Elamur Research⌕
☰
home / papers / 2023 / Exploring the Relationship between LLM Hallucinations and Prompt Linguistic Nuances: Readability, Formality, and Concreteness

Exploring the Relationship between LLM Hallucinations and Prompt Linguistic Nuances: Readability, Formality, and Concreteness

Summary

The paper looks at causes of hallucination in large language models rather than fixes. It asks whether three stylistic properties of a prompt change how often a model invents facts: readability, formality, and concreteness. The authors took 2,500 New York Times news-event tweets as factually grounded prompts and gave them to 15 LLMs, ranging from GPT-2, T5 and XLNet to GPT-4. Workers on Amazon Mechanical Turk labeled the outputs sentence by sentence, using four hallucination types: Person, Location, Number, and Acronym.

Each prompt was scored with three measures:

  • the Flesch Reading Ease Score,
  • the Heylighen–Dewaele formality formula, based on part-of-speech frequencies,
  • the mean word concreteness from the Brysbaert et al. (2014) ratings.

The scores were then grouped into low, mid and high bands. More formal and more concrete prompts were linked to fewer hallucinations. The readability results were mixed. The study is explicitly preliminary, and its results are mostly qualitative. Its value is in pointing to prompt wording as a practical, low-cost way to reduce hallucination.

Key points

  • The annotated dataset has 36,910 hallucinated sentences: Person 14,850, Location 13,050, Number 7,275 and Acronym 1,225.
  • The 15 models tested were GPT-4, GPT-3.5, GPT-3, GPT-2, MPT, OPT, LLaMA, BLOOM, Alpaca, Vicuna, Dolly, StableLM, XLNet, T5 and T0.
  • Score bands:
    • Readability and formality: low 0–30, mid 31–70, high 71–100.
    • Concreteness: 1–2.5, 2.5–3.5, 3.5–5.
  • Main associations:
    • Higher formality goes with fewer Person and Location errors.
    • Higher concreteness goes with fewer Number and Acronym errors.
    • Both effects are most visible in stronger models such as GPT-4 and OPT.
  • Easier-to-read prompts generally hallucinated less, but some hard-to-read formal prompts also did well. The authors therefore call the readability findings inconclusive, and no effect sizes are reported.

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: Machine translation · AI & machine learning

Cited by

14 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 12 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2309.11064 · arXiv 2309.11064

@misc{rawte2023exploring,
  title = {{Exploring the Relationship between LLM Hallucinations and Prompt Linguistic Nuances: Readability, Formality, and Concreteness}},
  author = {Rawte, Vipula and Priya, Prachi and Tonmoy, S.M Towhidul Islam and Zaman, S M Mehedi and Sheth, Amit and Das, Amitava},
  year = {2023},
  doi = {10.48550/arxiv.2309.11064},
  eprint = {2309.11064},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2309.11064}
}

