Elamur Research⌕
☰
home / papers / 2024 / Improving Scientific Hypothesis Generation with Knowledge Grounded Large Language Models

Improving Scientific Hypothesis Generation with Knowledge Grounded Large Language Models

Paper · 2024 · Guangzhi Xiong, Eric Xie, Amir Hassan Shariatmadari et al.

Summary

LLMs can propose new research directions, but they often produce convincing yet false claims. In science, where each reasoning step must be checkable, this is a serious problem. The authors present KG-CoI (Knowledge Grounded Chain of Ideas), a system that anchors LLM hypothesis generation in structured knowledge graphs (KGs). It has three stages. First, it retrieves context with KG guidance: it collects multi-hop relation paths between the two entities in question and uses them to generate search keywords for literature retrieval. Second, the model reasons step by step, producing a "chain of ideas" informed by both the KG and the retrieved documents. Third, each reasoning step is checked against the KG to detect hallucinations. For evaluation, the authors built a new biomedical dataset by hiding known KG links and asking models to infer the missing relations. KG-CoI outperformed direct prompting, CoT, and RAG baselines on both open and closed LLMs, and it produced fewer hallucinated steps. The work points toward more trustworthy LLM support for scientific discovery.

Key points

  • Architecture:
    • Retrievers: a KG retriever (KG-R), which finds k-step relation chains by breadth-first search, and a literature retriever (Lit-R).
    • LLM agents: one for query enrichment (LLM-E), one for hypothesis generation (LLM-G), and one for claim verification (LLM-V).
  • Models tested: Llama-3.1-8B, Llama-3.1-70B, GPT-4o-mini, and GPT-4o. Full KG-CoI accuracy ranged from 70.33% (Llama-3.1-8B) to 86.00% (GPT-4o).
  • Ablations: Removing any component (KG information, literature, query enrichment, chain of thoughts) lowered performance for every model.
    • Dropping KG information caused the largest losses for Llama-3.1-70B and GPT-4o. For GPT-4o, accuracy fell from 86.00% to 74.67%.
  • Literature matters less for stronger models: Removing literature cost Llama-3.1-8B 9.66 points but GPT-4o only 2.67 points. The authors attribute this to stronger models having absorbed more literature during training.
  • Self-consistency: Using self-consistency instead of greedy decoding generally improved accuracy and F1. However, it did not raise the confidence of generated hypotheses.

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

 Institutions

🎓University of VirginiaCharlottesville · United States36 works · 9,730 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksComputer Vision and Pattern Recognition · 2,869 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Guangzhi Xiong, Eric Xie, Amir Hassan Shariatmadari, Sikun Guo, Stefan Bekiranov and Aidong Zhang. 2024. Improving Scientific Hypothesis Generation with Knowledge Grounded Large Language Models. arXiv:2411.02382.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2411.02382 · arXiv 2411.02382

@misc{xiong2024improving,
  title = {{Improving Scientific Hypothesis Generation with Knowledge Grounded Large Language Models}},
  author = {Guangzhi Xiong, Eric Xie, Amir Hassan Shariatmadari, Sikun Guo, Stefan Bekiranov and Aidong Zhang},
  year = {2024},
  doi = {10.48550/arxiv.2411.02382},
  eprint = {2411.02382},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2411.02382}
}

