.; Xu, H.; Si, J.; Zhou, Z.; Yi, P.; Gao, X.; Sang, J.; Zhang, R.; Zhang, J.; Peng, C.; Huang, F.; and Zhou, J. 2023a. CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility. arXiv:2307.09705. find
  43. Xu, S.; Pang, L.; Shen, H.; Cheng, X.; and Chua, T.-S. 2023b. Search-in-the-Chain: Towards Accurate, Credible and Traceable Large Language Models for Knowledge-intensive Tasks. arXiv:2304.14732. — [arXiv:2304.14732](https://arxiv.org/abs/2304.14732) find
  44. Yunjie Ji, Y. G. Y. P. Q. N. B. M. X. L., Yong Deng. 2023. BELLE: Bloom-Enhanced Large Language model Engine. https://github.com/LianjiaTech/BELLE. ↗
  45. Zhang, W.; Aljunied, S. M.; Gao, C.; Chia, Y. K.; and Bing, L. 2023. M3Exam: A Multilingual, Multimodal, Multilevel Benchmark for Examining Large Language Models. find
  46. Zhong, W.; Cui, R.; Guo, Y.; Liang, Y.; Lu, S.; Wang, Y.; Saied, A.; Chen, W.; and Duan, N. 2023. AGIEval: A Human-Centric Benchmark for Evaluating Foundation Models. arXiv:2304.06364. Zhou, D.; Schärli, N.; Hou, L.; Wei, J.; Scales, N.; Wang, X.; Schuurmans, D.; Cui, C.; Bousquet, O.; Le, Q. V.; a… find our card
  47. Zhou, S.; Alon, U.; Xu, F. F.; Jiang, Z.; and Neubig, G. 2023b. DocPrompting: Generating Code by Retrieving the Docs. In The Eleventh International Conference on Learning Representations. find