Elamur Research
☰
home / papers / 2023 / Is ChatGPT a Good NLG Evaluator? A Preliminary Study

Is ChatGPT a Good NLG Evaluator? A Preliminary Study

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/2303.04048

Summary

Automatic metrics for natural language generation (NLG) often agree poorly with human judgments. This paper asks whether ChatGPT can do better if it is used as the judge. The authors treat ChatGPT as a stand-in for a human annotator. Each prompt names the task being judged (for example, news summarization) and the quality dimension (for example, fluency or relevance). ChatGPT then rates outputs either with a 0–100 direct assessment (DA) or with a 1–5 star scale, with or without a gold reference. They measure how well these ratings match human scores on five English meta-evaluation datasets covering summarization, story generation and data-to-text, and compare them with ROUGE, BERTScore, BARTScore and similar metrics. ChatGPT matches or beats these metrics in most settings. It does especially well on open-ended story generation, where many different outputs can be acceptable. However, its ratings change a lot depending on how the prompt is worded. It also loses its advantage on datasets whose human labels were built mainly by comparing outputs to references. The study is an early sign that a single LLM could serve as a general-purpose NLG evaluator.

Key points

  • The prompt templates are adapted from Kocmi and Federmann (2023). The DA prompt asks for a 0–100 score and the star prompt for one to five stars. Both come in reference-free and reference-based versions.
  • ChatGPT sets new state-of-the-art human correlation on SummEval and OpenMEVA, and is competitive on NewsRoom and BAGEL.
  • When a dataset's human labels lean heavily on reference overlap, simple lexical metrics like ROUGE already work well, and ChatGPT's advantage disappears.
  • Results swing with the prompt format. For example, story generation shows large gaps between DA with and without references.
  • Experiments were run through the ChatGPT web interface in February–March 2023, before the official API, with default temperature. Results are therefore hard to reproduce. The data is released at github.com/krystalan/chatgpt_as_nlg_evaluator.

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

268 papers of our library cite it (of the 140673 whose bibliographies we read); 14 of them have a page here, 254 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2303.04048 · arXiv 2303.04048

@misc{wang2023is,
  title = {{Is ChatGPT a Good NLG Evaluator? A Preliminary Study}},
  author = {Wang, Jiaan and Liang, Yunlong and Meng, Fandong and Sun, Zengkui and Shi, Haoxiang and Li, Zhixu and Xu, Jinan and Qu, Jianfeng and Zhou, Jie},
  year = {2023},
  doi = {10.48550/arxiv.2303.04048},
  eprint = {2303.04048},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2303.04048}
}

