Elamur Research⌕
☰
home / papers / 2023 / Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models

Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models

Paper · 2023 · Jiaao Chen, Xiaoman Pan, Dian Yu et al.

Summary

Large language models often fail at compositional generalization. They may master basic operations yet stumble on harder, unseen problems that need those operations combined in new ways. Chen et al. (Tencent AI Lab) tackle this through in-context learning and propose Skills-in-Context (SKiC), a single-stage prompt with three parts:

  • a short list of foundational skills with instructions, written by hand or drafted by an LLM;
  • a few worked exemplars in which every reasoning step is explicitly tied to a named skill;
  • the target problem.

They test SKiC on easy-to-hard generalization tasks, such as longer additions, multiplication, dynamic programming and last-letter concatenation, and on math reasoning with GSM8K and MATH. With only about two exemplars, SKiC reaches near-perfect systematic generalization, where CoT and Least-to-Most fall short. Models also start invoking pretrained skills that never appeared in the prompt. In addition, Llama2 fine-tuned on SKiC-style annotations transfers better to harder problems than when trained on CoT-style data. The work matters because it offers a cheap, plug-in alternative to CoT that improves reasoning on problems harder than the demonstrations.

Key points

  • Ablation results: both the in-context skills and the explicit step-to-skill grounding proved necessary for the gains.
  • Advantage over decomposition methods: SKiC handles problems with non-linear computation graphs (e.g., multiplication, DP), which Least-to-Most and Decomp cannot easily break into a sequence of sub-problems. Because it runs in one stage, it also limits error propagation.
  • Use of internal skills: on MATH, the model produced and applied skills absent from the prompt, such as the Angle Bisector Theorem and Heron's Formula.
  • Robustness and combinability: SKiC tolerates different skill sources and exemplar choices, and it combines with self-consistency and Progressive-Hint.
  • Fine-tuning experiment: GPT4 annotated GSM8K with SKiC-structured solutions, and Llama2 was trained on these and evaluated on the much harder MATH. It outperformed the same setup using GPT4-written CoT annotations.

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, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Jiaao Chen, Xiaoman Pan, Dian Yu, Kaiqiang Song, Xiaoyang Wang, Dong Yu and Jianshu Chen. 2023. Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models. arXiv:2308.00304.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2308.00304 · arXiv 2308.00304

@misc{chen2023skills,
  title = {{Skills-in-Context Prompting: Unlocking Compositionality in Large Language Models}},
  author = {Jiaao Chen, Xiaoman Pan, Dian Yu, Kaiqiang Song, Xiaoyang Wang, Dong Yu and Jianshu Chen},
  year = {2023},
  doi = {10.48550/arxiv.2308.00304},
  eprint = {2308.00304},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2308.00304}
}

