Elamur Research⌕
☰
home / papers / 2024 / In-Context Principle Learning from Mistakes

In-Context Principle Learning from Mistakes

Paper · 2024 · Tianjun Zhang, Aman Madaan, Luyu Gao et al.

Summary

Standard in-context learning shows a model only correct demonstrations, so it never learns from what goes wrong on a task. This paper introduces LEAP (Learning Principles), which gets more out of the same few labeled examples. First, the model answers each demonstration input zero-shot with step-by-step reasoning, sampled at non-zero temperature so that some answers fail. Next, the same model compares each failure with the gold reasoning and answer and explains the error. These explanations become low-level principles, which can be condensed into a short list of general high-level rules. The rules are prepended to the ordinary few-shot prompt and reused across the entire test set. No human feedback or fine-tuning is needed. The authors test LEAP on GSM8K, MATH, HotpotQA, DROP and 27 Big-Bench Hard tasks with GPT-3.5-turbo, GPT-4, GPT-4-turbo and Claude-2.1. It generally outperforms standard few-shot prompting, with and without CoT. The work matters because it brings learning from mistakes into ICL at almost no extra data or compute cost.

Key points

  • With GPT-4, LEAP beats few-shot CoT by 7.5% on DROP and 3.3% on HotpotQA, and it improves 16 Big-Bench Hard tasks.
  • Mistake generation draws n = 15 zero-shot CoT samples per example ("Think step-by-step"). Wrong outputs are found by checking them against the ground-truth answer.
  • There are two variants. LEAP LOW-LEVEL uses per-mistake insights directly. LEAP HIGH-LEVEL condenses them into about 5 example-agnostic bullet points.
  • The method is cheap: about 34 LLM calls in total, using only 3 labeled examples. By comparison, OPRO needs roughly 1,500 to 27,000 calls and a training set of about 260 GSM8K examples, and EvoPrompt needs about 10,000 calls.
  • A zero-shot variant (Zero-shot-LEAP) adds 1% for GPT-4 on GSM8K. In the authors' reproduction, the related Contrastive CoT approach performed no better than the baseline. The conclusion lists Gemini Pro among the improved models, whereas the abstract names Claude-2.1.

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

🎓Carnegie Mellon UniversityPittsburgh · United States539 works · 179,533 citations🎓University of California, BerkeleyBerkeley · United States175 works · 227,440 citations

 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

Cite

Cite as Tianjun Zhang, Aman Madaan, Luyu Gao, Steven Zheng, Swaroop Mishra, Yiming Yang, Niket Tandon and Uri Alon. 2024. In-Context Principle Learning from Mistakes. arXiv:2402.05403.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2402.05403 · arXiv 2402.05403

@misc{zhang2024context,
  title = {{In-Context Principle Learning from Mistakes}},
  author = {Tianjun Zhang, Aman Madaan, Luyu Gao, Steven Zheng, Swaroop Mishra, Yiming Yang, Niket Tandon and Uri Alon},
  year = {2024},
  doi = {10.48550/arxiv.2402.05403},
  eprint = {2402.05403},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2402.05403}
}

