Elamur Research⌕
☰
home / papers / 2023 / Evaluation Metrics in the Era of GPT-4: Reliably Evaluating Large Language Models on Sequence to Sequence Tasks

Evaluation Metrics in the Era of GPT-4: Reliably Evaluating Large Language Models on Sequence to Sequence Tasks

Paper · 2023 · Andrea Sottana, Bin Liang, Kai Zou et al.

Summary

The authors examine whether standard automatic metrics still judge modern generative LLMs fairly on sequence-to-sequence tasks. They tested open-source models (Flan-T5, T0pp, OPT-IML, Flan-UL2) and OpenAI models (GPT-3, InstructGPT, ChatGPT) zero-shot, with no fine-tuning or in-context examples. The tasks were summarisation on CNN / DailyMail, simplification on Newsela and grammatical error correction on the BEA-2019 development set. Outputs were scored three ways: with reference-based metrics, by three human annotators recruited via Prolific, and by GPT-4 acting as a reviewer.

The metrics and the humans disagreed. ChatGPT scored low on ROUGE and SARI, yet annotators usually preferred its outputs. Annotators also often rated the gold references below the strongest models, which suggests that weak benchmark data now limits evaluation. GPT-4's rankings matched human rankings fairly well for summarisation and simplification, and less well for GEC. The study argues against relying only on reference-based metrics. It also points to LLM-based judging as a partial, task-dependent alternative.

Key points

  • Data and metrics: 3,000 sampled items each from Newsela and CNN / DailyMail, plus all 4,384 BEA-2019 dev items. Scoring used SARI, the geometric mean of ROUGE-1/2/L F1, and ERRANT F0.5.
  • Automatic scores:
    • Summarisation: T0pp reached ROUGE 28.82 against 23.76 for ChatGPT.
    • Simplification: Flan-T5 reached SARI 44.98 against 37.55 for ChatGPT.
    • GEC: ChatGPT (39.54) and OPT-IML (39.05) were nearly tied.
  • Human evaluation: 100 samples per task were rated on Likert-style criteria taken from earlier work, with new Semantics and Grammaticality criteria added for GEC. Mean Krippendorff α was 0.88, with a minimum of 0.62.
  • Gold references: In summarisation, every annotator placed the gold reference last on all four criteria (Relevance, Fluency, Coherence, Consistency).
  • Limitations: Prompt tuning was limited, the sample of reviewers and items was small, test data may have leaked into pre-training, and GPT-4 might favour ChatGPT. Code is available at github.com/protagolabs/seq2seq_llm_evaluation.

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

🎓King's College LondonLondon · United Kingdom19 works · 1,454 citations

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 works

🧭 Topics 3Research subjects and languages the work is about.Language models & pretraining · 1,343 worksEvaluation & benchmarks · 685 worksSummarization · 159 works

Cite

Cite as Andrea Sottana, Bin Liang, Kai Zou and Zheng Yuan. 2023. Evaluation Metrics in the Era of GPT-4: Reliably Evaluating Large Language Models on Sequence to Sequence Tasks. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2023.emnlp-main.543

@inproceedings{sottana2023evaluation,
  title = {{Evaluation Metrics in the Era of GPT-4: Reliably Evaluating Large Language Models on Sequence to Sequence Tasks}},
  author = {Andrea Sottana, Bin Liang, Kai Zou and Zheng Yuan},
  booktitle = {Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing},
  year = {2023},
  doi = {10.18653/v1/2023.emnlp-main.543},
  url = {https://doi.org/10.18653/v1/2023.emnlp-main.543}
}