References

51 references, 25 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jacob Andreas. 2019. Good-enough compositional data augmentation. arXiv preprint arXiv:1904.09545.
  2. Arthur Bakker, Jantien Smit, and Rupert Wegerif. 2015. Scaffolding and dialogic teaching in mathematics education: Introduction and review. Zdm, 47:1047– 1065. find
  3. Dianne C Berry. 1983. Metacognitive experience and transfer of logical reasoning. The Quarterly Journal of Experimental Psychology Section A, 35(1):39–49. find
  4. Yao Fu, Hao Peng, Ashish Sabharwal, Peter Clark, and Tushar Khot. 2022. Complexity-based prompting for multi-step reasoning. arXiv preprint arXiv:2210.00720.
  5. Sebastian Borgeaud, Arthur Mensch, Jordan Hoffmann, Trevor Cai, Eliza Rutherford, Katie Millican, George Bm Van Den Driessche, Jean-Baptiste Lespiau, Bogdan Damoc, Aidan Clark, et al. 2022. Improving language models by retrieving from trillions of tokens. In International conference on machine learn… find
  6. Yiduo Guo, Yaobo Liang, Chenfei Wu, Wenshan Wu, Dongyan Zhao, and Nan Duan. 2023. Learning to program with natural language. find
  7. Jenny Hammond and Pauline Gibbons. 2005. Putting scaffolding to work: The contribution of scaffolding in articulating esl education. Prospect, 20(1):6–30. find
  8. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,… — Language Models are Few-Shot Learners
  9. Simeng Han, Hailey Schoelkopf, Yilun Zhao, Zhenting Qi, Martin Riddell, Luke Benson, Lucy Sun, Ekaterina Zubova, Yujie Qiao, Matthew Burtell, et al. 2022. Folio: Natural language reasoning with first-order logic. arXiv preprint arXiv:2209.00840.
  10. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring mathematical problem solving with the math dataset.
  11. Jordan Hoffmann, Sebastian Borgeaud, Arthur Mensch, Elena Buchatskaya, Trevor Cai, Eliza Rutherford, Diego de Las Casas, Lisa Anne Hendricks, Johannes Welbl, Aidan Clark, Tom Hennigan, Eric Noland, Katie Millican, George van den Driessche, Bogdan Damoc, Aurelia Guy, Simon Osindero, Karen Simonyan, E… — Training Compute-Optimal Large Language Models
  12. Collin Burns, Pavel Izmailov, Jan Hendrik Kirchner, Bowen Baker, Leo Gao, Leopold Aschenbrenner, Yining Chen, Adrien Ecoffet, Manas Joglekar, Jan Leike, et al. 2023. Weak-to-strong generalization: Eliciting strong capabilities with weak supervision. arXiv preprint arXiv:2312.09390.
  13. Robin Jia and Percy Liang. 2016. Data recombination for neural semantic parsing. arXiv preprint arXiv:1606.03622. find
  14. Xinyun Chen, Chen Liang, Adams Wei Yu, Dawn Song, and Denny Zhou. 2020. Compositional generalization via neural-symbolic stack machines. find
  15. Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. 2020. Scaling laws for neural language models.
  16. Hyung Won Chung, Le Hou, Shayne Longpre, Barret Zoph, Yi Tay, William Fedus, Eric Li, Xuezhi Wang, Mostafa Dehghani, Siddhartha Brahma, et al. 2022. Scaling instruction-finetuned language models. arXiv preprint arXiv:2210.11416. find
  17. Daniel Keysers, Nathanael Schärli, Nathan Scales, Hylke Buisman, Daniel Furrer, Sergii Kashubin, Nikola Momchev, Danila Sinopalnikov, Lukasz Stafiniak, Tibor Tihon, Dmitry Tsarkov, Xiao Wang, Marc van Zee, and Olivier Bousquet. 2020. Measuring compositional generalization: A comprehensive method on… find
  18. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, et al. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168.
  19. Honghua Dong, Jiayuan Mao, Tian Lin, Chong Wang, Lihong Li, and Denny Zhou. 2019. Neural logic machines. arXiv preprint arXiv:1904.11694. — [arXiv:1904.11694](https://arxiv.org/abs/1904.11694) find
  20. Shiela Kheirzadeh and Sarah Sadat Pakzadian. 2016. Depth of processing and age differences. Journal of psycholinguistic research, 45:1137–1149. find
  21. Dheeru Dua, Shivanshu Gupta, Sameer Singh, and Matt Gardner. 2022. Successive prompting for decomposing complex questions. arXiv preprint arXiv:2212.04092. find
  22. Tushar Khot, Kyle Richardson, Daniel Khashabi, and Ashish Sabharwal. 2021. Hey ai, can you solve complex tasks by talking to agents? arXiv preprint arXiv:2110.08542. find
  23. Nouha Dziri, Ximing Lu, Melanie Sclar, Xiang Lorraine Li, Liwei Jian, Bill Yuchen Lin, Peter West, Chandra Bhagavatula, Ronan Le Bras, Jena D Hwang, et al. 2023. Faith and fate: Limits of transformers on compositionality. arXiv preprint arXiv:2305.18654.
  24. Tushar Khot, Harsh Trivedi, Matthew Finlayson, Yao Fu, Kyle Richardson, Peter Clark, and Ashish Sabharwal. 2022. Decomposed prompting: A modular approach for solving complex tasks. arXiv preprint arXiv:2210.02406. find
  25. Brenden Lake and Marco Baroni. 2018. Generalization without systematicity: On the compositional skills of sequence-to-sequence recurrent networks. In International conference on machine learning, pages 2873–2882. PMLR. find
  26. Shaden Smith, Mostofa Patwary, Brandon Norick, Patrick LeGresley, Samyam Rajbhandari, Jared Casper, Zhun Liu, Shrimai Prabhumoye, George Zerveas, Vijay Korthikanti, et al. 2022. Using deepspeed and megatron to train megatron-turing nlg 530b, a large-scale generative language model. arXiv preprint ar…
