Elamur Research
☰
home / papers / 2023 / Instruction-Following Evaluation for Large Language Models

Instruction-Following Evaluation for Large Language Models

Written for the Research library (claude-opus-5-5) from the paper text. Original — arxiv.org/pdf/2311.07911

Summary

Judging whether an LLM actually obeys instructions is hard to standardize. Human raters are costly, slow and inconsistent, and an LLM acting as judge brings its own biases and blind spots. The authors propose IFEval, a benchmark limited to "verifiable instructions": constraints that a short deterministic program can check, such as a word-count range, a ban on commas, all-lowercase output or JSON formatting. They define 25 such instruction types and build 541 prompts, each pairing a task with one or more constraints. To create the prompts, they attached randomly chosen instructions to base prompts and used few-shot prompting to filter out contradictory combinations and diversify the wording. They then hand-edited every item. The checkers accept common surface variations, such as markdown bold around a required phrase, to avoid false negatives. GPT-4 and PaLM 2 serve as baselines. The work matters because it offers a cheap, fully reproducible and interpretable test that also shows which constraint types models tend to violate.

Key points

  • 25 instruction types are grouped into categories: Keywords, Language, Length Constraints, Detectable Content, Detectable Format, Combination, Change Cases, Start with / End with and Punctuation. Each type has several parameter values and phrasings.
  • The 541 prompts were built in four steps: (1) attach 1–3 random instructions to a base prompt, (2) filter illogical prompts via few-shot prompting, (3) rephrase via few-shot prompting, and (4) review each prompt manually.
  • Results are reported for GPT-4 and PaLM 2 as baselines. The numeric scores were in the portion of the text omitted from the provided excerpt, so they are not reproduced here.
  • The Response Language instruction and prompts such as a short article on Ukrainian morphology make the benchmark partially relevant to multilingual and Ukrainian NLP work.
  • Code and data are released in the google-research GitHub repository under instruction_following_eval.

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

550 papers of our library cite it (of the 140673 whose bibliographies we read); 24 of them have a page here, 526 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2311.07911 · arXiv 2311.07911

@misc{zhou2023instruction,
  title = {{Instruction-Following Evaluation for Large Language Models}},
  author = {Zhou, Jeffrey and Lu, Tianjian and Mishra, Swaroop and Brahma, Siddhartha and Basu, Sujoy and Luan, Yi and Zhou, Denny and Hou, Le},
  year = {2023},
  doi = {10.48550/arxiv.2311.07911},
  eprint = {2311.07911},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2311.07911}
}

References

27 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Rohan Anil, Andrew M Dai, Orhan Firat, Melvin Johnson, Dmitry Lepikhin, Alexandre Passos, Siamak Shakeri, Emanuel Taropa, Paige Bailey, Zhifeng Chen, et al. Palm 2 technical report. arXiv preprint arXiv:2305.10403, 2023. find our card
  2. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901, 2020. find
  3. Yupeng Chang, Xu Wang, Jindong Wang, Yuan Wu, Kaijie Zhu, Hao Chen, Linyi Yang, Xiaoyuan Yi, Cunxiang Wang, Yidong Wang, et al. A survey on evaluation of large language models. arXiv preprint arXiv:2307.03109, 2023. find our card
  4. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, et al. Evaluating large language models trained on code. arXiv preprint arXiv:2107.03374, 2021. find our card
  5. Cheng-Han Chiang and Hung-yi Lee. Can large language models be an alternative to human evaluations? arXiv preprint arXiv:2305.01937, 2023. find
  6. Aakanksha Chowdhery, Sharan Narang, Jacob Devlin, Maarten Bosma, Gaurav Mishra, Adam Roberts, Paul Barham, Hyung Won Chung, Charles Sutton, Sebastian Gehrmann, et al. PaLM: Scaling language modeling with pathways. arXiv preprint arXiv:2204.02311, 2022. find
  7. Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al. Scaling instruction-finetuned language models. arXiv preprint arXiv:2210.11416, 2022. find
  8. Jinlan Fu, See-Kiong Ng, Zhengbao Jiang, and Pengfei Liu. Gptscore: Evaluate as you desire. arXiv preprint arXiv:2302.04166, 2023. find
  9. Daniel Martin Katz, Michael James Bommarito, Shang Gao, and Pablo Arredondo. Gpt-4 passes the bar exam. Available at SSRN 4389233, 2023. find
  10. Anis Koubaa. Gpt-4 vs. gpt-3.5: A concise showdown. 2023. find
  11. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. Gpteval: Nlg evaluation using gpt-4 with better human alignment. arXiv preprint arXiv:2303.16634, 2023. find
  12. Swaroop Mishra, Daniel Khashabi, Chitta Baral, and Hannaneh Hajishirzi. Cross-task generalization via natural language crowdsourcing instructions. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 3470–3487, 2022. find
  13. Ben Naismith, Phoebe Mulcaire, and Jill Burstein. Automated evaluation of written discourse coherence using gpt-4. In Proceedings of the 18th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2023), pp. 394–403, 2023. find
  14. OpenAI. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023. find our card
  15. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35:27730–27744, 2022. find
  16. Baolin Peng, Chunyuan Li, Pengcheng He, Michel Galley, and Jianfeng Gao. Instruction tuning with gpt-4. arXiv preprint arXiv:2304.03277, 2023. find our card
  17. Chenhui Shen, Liying Cheng, Yang You, and Lidong Bing. Are large language models good evaluators for abstractive summarization? arXiv preprint arXiv:2305.13091, 2023. find
  18. Ondrej Skopek, Rahul Aralikatte, Sian Gooding, and Victor Carbune. Towards better evaluation of instruction-following: A case-study in summarization. arXiv preprint arXiv:2310.08394, 2023. find
  19. Jiao Sun, Yufei Tian, Wangchunshu Zhou, Nan Xu, Qian Hu, Rahul Gupta, John Frederick Wieting, Nanyun Peng, and Xuezhe Ma. Evaluating large language models on controlled generation tasks. arXiv preprint arXiv:2310.14542, 2023. find
  20. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. Stanford alpaca: An instruction-following llama model. https://github. com/tatsu-lab/stanford\_alpaca, 2023. find
  21. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023. find
  22. Sanh Victor, Webson Albert, Raffel Colin, Bach Stephen, Sutawika Lintang, Alyafeai Zaid, Chaffin Antoine, Stiegler Arnaud, Raja Arun, Dey Manan, et al. Multitask prompted training enables zero-shot task generalization. In ICLR, 2022. find our card
  23. Peiyi Wang, Lei Li, Liang Chen, Dawei Zhu, Binghuai Lin, Yunbo Cao, Qi Liu, Tianyu Liu, and Zhifang Sui. Large language models are not fair evaluators. arXiv preprint arXiv:2305.17926, 2023. find
  24. Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. Finetuned language models are zero-shot learners. In ICLR, 2022. find
  25. Ning Wu, Ming Gong, Linjun Shou, Shining Liang, and Daxin Jiang. Large language models are diverse role-players for summarization evaluation. arXiv preprint arXiv:2303.15078, 2023. find
  26. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric. P Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. Judging llm-as-a-judge with mt-bench and chatbot arena, 2023. find
  27. Ruiqi Zhong, Kristy Lee, Zheng Zhang, and Dan Klein. Adapting language models for zero-shot learning by meta-tuning on dataset and prompt collections. In Findings of the Association for Computational Linguistics: EMNLP 2021, pp. 2856–2878, 2021. *Figure 3: Instruction following accuracy per detailed… find