References

33 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. AI4Science, M. R.; and Quantum, M. A. 2023. The impact of large language models on scientific discovery: a preliminary study using gpt-4. arXiv preprint arXiv:2311.07361. — [arXiv:2311.07361](https://arxiv.org/abs/2311.07361) find
  2. Besta, M.; Blach, N.; Kubicek, A.; Gerstenberger, R.; Podstawski, M.; Gianinazzi, L.; Gajda, J.; Lehmann, T.; Niewiadomski, H.; Nyczyk, P.; et al. 2024. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume… find
  3. Borgeaud, S.; Mensch, A.; Hoffmann, J.; Cai, T.; Rutherford, E.; Millican, K.; Van Den Driessche, G. B.; Lespiau, J.-B.; Damoc, B.; Clark, A.; et al. 2022. Improving language models by retrieving from trillions of tokens. In International conference on machine learning, 2206–2240. PMLR. find
  4. Cobbe, K.; Kosaraju, V.; Bavarian, M.; Chen, M.; Jun, H.; Kaiser, L.; Plappert, M.; Tworek, J.; Hilton, J.; Nakano, R.; et al. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168.
  5. Delile, J.; Mukherjee, S.; Van Pamel, A.; and Zhukov, L. 2024. Graph-Based Retriever Captures the Long Tail of Biomedical Knowledge. arXiv preprint arXiv:2402.12352. — [arXiv:2402.12352](https://arxiv.org/abs/2402.12352) find
  6. Gao, Y.; Xiong, Y.; Gao, X.; Jia, K.; Pan, J.; Bi, Y.; Dai, Y.; Sun, J.; and Wang, H. 2023. Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997.
  7. Guu, K.; Lee, K.; Tung, Z.; Pasupat, P.; and Chang, M. 2020. Retrieval augmented language model pre-training. In International conference on machine learning, 3929–3938. PMLR.
  8. Hendrycks, D.; Burns, C.; Basart, S.; Zou, A.; Mazeika, M.; Song, D.; and Steinhardt, J. 2021. Measuring Massive Multitask Language Understanding. In 9th International Conference on Learning Representations, ICLR 2021, Virtual Event, Austria, May 3-7, 2021. OpenReview.net.
  9. Hou, W.; and Ji, Z. 2023. Assessing GPT-4 for cell type annotation in single-cell RNA-seq analysis. BioRxiv. find
  10. Huang, L.; Yu, W.; Ma, W.; Zhong, W.; Feng, Z.; Wang, H.; Chen, Q.; Peng, W.; Feng, X.; Qin, B.; et al. 2023. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. arXiv preprint arXiv:2311.05232.
  11. Izacard, G.; and Grave, E. 2020. Leveraging passage retrieval with generative models for open domain question answering. arXiv preprint arXiv:2007.01282.
  12. Lewis, P.; Perez, E.; Piktus, A.; Petroni, F.; Karpukhin, V.; Goyal, N.; Küttler, H.; Lewis, M.; Yih, W.-t.; Rocktäschel, T.; et al. 2020. Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in Neural Information Processing Systems, 33: 9459–9474.
  13. Liu, H.; Soroush, A.; Nestor, J. G.; Park, E.; Idnay, B.; Fang, Y.; Pan, J.; Liao, S.; Bernard, M.; Peng, Y.; et al. 2024. Retrieval augmented scientific claim verification. JAMIA open, 7(1): ooae021. find
  14. Liu, Y.; Yao, Y.; Ton, J.-F.; Zhang, X.; Cheng, R. G. H.; Klochkov, Y.; Taufiq, M. F.; and Li, H. 2023. Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models’ Alignment. arXiv preprint arXiv:2308.05374.
  15. Luo, L.; Li, Y.-F.; Haffari, G.; and Pan, S. 2023. Reasoning on graphs: Faithful and interpretable large language model reasoning. arXiv preprint arXiv:2310.01061. find
  16. Matsumoto, N.; Moran, J.; Choi, H.; Hernandez, M. E.; Venkatesan, M.; Wang, P.; and Moore, J. H. 2024. KRA-GEN: a knowledge graph-enhanced RAG framework for biomedical problem solving using large language models. Bioinformatics, 40(6). find