References

31 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Amazon. Amazon mechanical turk. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, *Figure 3: Hallucination vs Readability* *Figure 4: Hallucination vs Formality* Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901. find
  2. Marc Brysbaert, Amy Beth Warriner, and Victor Kuperman. 2014. Concreteness ratings for 40 thousand generally known english word lemmas. Behavior research methods, 46:904–911. ↗
  3. Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E. Gonzalez, Ion Stoica, and Eric P. Xing. 2023. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality. *Figure 5: Hallucination vs Concreteness* databricks. 2023. Dolly. find
  4. Tristan Deleu, David Kanaa, Leo Feng, Giancarlo Kerg, Yoshua Bengio, Guillaume Lajoie, and Pierre-Luc Bacon. 2022. Continuous-time meta-learning with forward mode differentiation. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenReview.net. ↗
  5. R Flesch. 1948. A new readability yardstick journal of applied psychology 32: 221–233. find
  6. Francis Heylighen and Jean-Marc Dewaele. 1999. Formality of language: definition, measurement and behavioral determinants. Interner Bericht, Center “Leo Apostel”, Vrije Universiteit Brüssel, 4(1). find
  7. Xingxuan Li, Ruochen Zhao, Yew Ken Chia, Bosheng Ding, Lidong Bing, Shafiq Joty, and Soujanya Poria. 2023. Chain of knowledge: A framework for grounding large language models with structured knowledge bases. arXiv preprint arXiv:2305.13269. ↗
  8. Jiawei Liu, Chunqiu Steven Xia, Yuyao Wang, and Lingming Zhang. 2023a. Is your code generated by chatgpt really correct? rigorous evaluation of large language models for code generation. arXiv preprint arXiv:2305.01210. ↗ our card
  9. Nelson F Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2023b. Lost in the middle: How language models use long contexts. arXiv preprint arXiv:2307.03172. ↗ our card
  10. Eduard Hovy. 1987. Generating natural language under pragmatic constraints. Journal of Pragmatics, 11(6):689–719. ↗
  11. Midjourney. 2022. https://www.midjourney.com. ↗
  12. Faisal Ladhak, Esin Durmus, Mirac Suzgun, Tianyi Zhang, Dan Jurafsky, Kathleen Mckeown, and Tatsunori B Hashimoto. 2023. When do pre-training biases propagate to downstream tasks? a case study in text summarization. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 3198–3211. NYT. https://www.nytimes.com/topic/company/twitter. ↗
  13. Niels Mündler, Jingxuan He, Slobodan Jenko, and Martin Vechev. 2023. Self-contradictory hallucinations of large language models: Evaluation, detection and mitigation. ↗
  14. OpenAI. 2022. Introducing chatgpt. find
  15. OpenAI. 2023. Gpt-4 technical report. find our card
  16. Allan Paivio. 2013. Dual coding theory, word abstractness, and emotion: a critical review of kousta et al.(2011). ↗
  17. Ellie Pavlick and Joel Tetreault. 2016. An empirical analysis of formality in online communication. Transactions of the Association for Computational Linguistics, 4:61–74. ↗ our card
  18. Baolin Peng, Michel Galley, Pengcheng He, Hao Cheng, Yujia Xie, Yu Hu, Qiuyuan Huang, Lars Liden, Zhou Yu, Weizhu Chen, et al. 2023. Check your facts and try again: Improving large language models with external knowledge and automated feedback. arXiv preprint arXiv:2302.12813. ↗ our card
  19. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. 2023. Stanford alpaca: An instruction-following llama model. https://github.com/tatsu-lab/ stanford_alpaca. ↗
  20. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971. ↗
  21. Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever, et al. 2019. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9. find
  22. Neeraj Varshney, Wenlin Yao, Hongming Zhang, Jianshu Chen, and Dong Yu. 2023. A stitch in time saves nine: Detecting and mitigating hallucinations of llms by validating low-confidence generation. arXiv preprint arXiv:2307.03987. ↗ our card
  23. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. The Journal of Machine Learning Research, 21(1):5485–5551. ↗
  24. Zhen Wang, Rameswar Panda, Leonid Karlinsky, Rogerio Feris, Huan Sun, and Yoon Kim. 2023. Multitask prompt tuning enables parameter-efficient transfer learning. In The Eleventh International Conference on Learning Representations. ↗
  25. Aditya Ramesh, Prafulla Dhariwal, Alex Nichol, Casey Chu, and Mark Chen. 2022. Hierarchical text-conditional image generation with clip latents. arXiv preprint arXiv:2204.06125. ↗ our card
  26. Zhilin Yang, Zihang Dai, Yiming Yang, Jaime Carbonell, Russ R Salakhutdinov, and Quoc V Le. 2019. Xlnet: Generalized autoregressive pre-training for language understanding. Advances in neural information processing systems, 32. find
  27. Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Radford, Mark Chen, and Ilya Sutskever. 2021. Zero-shot text-to-image generation. In International Conference on Machine Learning, pages 8821–8831. PMLR. ↗ our card
  28. Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10684–10695. ↗
  29. Teven Le Scao, Angela Fan, Christopher Akiki, Ellie Pavlick, Suzana Ilić, Daniel Hesslow, Roman Castagné, Alexandra Sasha Luccioni, François Yvon, Matthias Gallé, et al. 2022. Bloom: A 176b-parameter open-access multilingual language model. arXiv preprint arXiv:2211.05100. ↗
  30. Shuo Zhang, Liangming Pan, Junzhou Zhao, and William Yang Wang. 2023. Mitigating language model hallucination with interactive question-knowledge alignment. find
  31. Susan Zhang, Stephen Roller, Naman Goyal, Mikel Artetxe, Moya Chen, Shuohui Chen, Christopher Dewan, Mona Diab, Xian Li, Xi Victoria Lin, Todor Mihaylov, Myle Ott, Sam Shleifer, Kurt Shuster, Daniel Simig, Punit Singh Koura, Anjali Sridhar, Tianlu Wang, and Luke Zettlemoyer. 2022. Opt: Open pre-trained transformer language models. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/rawte-2023-exploring-relationship-between-llm-hallucinations/