References

57 references, 38 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Gemini Team, A., Anil, R., Borgeaud, S., Wu, Y., Alayrac, J.-B., Yu, J., Soricut, R., Schalkwyk, J., Dai, A. M., Hauth, A., et al. Gemini: a family of highly capable multimodal models. arXiv preprint arXiv:2312.11805, 2023.
  2. Alayrac, J.-B., Donahue, J., Luc, P., Miech, A., Barr, I., Hasson, Y., Lenc, K., Mensch, A., Millican, K., Reynolds, M., et al. Flamingo: a visual language model for few-shot learning. Advances in Neural Information Processing Systems, 35:23716–23736, 2022.
  3. Guo, Q., Wang, R., Guo, J., Li, B., Song, K., Tan, X., Liu, G., Bian, J., and Yang, Y. Connecting large language models with evolutionary algorithms yields powerful prompt optimizers. arXiv preprint arXiv:2309.08532, 2023. find
  4. Anil, R., Dai, A. M., Firat, O., Johnson, M., Lepikhin, D., Passos, A., Shakeri, S., Taropa, E., Bailey, P., Chen, Z., et al. Palm 2 technical report. arXiv preprint arXiv:2305.10403, 2023.
  5. Hendrycks, D., Burns, C., Kadavath, S., Arora, A., Basart, S., Tang, E., Song, D., and Steinhardt, J. Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874, 2021.
  6. Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J. D., Dhariwal, P., Neelakantan, A., Shyam, P., Sastry, G., Askell, A., et al. Language models are few-shot learners. Advances in neural information processing systems, 33: 1877–1901, 2020.
  7. Huang, J., Chen, X., Mishra, S., Zheng, H. S., Yu, A. W., Song, X., and Zhou, D. Large language models cannot self-correct reasoning yet. arXiv preprint arXiv:2310.01798, 2023a.
  8. Chen, X., Lin, M., Schärli, N., and Zhou, D. Teaching large language models to self-debug. arXiv preprint arXiv:2304.05128, 2023. find
  9. Chialvo, D. R. and Bak, P. Learning from mistakes. Neuroscience, 90(4):1137–1148, 1999. find
  10. Huang, J., Gu, S., Hou, L., Wu, Y., Wang, X., Yu, H., and Han, J. Large language models can self-improve. In Bouamor, H., Pino, J., and Bali, K. (eds.), Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pp. 1051–1068, Singapore, December 2023b. Association for C…
  11. Chowdhery, A., Narang, S., Devlin, J., Bosma, M., Mishra, G., Roberts, A., Barham, P., Chung, H., Sutton, C., Gehrmann, S., et al. Palm: Scaling language modeling with pathways (no. arxiv: 2204.02311). arxiv, 2022.
  12. Kaddour, J., Harris, J., Mozes, M., Bradley, H., Raileanu, R., and McHardy, R. Challenges and applications of large language models. arXiv preprint arXiv:2307.10169, 2023.
  13. Cobbe, K., Kosaraju, V., Bavarian, M., Chen, M., Jun, H., Kaiser, L., Plappert, M., Tworek, J., Hilton, J., Nakano, R., et al. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168, 2021.
  14. Kojima, T., Gu, S. S., Reid, M., Matsuo, Y., and Iwasawa, Y. Large Language Models are Zero-Shot Reasoners. arXiv preprint arXiv:2205.11916, 2022.
  15. Chia, Y. K., Chen, G., Tuan, L. A., Poria, S., and Bing, L. Contrastive chain-of-thought prompting. arXiv preprint arXiv:2311.09277, 2023.
  16. Li, Y., Choi, D., Chung, J., Kushman, N., Schrittwieser, J., Leblond, R., Eccles, T., Keeling, J., Gimeno, F., Dal Lago, A., et al. Competition-level code generation with alphacode. Science, 378(6624):1092–1097, 2022.
  17. Dua, D., Wang, Y., Dasigi, P., Stanovsky, G., Singh, S., and Gardner, M. DROP: A reading comprehension benchmark requiring discrete reasoning over paragraphs. In Proc. of NAACL, 2019a.
  18. Liu, J., Shen, D., Zhang, Y., Dolan, B., Carin, L., and Chen, W. What makes good in-context examples for gpt-3? In Workshop on Knowledge Extraction and Integration for Deep Learning Architectures; Deep Learning Inside Out, 2021. URL https://api.semanticscholar.org/CorpusID:231632658. ↗
  19. Dua, D., Wang, Y., Dasigi, P., Stanovsky, G., Singh, S., and Gardner, M. Drop: A reading comprehension benchmark requiring discrete reasoning over paragraphs. In North American Chapter of the Association for Computational Linguistics, 2019b. URL https://api.semanticscholar.org/CorpusID:67855846. ↗