References

46 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Yejin Bang, Samuel Cahyawijaya, Nayeon Lee, Wenliang Dai, Dan Su, Bryan Wilie, Holy Lovenia, Ziwei Ji, Tiezheng Yu, Willy Chung, et al. 2023. A multi-task, multilingual, multimodal evaluation of chatgpt on reasoning, hallucination, and interactivity. arXiv preprint arXiv:2302.04023. find our card
  2. Manik Bhandari, Pranav Narayan Gour, Atabak Ashfaq, Pengfei Liu, and Graham Neubig. 2020. Re-evaluating evaluation in text summarization. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 9347–9359, Online. Association for Computational Linguist… find
  3. Hong Chen, Duc Vo, Hiroya Takamura, Yusuke Miyao, and Hideki Nakayama. 2022. StoryER: Automatic story evaluation via ranking, rating and reasoning. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 1739–1753, Abu Dhabi, United Arab Emirates. Association… find
  4. Michael Denkowski and Alon Lavie. 2011. Meteor 1.3: Automatic metric for reliable optimization and evaluation of machine translation systems. In Proceedings of the Sixth Workshop on Statistical Machine Translation, pages 85–91, Edinburgh, Scotland. Association for Computational Linguistics. find
  5. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,… find
  6. Bhuwan Dhingra, Manaal Faruqui, Ankur Parikh, Ming-Wei Chang, Dipanjan Das, and William Cohen. 2019. Handling divergent reference texts when evaluating table-to-text generation. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4884–4895, Florence, Ita… find our card
  7. Alexander R. Fabbri, Wojciech Kryściński, Bryan Mc-Cann, Caiming Xiong, Richard Socher, and Dragomir Radev. 2021. SummEval: Re-evaluating summarization evaluation. Transactions of the Association for Computational Linguistics, 9:391–409. find
  8. Jinlan Fu, See-Kiong Ng, Zhengbao Jiang, and Pengfei Liu. 2023. Gptscore: Evaluate as you desire. find
  9. Max Grusky, Mor Naaman, and Yoav Artzi. 2018. Newsroom: A dataset of 1.3 million summaries with diverse extractive strategies. In Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers), pa… find our card
  10. Jian Guan, Zhexin Zhang, Zhuoer Feng, Zitao Liu, Wenbiao Ding, Xiaoxi Mao, Changjie Fan, and Minlie Huang. 2021. OpenMEVA: A benchmark for evaluating open-ended story generation metrics. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th Internati… find
  11. Karl Moritz Hermann, Tomás Kociský, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom. 2015. Teaching machines to read and comprehend. In Advances in Neural Information Processing Systems 28: Annual Conference on Neural Information Processing Systems 2015, December 7-… find our card
  12. J. Edward Hu, Abhinav Singh, Nils Holzenberger, Matt Post, and Benjamin Van Durme. 2019. Large-scale, diverse, paraphrastic bitexts via sampling and clustering. In Proceedings of the 23rd Conference on Computational Natural Language Learning (CoNLL), pages 44–54, Hong Kong, China. Association for Co… find
  13. Fan Huang, Haewoon Kwak, and Jisun An. 2023. Is chatgpt better than human annotators? potential and limitations of chatgpt in explaining implicit hate speech. find
  14. Katharina Jeblick, Balthasar Schachtner, Jakob Dexl, Andreas Mittermeier, Anna Theresa Stüber, Johanna Topalis, Tobias Weber, Philipp Wesp, Bastian Sabel, Jens Ricke, and Michael Ingrisch. 2022. Chatgpt makes medicine easy to swallow: An exploratory case study on simplified radiology reports. find
  15. Wenxiang Jiao, Wenxuan Wang, Jen-tse Huang, Xing Wang, and Zhaopeng Tu. 2023. Is chatgpt a good translator? a preliminary study. arXiv preprint arXiv:2301.08745. find our card
  16. Maurice G Kendall. 1938. A new measure of rank correlation. Biometrika, 30(1/2):81–93. find
  17. Tom Kocmi and Christian Federmann. 2023. Large language models are state-of-the-art evaluators of translation quality. arXiv preprint arXiv:2302.14520. find
  18. Jan Kocoń, Igor Cichecki, Oliwier Kaszyca, Mateusz Kochanek, Dominika Szydło, Joanna Baran, Julita Bielaniewicz, Marcin Gruza, Arkadiusz Janz, Kamil Kanclerz, Anna Kocoń, Bartłomiej Koptyra, Wiktoria Mieleszczenko-Kowszewicz, Piotr Miłkowski, Marcin Oleksy, Maciej Piasecki, Łukasz Radliński, Konr… find
  19. Matt Kusner, Yu Sun, Nicholas Kolkin, and Kilian Weinberger. 2015. From word embeddings to document distances. In Proceedings of the 32nd International Conference on Machine Learning, volume 37 of Proceedings of Machine Learning Research, pages 957– 966, Lille, France. PMLR. find
  20. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In Proceedings of the 58th Annual Meeting of the… find our card
  21. Jiwei Li, Michel Galley, Chris Brockett, Jianfeng Gao, and Bill Dolan. 2016. A diversity-promoting objective function for neural conversation models. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pag… find our card
