Elamur Research⌕
☰
home / papers / 2024 / Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs

Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs

Paper · 2024 · Krista Opsahl-Ong, Michael J Ryan, Josh Purtell et al.

Summary

Pipelines that chain several language model calls (LM programs) need prompts that work well together across all stages. In practice, though, only the final output can usually be scored. There are no labels for intermediate steps, and the authors assume no access to gradients, weights or log-probabilities. The paper formalizes this setting and reduces it to tuning two things per module: a free-form instruction and a set of few-shot demonstrations. It addresses two obstacles. The first is generating a few strong candidate prompts from an enormous space. The second is credit assignment, i.e. working out which module-level choices drive overall performance. The proposed techniques are:

  • demonstrations bootstrapped from successful program runs;
  • instruction proposals grounded in summaries of the data and the program;
  • a Bayesian surrogate model fitted on noisy mini-batch evaluations;
  • meta-learning of how the proposer LM should be configured.

Combined, these form MIPRO. Using Llama3-8B, it beats the baseline optimizers on five of the seven tasks in a newly released benchmark, with accuracy gains reaching 13%. This matters because it automates the manual prompt engineering that multi-stage systems currently require.

Key points

  • MIPRO (Multi-prompt Instruction PRoposal Optimizer) jointly tunes instructions and demonstrations. It is best on 5 of 7 benchmark programs, with gains of up to 13% accuracy, using Llama3-8B.
  • Three proposal strategies are studied: bootstrapping demonstrations, grounding the proposer, and learning to propose. Three credit-assignment strategies are compared: greedy, surrogate and history-based.
  • Bootstrapped few-shot demonstrations often give the largest gains. Instruction optimization becomes crucial for tasks governed by conditional rules, and optimizing both together usually works best.
  • Results are averaged over 5 runs, with significance checked via Wilcoxon signed-rank tests (p < .05). Comparisons on ScoNe, HotPotQA and HoVer show mixed results among the instruction-only variants (Module-Level OPRO, 0-Shot MIPRO, 0-Shot MIPRO++).
  • The optimizers and benchmark are released in DSPy (dspy.ai). Stated limitations include a fixed budget, fixed proposer and task LMs, and difficulty inferring complex task rules without a handwritten seed prompt.

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

🎓Stanford UniversityStanford · United States310 works · 284,480 citations🎓University of California, BerkeleyBerkeley · United States127 works · 212,523 citations🎓KTH Royal Institute of TechnologyStockholm · Sweden7 works · 605 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 1Research subjects and languages the work is about.Language models & pretraining · 1,343 works

Cite

Cite as Krista Opsahl-Ong, Michael J Ryan, Josh Purtell, David Broman, Christopher Potts, Matei Zaharia and Omar Khattab. 2024. Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs. Proceedings of the 2024 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/v1/2024.emnlp-main.525

@inproceedings{opsahlong2024optimizing,
  title = {{Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs}},
  author = {Krista Opsahl-Ong, Michael J Ryan, Josh Purtell, David Broman, Christopher Potts, Matei Zaharia and Omar Khattab},
  booktitle = {Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing},
  year = {2024},
  doi = {10.18653/v1/2024.emnlp-main.525},
  url = {https://doi.org/10.18653/v1/2024.emnlp-main.525}
}