References

42 references, 22 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.
  2. Emily M Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pages 610–623. find
  3. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind The base pay was £12 with a 10% bonus if they completed the full 16-hour study. All 3 reviewers completed the study and were awarded this bonus. find
  4. 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.
  5. Christopher Bryant, Mariano Felice, Øistein E. Andersen, and Ted Briscoe. 2019. The BEA-2019 shared task on grammatical error correction. In Proceedings of the Fourteenth Workshop on Innovative Use of NLP for Building Educational Applications, pages 52–75, Florence, Italy. Association for Computatio…
  6. Christopher Bryant, Mariano Felice, and Ted Briscoe. 2017. Automatic annotation and evaluation of error types for grammatical error correction. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 793–805, Vancouver, Canada. Associ…
  7. Cheng-Han Chiang and Hung-yi Lee. 2023. Can large language models be an alternative to human evaluations? In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15607–15631, Toronto, Canada. Association for Computational Linguistics. find
  8. Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, Albert Webson, Shixiang Shane Gu, Zhuyun Dai, Mirac Suzgun, Xinyun Chen, Aakanksha Chowdhery, Sharan Narang, Gaurav Mishra, Adams Yu, Vincent Zhao, Yanping Huang, A…
  9. Steven Coyne, Keisuke Sakaguchi, Diana Galvan-Sosa, Michael Zock, and Kentaro Inui. 2023. Analyzing the performance of gpt-3.5 and gpt-4 in grammatical error correction. arXiv preprint arXiv:2303.14342. find
  10. Oscar M Cumbicus-Pineda, Itziar Gonzalez-Dios, and Aitor Soroa. 2021. Linguistic capabilities for a checklist-based evaluation in automatic text simplification. In CTTS@ SEPLN. find
  11. 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,…
  12. 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
  13. Tao Fang, Shu Yang, Kaixin Lan, Derek F Wong, Jinpeng Hu, Lidia S Chao, and Yue Zhang. 2023. Is chatgpt a highly fluent grammatical error correction system? a comprehensive evaluation. arXiv preprint arXiv:2304.01746. find
  14. Jinlan Fu, See-Kiong Ng, Zhengbao Jiang, and Pengfei Liu. 2023. Gptscore: Evaluate as you desire. arXiv preprint arXiv:2302.04166. find
  15. Fabrizio Gilardi, Meysam Alizadeh, and Maël Kubli. 2023. Chatgpt outperforms crowd-workers for text-annotation tasks. arXiv preprint arXiv:2303.15056.
  16. Natalia Grabar and Horacio Saggion. 2022. Evaluation of automatic text simplification: Where are we now, where should we go from here. In Actes de la 29e Conférence sur le Traitement Automatique des Langues Naturelles. Volume 1: conférence principale, pages 453–463. find
  17. Karl Moritz Hermann, Tomáš Kočiský, Edward Grefenstette, Lasse Espeholt, Will Kay, Mustafa Suleyman, and Phil Blunsom. 2015. Teaching machines to read and comprehend. In Proceedings of the 28th International Conference on Neural Information Processing Systems - Volume 1, NIPS’15, page 1693–1701, Ca…
  18. Neslihan Iskender, Tim Polzehl, and Sebastian Möller. 2021. Reliability of human evaluation for text summarization: Lessons learned and challenges ahead. In Proceedings of the Workshop on Human Evaluation of NLP Systems (HumEval), pages 86–96, Online. Association for Computational Linguistics. find
  19. Srinivasan Iyer, Xi Victoria Lin, Ramakanth Pasunuru, Todor Mihaylov, Dániel Simig, Ping Yu, Kurt Shuster, Tianlu Wang, Qing Liu, Punit Singh Koura, et al. 2022. Opt-iml: Scaling language model instruction meta learning through the lens of generalization.
  20. Chao Jiang, Mounica Maddela, Wuwei Lan, Yang Zhong, and Wei Xu. 2020. Neural CRF model for sentence alignment in text simplification. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 7943–7960, Online. Association for Computational Linguistics.
  21. Klaus Krippendorff. 2011. Computing krippendorff’s alpha-reliability. find