  22. Chin-Yew Lin. 2004. ROUGE: A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74–81, Barcelona, Spain. Association for Computational Linguistics. find
  23. François Mairesse, Milica Gašić, Filip Jurčíček, Simon Keizer, Blaise Thomson, Kai Yu, and Steve Young. 2010. Phrase-based statistical language generation using graphical models and active learning. In Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages… find
  24. Mavuto M Mukaka. 2012. A guide to appropriate use of correlation coefficient in medical research. Malawi medical journal, 24(3):69–71. find
  25. Ani Nenkova and Rebecca Passonneau. 2004. Evaluating content selection in summarization: The pyramid method. In Proceedings of the Human Language Technology Conference of the North American Chapter of the Association for Computational Linguistics: HLT-NAACL 2004, pages 145–152, Boston, Massachusetts… find
  26. OpenAI. 2022. Introducing chatgpt. openai.com/blog/chatgpt. https:// find
  27. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. arXiv preprint arXiv:2203.02155. find
  28. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: A method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting on Association for Computational Linguistics, ACL ’02, page 311–318, USA. Association for Computational Linguistics. find our card
  29. Krishna Pillutla, Lang Liu, John Thickstun, Sean Welleck, Swabha Swayamdipta, Rowan Zellers, Sewoong Oh, Yejin Choi, and Zaid Harchaoui. 2022. Mauve scores for generative models: Theory and practice. find
  30. Chengwei Qin, Aston Zhang, Zhuosheng Zhang, Jiaao Chen, Michihiro Yasunaga, and Diyi Yang. 2023. Is chatgpt a general-purpose natural language processing task solver? arXiv preprint arXiv:2302.06476. find our card
  31. Haocong Rao, Cyril Leung, and Chunyan Miao. 2023. Can chatgpt assess human personalities? a general evaluation framework. find
  32. Ricardo Rei, Craig Stewart, Ana C Farinha, and Alon Lavie. 2020. COMET: A neural framework for MT evaluation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 2685–2702, Online. Association for Computational Linguistics. find our card
  33. Thibault Sellam, Dipanjan Das, and Ankur Parikh. 2020. BLEURT: Learning robust metrics for text generation. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7881–7892, Online. Association for Computational Linguistics. find
  34. Brian Thompson and Matt Post. 2020. Automatic machine translation evaluation in many languages via zero-shot paraphrasing. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 90–121, Online. Association for Computational Linguistics. find our card
  35. Jiaan Wang, Yunlong Liang, Fandong Meng, Zhixu Li, Jianfeng Qu, and Jie Zhou. 2023a. Cross-lingual summarization via chatgpt. arXiv preprint. find
  36. Shuai Wang, Harrisen Scells, Bevan Koopman, and Guido Zuccon. 2023b. Can chatgpt write a good boolean query for systematic review literature search? find
  37. Xiang Wei, Xingyu Cui, Ning Cheng, Xiaobin Wang, Xin Zhang, Shen Huang, Pengjun Xie, Jinan Xu, Yufeng Chen, Meishan Zhang, Yong Jiang, and Wenjuan Han. 2023. Zero-shot information extraction via chatting with chatgpt. find
  38. Xianjun Yang, Yan Li, Xinlu Zhang, Haifeng Chen, and Wei Cheng. 2023. Exploring the limits of chatgpt for query or aspect-based text summarization. arXiv preprint arXiv:2302.08081. find
  39. Weizhe Yuan, Graham Neubig, and Pengfei Liu. 2021. Bartscore: Evaluating generated text as text generation. find our card
  40. Jerrold H Zar. 2005. Spearman rank correlation. Encyclopedia of biostatistics, 7. find
  41. Zhiyuan Zeng, Jiatong Yu, Tianyu Gao, Yu Meng, Tanya Goyal, and Danqi Chen. 2023. Evaluating large language models at evaluating instruction following. arXiv preprint arXiv:2310.07641. find
  42. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q. Weinberger, and Yoav Artzi. 2020. Bertscore: Evaluating text generation with BERT. In 8th International Conference on Learning Representations, ICLR 2020, Addis Ababa, Ethiopia, April 26-30, 2020. OpenReview.net. find
  43. Wei Zhao, Maxime Peyrard, Fei Liu, Yang Gao, Christian M. Meyer, and Steffen Eger. 2019. MoverScore: Text generation evaluating with contextualized embeddings and earth mover distance. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International… find our card
  44. Qihuang Zhong, Liang Ding, Juhua Liu, Bo Du, and Dacheng Tao. 2023. Can chatgpt understand too? a comparative study on chatgpt and fine-tuned bert. find
  45. Kaijie Zhu, Jindong Wang, Jiaheng Zhou, Zichen Wang, Hao Chen, Yidong Wang, Linyi Yang, Wei Ye, Neil Zhenqiang Gong, Yue Zhang, et al. 2023. Promptbench: Towards evaluating the robustness of large language models on adversarial prompts. arXiv preprint arXiv:2306.04528. find our card
  46. Guido Zuccon and Bevan Koopman. 2023. Dr chatgpt, tell me what i want to hear: How prompt knowledge impacts health answer correctness. find