  20. Edmondson, A. Psychological safety and learning behavior in work teams. Administrative science quarterly, 44(2): 350–383, 1999. find
  21. Liu, P., Yuan, W., Fu, J., Jiang, Z., Hayashi, H., and Neubig, G. Pre-train, prompt, and predict: A systematic survey of prompting methods in natural language processing. ACM Computing Surveys, 55(9):1–35, 2023.
  22. Edmondson, A. C. Learning from mistakes is easier said than done: Group and organizational influences on the detection and correction of human error. The Journal of Applied Behavioral Science, 32(1):5–28, 1996. find
  23. Lu, S., Bigoulaeva, I., Sachdeva, R., Madabushi, H. T., and Gurevych, I. Are emergent abilities in large language models just in-context learning? arXiv preprint arXiv:2309.01809, 2023. find
  24. Madaan, A., Tandon, N., Clark, P., and Yang, Y. Memory-assisted prompt editing to improve gpt-3 after deployment. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pp. 2833–2861, 2022. find
  25. Shin, T., Razeghi, Y., Logan IV, R. L., Wallace, E., and Singh, S. Autoprompt: Eliciting knowledge from language models with automatically generated prompts. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pp. 4222– 4235, 2020.
  26. Madaan, A., Tandon, N., Gupta, P., Hallinan, S., Gao, L., Wiegreffe, S., Alon, U., Dziri, N., Prabhumoye, S., Yang, Y., et al. Self-refine: Iterative refinement with self-feedback. arXiv preprint arXiv:2303.17651, 2023. find
  27. Sun, Z., Shen, Y., Zhou, Q., Zhang, H., Chen, Z., Cox, D., Yang, Y., and Gan, C. Principle-driven self-alignment of language models from scratch with minimal human supervision. arXiv preprint arXiv:2305.03047, 2023.
  28. Min, S., Lewis, M., Zettlemoyer, L., and Hajishirzi, H. Metaicl: Learning to learn in context. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pp. 2791–2809, 2022a. find
  29. Suzgun, M., Scales, N., Schärli, N., Gehrmann, S., Tay, Y., Chung, H. W., Chowdhery, A., Le, Q. V., Chi, E. H., Zhou, D., et al. Challenging big-bench tasks and whether chain-of-thought can solve them. arXiv preprint arXiv:2210.09261, 2022.
  30. Min, S., Lyu, X., Holtzman, A., Artetxe, M., Lewis, M., Hajishirzi, H., and Zettlemoyer, L. Rethinking the role of demonstrations: What makes in-context learning work? In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pp. 11048– 11064, 2022b.
  31. Touvron, H., Lavril, T., Izacard, G., Martinet, X., Lachaux, M.-A., Lacroix, T., Rozière, B., Goyal, N., Hambro, E., Azhar, F., et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023a.
  32. Touvron, H., Martin, L., Stone, K., Albert, P., Almahairi, A., Babaei, Y., Bashlykov, N., Batra, S., Bhargava, P., Bhosale, S., et al. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288, 2023b.
  33. Minsky, M. and Papert, S. An introduction to computational geometry. Cambridge tiass., HIT, 479(480):104, 1969. find
  34. Mishra, S., Khashabi, D., Baral, C., and Hajishirzi, H. Cross-task generalization via natural language crowdsourcing instructions. arXiv preprint arXiv:2104.08773, 2021. find
  35. Valmeekam, K., Marquez, M., and Kambhampati, S. Investigating the effectiveness of self-critiquing in LLMs solving planning tasks. In NeurIPS 2023 Foundation Models for Decision Making Workshop, 2023. URL https://openreview.net/forum?id=gGQfkyb0KL. ↗
  36. Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., et al. Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems, 35:27730–27744, 2022.
  37. Qiao, S., Ou, Y., Zhang, N., Chen, X., Yao, Y., Deng, S., Tan, C., Huang, F., and Chen, H. Reasoning with language model prompting: A survey. In Rogers, A., Boyd-Graber, J., and Okazaki, N. (eds.), Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Lon…
  38. Wang, X., Wei, J., Schuurmans, D., Le, Q., Chi, E., and Zhou, D. Rationale-Augmented Ensembles in Language Models. arXiv preprints arXiv:2207.00747, 2022a. find