  17. Neumann, M.; King, D.; Beltagy, I.; and Ammar, W. 2019. ScispaCy: fast and robust models for biomedical natural language processing. arXiv preprint arXiv:1902.07669.
  18. Nori, H.; King, N.; McKinney, S. M.; Carignan, D.; and Horvitz, E. 2023. Capabilities of gpt-4 on medical challenge problems. arXiv preprint arXiv:2303.13375. OpenAI; Achiam, J.; Adler, S.; Agarwal, S.; Ahmad, L.; Akkaya, I.; Aleman, F. L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; Avi…
  19. Qi, B.; Zhang, K.; Li, H.; Tian, K.; Zeng, S.; Chen, Z.-R.; and Zhou, B. 2023. Large Language Models are Zero Shot Hypothesis Proposers. arXiv preprint arXiv:2311.05965. find
  20. Robertson, S.; Zaragoza, H.; et al. 2009. The probabilistic relevance framework: BM25 and beyond. Foundations and Trends® in Information Retrieval, 3(4): 333–389. find
  21. Sebastian, Y.; Siew, E.-G.; and Orimaye, S. O. 2017. Emerging approaches in literature-based discovery: techniques and performance review. The Knowledge Engineering Review, 32: e12. find
  22. Stribling, D.; Xia, Y.; Amer, M. K.; Graim, K. S.; Mulligan, C. J.; and Renne, R. 2024. The model student: GPT-4 performance on graduate biomedical science exams. Scientific Reports, 14(1): 5670. find
  23. Trivedi, H.; Balasubramanian, N.; Khot, T.; and Sabharwal, A. 2022. Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions. arXiv preprint arXiv:2212.10509.
  24. Wang, Q.; Downey, D.; Ji, H.; and Hope, T. 2023. Scimon: Scientific inspiration machines optimized for novelty. arXiv preprint arXiv:2305.14259. find
  25. Wang, X.; Wei, J.; Schuurmans, D.; Le, Q.; Chi, E.; Narang, S.; Chowdhery, A.; and Zhou, D. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.
  26. Wei, C.-H.; Allot, A.; Lai, P.-T.; Leaman, R.; Tian, S.; Luo, L.; Jin, Q.; Wang, Z.; Chen, Q.; and Lu, Z. 2024. Pub-Tator 3.0: an AI-powered literature resource for unlocking biomedical knowledge. Nucleic Acids Research, gkae235. find
  27. Wei, J.; Wang, X.; Schuurmans, D.; Bosma, M.; Xia, F.; Chi, E.; Le, Q. V.; Zhou, D.; et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35: 24824–24837.
  28. Wen, Y.; Wang, Z.; and Sun, J. 2023. Mindmap: Knowledge graph prompting sparks graph of thoughts in large language models. arXiv preprint arXiv:2308.09729.
  29. Xiong, G.; Jin, Q.; Lu, Z.; and Zhang, A. 2024. Benchmarking retrieval-augmented generation for medicine. In Findings of the Association for Computational Linguistics: ACL 2024.
  30. Yao, S.; Yu, D.; Zhao, J.; Shafran, I.; Griffiths, T.; Cao, Y.; and Narasimhan, K. 2024. Tree of thoughts: Deliberate problem solving with large language models. Advances in Neural Information Processing Systems, 36. find
  31. Zakka, C.; Shad, R.; Chaurasia, A.; Dalal, A. R.; Kim, J. L.; Moor, M.; Fong, R.; Phillips, C.; Alexander, K.; Ashley, E.; et al. 2024. Almanac—retrieval-augmented language models for clinical medicine. NEJM AI, 1(2): AIoa2300068. find
  32. Zhao, P.; Zhang, H.; Yu, Q.; Wang, Z.; Geng, Y.; Fu, F.; Yang, L.; Zhang, W.; and Cui, B. 2024. Retrieval-Augmented Generation for AI-Generated Content: A Survey. arXiv preprint arXiv:2402.19473.
  33. Zhou, Y.; Liu, H.; Srivastava, T.; Mei, H.; and Tan, C. 2024. Hypothesis Generation with Large Language Models. arXiv preprint arXiv:2404.04326. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/xiong-2024-improving-scientific-hypothesis-generation-knowledge/