  27. Aitor Lewkowycz, Anders Andreassen, David Dohan, Ethan Dyer, Henryk Michalewski, Vinay Ramasesh, Ambrose Slone, Cem Anil, Imanol Schlag, Theo Gutman-Solo, et al. 2022. Solving quantitative reasoning problems with language models. arXiv preprint arXiv:2206.14858.
  28. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023a. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971.
  29. Swaroop Mishra, Daniel Khashabi, Chitta Baral, and Hannaneh Hajishirzi. 2022. 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), pages 3470–3487. find
  30. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, Brian Fuller,… — Llama 2: Open Foundation and Fine-Tuned Chat Models
  31. Maxwell Nye, Anders Johan Andreassen, Guy Gur-Ari, Henryk Michalewski, Jacob Austin, David Bieber, David Dohan, Aitor Lewkowycz, Maarten Bosma, David Luan, et al. 2021. Show your work: Scratchpads for intermediate computation with language models. arXiv preprint arXiv:2112.00114.
  32. OpenAI. 2023. Gpt-4 technical report.
  33. 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.
  34. Siru Ouyang, Jiaao Chen, Jiawei Han, and Diyi Yang. 2023. Compositional data augmentation for abstractive conversation summarization. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1471–1488. find
  35. Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R Bowman. 2018. Glue: A multi-task benchmark and analysis platform for natural language understanding. arXiv preprint arXiv:1804.07461.
  36. Xiaoman Pan, Wenlin Yao, Hongming Zhang, Dian Yu, Dong Yu, and Jianshu Chen. 2022. Knowledge-in-context: Towards knowledgeable semi-parametric language models. arXiv preprint arXiv:2210.16433. — [arXiv:2210.16433](https://arxiv.org/abs/2210.16433) find
  37. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.
  38. Abulhair Saparov and He He. 2022. Language models are greedy reasoners: A systematic formal analysis of chain-of-thought. arXiv preprint arXiv:2210.01240.
  39. Jason Wei, Maarten Bosma, Vincent Y Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M Dai, and Quoc V Le. 2021. Finetuned language models are zero-shot learners. arXiv preprint arXiv:2109.01652.
  40. Avi Schwarzschild, Eitan Borgnia, Arjun Gupta, Furong Huang, Uzi Vishkin, Micah Goldblum, and Tom Goldstein. 2021. Can you learn an algorithm? generalizing from easy to hard problems with recurrent networks. Advances in Neural Information Processing Systems, 34:6695–6706. find
  41. Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, Ed H. Chi, Tatsunori Hashimoto, Oriol Vinyals, Percy Liang, Jeff Dean, and William Fedus. 2022a. Emergent abilities of large language models.
  42. NAN SHAO, Zefan Cai, Hanwei xu, Chonghua Liao, Yanan Zheng, and Zhilin Yang. 2023. Compositional task representations for large language models. In The Eleventh International Conference on Learning Representations. find
  43. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. 2022b. Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35:24824–24837.
  44. Sean Welleck, Jiacheng Liu, Ximing Lu, Hannaneh Hajishirzi, and Yejin Choi. 2022. Naturalprover: Grounded mathematical proof generation with language models. Advances in Neural Information Processing Systems, 35:4913–4927. find
  45. Yifan Zhang, Jingqin Yang, Yang Yuan, and Andrew Chi-Chih Yao. 2023. Cumulative reasoning with large language models. arXiv preprint arXiv:2308.04371. find
  46. Chuanyang Zheng, Zhengying Liu, Enze Xie, Zhenguo Li, and Yu Li. 2023a. Progressive-hint prompting improves reasoning in large language models. arXiv preprint arXiv:2304.09797. find
  47. Huaixiu Steven Zheng, Swaroop Mishra, Xinyun Chen, Heng-Tze Cheng, Ed H Chi, Quoc V Le, and Denny Zhou. 2023b. Take a step back: Evoking reasoning via abstraction in large language models. arXiv preprint arXiv:2310.06117. find
  48. Aojun Zhou, Ke Wang, Zimu Lu, Weikang Shi, Sichun Luo, Zipeng Qin, Shaoqing Lu, Anya Jia, Linqi Song, Mingjie Zhan, et al. 2023. Solving challenging math word problems using gpt-4 code interpreter with code-based self-verification. arXiv preprint arXiv:2308.07921.
  49. Denny Zhou, Nathanael Schärli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Olivier Bousquet, Quoc Le, and Ed Chi. 2022a. Least-to-most prompting enables complex reasoning in large language models. arXiv preprint arXiv:2205.10625.
  50. Hattie Zhou, Azade Nova, Hugo Larochelle, Aaron Courville, Behnam Neyshabur, and Hanie Sedghi. 2022b. Teaching algorithmic reasoning via in-context learning. arXiv preprint arXiv:2211.09066. find
  51. Pei Zhou, Jay Pujara, Xiang Ren, Xinyun Chen, Heng-Tze Cheng, Quoc V Le, Ed H Chi, Denny Zhou, Swaroop Mishra, and Huaixiu Steven Zheng. 2024. Self-discover: Large language models self-compose reasoning structures. arXiv preprint arXiv:2402.03620. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/chen-2023-skills-context-prompting-unlocking-compositionality/