References

43 references, 19 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. AI@Meta. 2024. Llama 3 model card. find
  2. Takuya Akiba, Shotaro Sano, Toshihiko Yanase, Takeru Ohta, and Masanori Koyama. 2019. Optuna: A next-generation hyperparameter optimization framework. In Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, pages 2623–2631.
  3. James Bergstra, Rémi Bardenet, Yoshua Bengio, and Balázs Kégl. 2011. Algorithms for hyper-parameter optimization. Advances in neural information processing systems, 24. find
  4. Luca Beurer-Kellner, Marc Fischer, and Martin Vechev. 2023. Prompting is programming: A query language for large language models. Proceedings of the ACM on Programming Languages, 7(PLDI):1946–1969. find
  5. Wenhu Chen, Xueguang Ma, Xinyi Wang, and William W Cohen. 2022. Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks. arXiv preprint arXiv:2211.12588. find
  6. Antonia Creswell and Murray Shanahan. 2022. Faithful reasoning using large language models. Preprint, arXiv:2208.14271. find
  7. Mingkai Deng, Jianyu Wang, Cheng-Ping Hsieh, Yihan Wang, Han Guo, Tianmin Shu, Meng Song, Eric Xing, and Zhiting Hu. 2022. RLPrompt: Optimizing discrete text prompts with reinforcement learning. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3369–339…
  8. Robert C. Detrano, András Jánosi, Walter Steinbrunn, Matthias Emil Pfisterer, Johann-Jakob Schmid, Sar-bjit Sandhu, Kern Guppy, Stella Lee, and Victor Froelicher. 1989. International application of a new probability algorithm for the diagnosis of coronary artery disease. The American journal of card… find
  9. David Dohan, Winnie Xu, Aitor Lewkowycz, Jacob Austin, David Bieber, Raphael Gontijo Lopes, Yuhuai Wu, Henryk Michalewski, Rif A Saurous, Jascha Sohl-Dickstein, et al. 2022. Language model cascades. arXiv preprint arXiv:2207.10342. find
  10. Stefan Falkner, Aaron Klein, and Frank Hutter. 2018. BOHB: Robust and efficient hyperparameter optimization at scale. In Proceedings of the 35th International Conference on Machine Learning, volume 80 of Proceedings of Machine Learning Research, pages 1437–1446. PMLR. find
  11. Chrisantha Fernando, Dylan Banarse, Henryk Michalewski, Simon Osindero, and Tim Rocktäschel. 2023. Promptbreeder: Self-referential self-improvement via prompt evolution. arXiv preprint arXiv:2309.16797.
  12. Rory A. Fisher. 1936. The use of multiple measurements in taxonomic problems. Annals of Human Genetics, 7:179–188. find
  13. Tianyu Gao, Adam Fisch, and Danqi Chen. 2021. Making pre-trained language models better few-shot learners. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pa…
  14. Qingyan Guo, Rui Wang, Junliang Guo, Bei Li, Kaitao Song, Xu Tan, Guoqing Liu, Jiang Bian, and Yujiu Yang. 2024. Connecting large language models with evolutionary algorithms yields powerful prompt optimizers. Preprint, arXiv:2309.08532. find
  15. Yaru Hao, Zewen Chi, Li Dong, and Furu Wei. 2022. Optimizing prompts for text-to-image generation. arXiv preprint arXiv:2212.09611. — [arXiv:2212.09611](https://arxiv.org/abs/2212.09611) find
  16. Yichen Jiang, Shikha Bordia, Zheng Zhong, Charles Dognin, Maneesh Singh, and Mohit Bansal. 2020. HoVer: A dataset for many-hop fact extraction and claim verification. In Findings of the Association for Computational Linguistics: EMNLP 2020, pages 3441–3460, Online. Association for Computational Ling…
  17. Omar Khattab, Christopher Potts, and Matei Zaharia. 2021. Baleen: Robust Multi-Hop Reasoning at Scale via Condensed Retrieval. In Thirty-Fifth Conference on Neural Information Processing Systems. find
  18. Omar Khattab, Keshav Santhanam, Xiang Lisa Li, David Hall, Percy Liang, Christopher Potts, and Matei Zaharia. 2022. Demonstrate-search-predict: Composing retrieval and language models for knowledge-intensive nlp. arXiv preprint arXiv:2212.14024.
  19. Omar Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav Santhanam, Sri Vardhamanan A, Saiful Haq, Ashutosh Sharma, Thomas T. Joshi, Hanna Moazam, Heather Miller, Matei Zaharia, and Christopher Potts. 2024. DSPy: Compiling declarative language model calls into state-of-the-art pipeline… find
  20. Xiang Lisa Li and Percy Liang. 2021. Prefix-tuning: Optimizing continuous prompts for generation. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 4582–… find
  21. Mohammadreza Pourreza and Davood Rafiei. 2023. Din-sql: Decomposed in-context learning of text-to-sql with self-correction. arXiv preprint arXiv:2304.11015.