  39. Wang, Y., Kordi, Y., Mishra, S., Liu, A., Smith, N. A., Khashabi, D., and Hajishirzi, H. Self-instruct: Aligning language model with self generated instructions. arXiv preprint arXiv:2212.10560, 2022b. find
  40. Wei, J., Bosma, M., Zhao, V. Y., Guu, K., Yu, A. W., Lester, B., Du, N., Dai, A. M., and Le, Q. V. Finetuned language models are zero-shot learners. In ICLR, 2022a.
  41. Wei, J., Tay, Y., Bommasani, R., Raffel, C., Zoph, B., Borgeaud, S., Yogatama, D., Bosma, M., Zhou, D., Metzler, D., Chi, E. H., Hashimoto, T., Vinyals, O., Liang, P., Dean, J., and Fedus, W. Emergent abilities of large language models. Transactions on Machine Learning Research, 2022b. ISSN 2835-885…
  42. Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., Sutskever, I., et al. Language models are unsupervised multitask learners. OpenAI blog, 1(8):9, 2019. find
  43. Rosenblatt, F. The perceptron, a perceiving and recognizing automaton Project Para. Cornell Aeronautical Laboratory, 1957. find
  44. Wei, J., Wang, X., Schuurmans, D., Bosma, M., Xia, F., Chi, E., Le, Q. V., Zhou, D., et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35: 24824–24837, 2022c.
  45. Zhao, W. X., Zhou, K., Li, J., Tang, T., Wang, X., Hou, Y., Min, Y., Zhang, B., Zhang, J., Dong, Z., et al. A survey of large language models. arXiv preprint arXiv:2303.18223, 2023.
  46. Zhou, D., Schärli, N., Hou, L., Wei, J., Scales, N., Wang, X., Schuurmans, D., Bousquet, O., Le, Q., and Chi, E. Least-to-Most Prompting Enables Complex Reasoning in Large Language Models. arXiv preprint arXiv:2205.10625, 2022.
  47. Wei, J., Wei, J., Tay, Y., Tran, D., Webson, A., Lu, Y., Chen, X., Liu, H., Huang, D., Zhou, D., et al. Larger language models do in-context learning differently. arXiv preprint arXiv:2303.03846, 2023.
  48. Wiener, N. Cybernetics; or control and communication in the animal and the machine. 1948. find
  49. Xi, Z., Chen, W., Guo, X., He, W., Ding, Y., Hong, B., Zhang, M., Wang, J., Jin, S., Zhou, E., et al. The rise and potential of large language model based agents: A survey. arXiv preprint arXiv:2309.07864, 2023.
  50. Yang, C., Wang, X., Lu, Y., Liu, H., Le, Q. V., Zhou, D., and Chen, X. Large language models as optimizers. arXiv preprint arXiv:2309.03409, 2023.
  51. Yang, Z., Qi, P., Zhang, S., Bengio, Y., Cohen, W. W., Salakhutdinov, R., and Manning, C. D. Hotpotqa: A dataset for diverse, explainable multi-hop question answering. In Conference on Empirical Methods in Natural Language Processing, 2018a. URL https://api.semanticscholar.org/CorpusID:52822214. ↗
  52. Yang, Z., Qi, P., Zhang, S., Bengio, Y., Cohen, W. W., Salakhutdinov, R., and Manning, C. D. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Conference on Empirical Methods in Natural Language Processing (EMNLP), 2018b.
  53. Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K. R., and Cao, Y. React: Synergizing reasoning and acting in language models. In The Eleventh International Conference on Learning Representations, 2022.
  54. Zelikman, E., Wu, Y., Mu, J., and Goodman, N. STar: Bootstrapping reasoning with reasoning. In Oh, A. H., Agarwal, A., Belgrave, D., and Cho, K. (eds.), Advances in Neural Information Processing Systems, 2022. URL https://openreview.net/forum? id=_3ELRdg2sgI. ↗
  55. Zhang, S., Roller, S., Goyal, N., Artetxe, M., Chen, M., Chen, S., Dewan, C., Diab, M., Li, X., Lin, X. V., et al. Opt: Open pre-trained transformer language models. arXiv preprint arXiv:2205.01068, 2022.
  56. Zhang, Z., Yao, Y., Zhang, A., Tang, X., Ma, X., He, Z., Wang, Y., Gerstein, M., Wang, R., Liu, G., et al. Igniting language intelligence: The hitchhiker’s guide from chain-of-thought reasoning to language agents. arXiv preprint arXiv:2311.11797, 2023. — [arXiv:2311.11797](https://arxiv.org/abs/2311.11797) find
  57. A. Additional Results on Text Reasoning find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhang-2024-context-principle-learning-mistakes/