  22. Percy Liang, Rishi Bommasani, Tony Lee, Dimitris Tsipras, Dilara Soylu, Michihiro Yasunaga, Yian Zhang, Deepak Narayanan, Yuhuai Wu, Ananya Kumar, et al. 2022. Holistic evaluation of language models. arXiv preprint arXiv:2211.09110. find
  23. Rensis Likert. 1932. A technique for the measurement of attitudes. Archives of psychology, 140:5–55. find
  24. 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.
  25. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. Gpteval: Nlg evaluation using gpt-4 with better human alignment. arXiv preprint arXiv:2303.16634. find
  26. Seyed Mahed Mousavi, Gabriel Roccabruna, Michela Lorandi, Simone Caldarella, and Giuseppe Riccardi. 2022. Evaluation of response generation models: Shouldn’t it be shareable and replicable? In Proceedings of the 2nd Workshop on Natural Language Generation, Evaluation, and Metrics (GEM), pages 136–14… find
  27. Ramesh Nallapati, Bowen Zhou, Cicero dos Santos, Çağlar Gulçehre, and Bing Xiang. 2016. Abstractive text summarization using sequence-to-sequence RNNs and beyond. In Proceedings of the 20th SIGNLL Conference on Computational Natural Language Learning, pages 280–290, Berlin, Germany. Association for… find
  28. OpenAI. 2023. Gpt-4 technical report.
  29. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. 2022. Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35:27730–27744.
  30. Jiaxin Pei, Aparna Ananthasubramaniam, Xingyao Wang, Naitian Zhou, Apostolos Dedeloudis, Jackson Sargent, and David Jurgens. 2022. Potato: The portable text annotation tool. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing: System Demonstrations.
  31. Jason Phang, Yao Zhao, and Peter J Liu. 2022. Investigating efficiently extending transformers for long input summarization. arXiv preprint arXiv:2208.04347. find
  32. Victor Sanh, Albert Webson, Colin Raffel, Stephen H. Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Teven Le Scao, Arun Raja, Manan Dey, M Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Nihal Nayak, Debajyoti Da…
  33. Yi Tay, Mostafa Dehghani, Vinh Q Tran, Xavier Garcia, Dara Bahri, Tal Schuster, Huaixiu Steven Zheng, Neil Houlsby, and Donald Metzler. 2022. Unifying language learning paradigms. arXiv preprint arXiv:2205.05131. – Source: Ron Bee , a professor at San Diego State University , is worried that so few…
  34. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971.
  35. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems, 30.
  36. Original dataset, sample 2 – Source: Ron Bee , a professor at San Diego State University , is worried that so few Americans serve in the military . – Destination: Very few people join the military now . find
  37. Peiyi Wang, Lei Li, Liang Chen, Dawei Zhu, Binghuai Lin, Yunbo Cao, Qi Liu, Tianyu Liu, and Zhifang Sui. 2023. Large language models are not fair evaluators. arXiv preprint arXiv:2305.17926. find
  38. Our merged sample Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35:24824–24837.
  39. Haoran Wu, Wenxuan Wang, Yuxuan Wan, Wenxiang Jiao, and Michael Lyu. 2023. Chatgpt or grammarly? evaluating chatgpt on grammatical error correction benchmark. arXiv preprint arXiv:2303.13648. find
  40. Wei Xu, Chris Callison-Burch, and Courtney Napoles. 2015. Problems in current text simplification research: New data can help. Transactions of the Association for Computational Linguistics, 3:283–297.
  41. Wei Xu, Courtney Napoles, Ellie Pavlick, Quanze Chen, and Chris Callison-Burch. 2016. Optimizing statistical machine translation for text simplification. Transactions of the Association for Computational Linguistics, 4:401–415.
  42. Tianyi Zhang, Faisal Ladhak, Esin Durmus, Percy Liang, Kathleen McKeown, and Tatsunori B Hashimoto. 2023. Benchmarking large language models for news summarization. arXiv preprint arXiv:2301.13848.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/sottana-2023-evaluation-metrics-era-gpt-4/