  22. Reid Pryzant, Dan Iter, Jerry Li, Yin Tat Lee, Chenguang Zhu, and Michael Zeng. 2023. Automatic prompt optimization with" gradient descent" and beam search. arXiv preprint arXiv:2305.03495.
  23. Yujia Qin, Shihao Liang, Yining Ye, Kunlun Zhu, Lan Yan, Yaxi Lu, Yankai Lin, Xin Cong, Xiangru Tang, Bill Qian, Sihan Zhao, Lauren Hong, Runchu Tian, Ruobing Xie, Jie Zhou, Mark Gerstein, Dahai Li, Zhiyuan Liu, and Maosong Sun. 2023. Toolllm: Facilitating large language models to master 16000+ real…
  24. Tal Ridnik, Dedy Kredo, and Itamar Friedman. 2024. Code generation with alphacodium: From prompt engineering to flow engineering. arXiv preprint arXiv:2401.08500. find
  25. Imanol Schlag, Sainbayar Sukhbaatar, Asli Celikyilmaz, Wen-tau Yih, Jason Weston, Jürgen Schmidhuber, and Xian Li. 2023. Large language model programs. arXiv preprint arXiv:2305.05364. find
  26. Jingyuan S. She, Christopher Potts, Samuel R. Bowman, and Atticus Geiger. 2023. ScoNe: Benchmarking negation reasoning in language models with fine-tuning and in-context learning. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pag… find
  27. Taylor Shin, Yasaman Razeghi, Robert L. Logan IV, Eric Wallace, and Sameer Singh. 2020. AutoPrompt: Eliciting Knowledge from Language Models with Automatically Generated Prompts. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 4222–4235, Onlin…
  28. Jasper Snoek, Hugo Larochelle, and Ryan P. Adams. 2012. Practical bayesian optimization of machine learning algorithms. Preprint, arXiv:1206.2944. — [arXiv:1206.2944](https://arxiv.org/abs/1206.2944) find
  29. Alessandro Sordoni, Xingdi Yuan, Marc-Alexandre Côté, Matheus Pereira, Adam Trischler, Ziang Xiao, Arian Hosseini, Friederike Niedtner, and Nicolas Le Roux. 2023. Joint prompt optimization of stacked llms using variational inference. Preprint, arXiv:2306.12509. find
  30. Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. 2023. Lost in the middle: How language models use long contexts. arXiv preprint arXiv:2307.03172.
  31. Dilara Soylu, Christopher Potts, and Omar Khattab. 2024. Fine-tuning and prompt optimization: Two great steps that work better together. Preprint, arXiv:2407.10930. find
  32. Jiayi Pan, Yichi Zhang, Nicholas Tomlin, Yifei Zhou, Sergey Levine, and Alane Suhr. 2024. Autonomous evaluation and refinement of digital agents. Preprint, arXiv:2404.06474. find
  33. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. 2022. Chain of thought prompting elicits reasoning in large language models. arXiv preprint arXiv:2201.11903.
  34. Yuxin Wen, Neel Jain, John Kirchenbauer, Micah Goldblum, Jonas Geiping, and Tom Goldstein. 2023. Hard prompts made easy: Gradient-based discrete optimization for prompt tuning and discovery. arXiv preprint arXiv:2302.03668. find
  35. Tongshuang Wu, Michael Terry, and Carrie Jun Cai. 2022. Ai chains: Transparent and controllable human-ai interaction by chaining large language model prompts. In Proceedings of the 2022 CHI conference on human factors in computing systems, pages 1–22. find
  36. Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V Le, Denny Zhou, and Xinyun Chen. 2023. Large language models as optimizers. arXiv preprint arXiv:2309.03409.
  37. Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V. Le, Denny Zhou, and Xinyun Chen. 2024. Large language models as optimizers. Preprint, arXiv:2309.03409.
  38. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2369…
  39. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. React: Synergizing reasoning and acting in language models. Preprint, arXiv:2210.03629.
  40. Tianjun Zhang, Xuezhi Wang, Denny Zhou, Dale Schuurmans, and Joseph E Gonzalez. 2022. Tempera: Test-time prompt editing via reinforcement learning. In The Eleventh International Conference on Learning Representations. find
  41. Lianmin Zheng, Liangsheng Yin, Zhiqiang Xie, Chuyue Sun, Jeff Huang, Cody Hao Yu, Shiyi Cao, Christos Kozyrakis, Ion Stoica, Joseph E. Gonzalez, Clark Barrett, and Ying Sheng. 2024. Sglang: Efficient execution of structured language model programs. Preprint, arXiv:2312.07104.
  42. Yongchao Zhou, Andrei Ioan Muresanu, Ziwen Han, Keiran Paster, Silviu Pitis, Harris Chan, and Jimmy Ba. 2022. Large language models are human-level prompt engineers. arXiv preprint arXiv:2211.01910.
  43. Yongchao Zhou, Andrei Ioan Muresanu, Ziwen Han, Keiran Paster, Silviu Pitis, Harris Chan, and Jimmy Ba. 2023. Large language models are human-level prompt engineers. Preprint, arXiv:2211.01910.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ong-2024-optimizing-instructions-demonstrations-multi-stage/