Elamur Research⌕
☰
home / papers / 2023 / Lion: Adversarial Distillation of Proprietary Large Language Models

Lion: Adversarial Distillation of Proprietary Large Language Models

Paper · 2023 · Yuxin Jiang, Chunkit Chan, Mingyang Chen et al.

Summary

Distilling a closed, API-only LLM into a small open model has usually been one-directional. The student is fine-tuned to copy the teacher's answers on a fixed instruction set and gets no signal about where it is still weak. Jiang et al. adapt adversarial knowledge distillation to this black-box setting. The teacher's weights and gradients are inaccessible, so they prompt ChatGPT to play three roles:

  • teacher, which writes reference answers;
  • referee, which judges how far the student's output falls behind the teacher's;
  • generator, which creates new instructions resembling the ones the student handled poorly.

Each round therefore cycles through imitation, discrimination and generation, so training keeps focusing on difficult cases. The process starts from Alpaca's instructions and a LLaMA base model. Three rounds produce about 70K training examples and the student model Lion. Lion-13B reportedly comes close to ChatGPT on open-ended generation and clearly outperforms Vicuna-13B on zero-shot reasoning. The study shows that targeting a student's failures makes imitation-based distillation much more data-efficient, without any human labeling.

Key points

  • The framework uses one proprietary LLM (gpt-3.5-turbo) as Teacher, Referee and Generator through different prompt templates. It also maintains two data pools: a Train Pool for training and a Cache Pool for evaluation.
  • The initial data is Alpaca's training set, which grew from 175 hand-picked seed instructions. After three adversarial iterations on LLaMA, the student is trained on roughly 70K examples.
  • Lion-13B beats Vicuna-13B by 55.4% on BIG-Bench Hard (BBH) and by 16.7% on AGIEval. Overall evaluation covers 40 sub-tasks across generation and reasoning.
  • Training one student took about 450k API calls, around 70% of WizardLM's 624k, and cost close to $900.
  • The authors list several limitations:
    • Lion still trails ChatGPT on complex reasoning.
    • It handles multi-turn dialogue poorly.
    • It is capped at 1024 tokens.
    • Weights are restricted to research use under the LLaMA license.

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

🎓Hong Kong University of Science and TechnologyHong Kong · China170 works · 26,341 citations🎓The Hong Kong University of Science and Technology (Guangzhou)Guangzhou · China67 works · 1,765 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,894 worksEfficient models & compression · 297 works

Cite

Cite as Yuxin Jiang, Chunkit Chan, Mingyang Chen and Wei Wang. 2023. Lion: Adversarial Distillation of Proprietary Large Language Models. 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.189

@inproceedings{jiang2023lion,
  title = {{Lion: Adversarial Distillation of Proprietary Large Language Models}},
  author = {Yuxin Jiang, Chunkit Chan, Mingyang Chen and Wei Wang},
  booktitle = {Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing},
  year = {2023},
  doi = {10.18653/v1/2023.emnlp-main.189},
  url = {https://doi.org/10.18653/v1/2023.emnlp-main.189}
}

References

47 references, 22 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Sravanti Addepalli, Gaurav Kumar Nayak, Anirban Chakraborty, and Venkatesh Babu Radhakrishnan. 2020. Degan: Data-enriching gan for retrieving representative samples from a trained classifier. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, pages 3130–3137. find
  2. Vamsi Aribandi, Yi Tay, Tal Schuster, Jinfeng Rao, Huaixiu Steven Zheng, Sanket Vaibhav Mehta, Honglei Zhuang, Vinh Q. Tran, Dara Bahri, Jianmo Ni, Jai Prakash Gupta, Kai Hui, Sebastian Ruder, and Donald Metzler. 2022. Ext5: Towards extreme multi-task scaling for transfer learning. In The Tenth Inte…
  3. Yejin Bang, Samuel Cahyawijaya, Nayeon Lee, Wenliang Dai, Dan Su, Bryan Wilie, Holy Lovenia, Ziwei Ji, Tiezheng Yu, Willy Chung, Quyet V. Do, Yan Xu, and Pascale Fung. 2023. A multitask, multilingual, multimodal evaluation of chatgpt on reasoning, hallucination, and interactivity. CoRR, abs/2302.040…
  4. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott M. Lundberg, Harsha Nori, Hamid Palangi, Marco Túlio Ribeiro, and Yi Zhang. 2023. Sparks of artificial general intelligence: Early experiments with GPT-4. CoRR, ab…
  5. Chunkit Chan and Tsz Ho Chan. 2023. Discourse-aware prompt for argument impact classification. In Proceedings of the 15th International Conference on Machine Learning and Computing, ICMLC 2023, Zhuhai, China, February 17-20, 2023, pages 165– 171. ACM. find
  6. Chunkit Chan, Jiayang Cheng, Weiqi Wang, Yuxin Jiang, Tianqing Fang, Xin Liu, and Yangqiu Song. 2023a. Chatgpt evaluation on sentence level relations: A focus on temporal, causal, and discourse relations. CoRR, abs/2304.14827.
  7. Chunkit Chan, Xin Liu, Tsz Ho Chan, Jiayang Cheng, Yangqiu Song, Ginny Y. Wong, and Simon See. 2023b. Self-consistent narrative prompts on abductive natural language inference. CoRR, abs/2309.08303. find
  8. Chunkit Chan, Xin Liu, Jiayang Cheng, Zihan Li, Yangqiu Song, Ginny Y. Wong, and Simon See. 2023c. Discoprompt: Path prediction prompt tuning for implicit discourse relation recognition. In Findings of the Association for Computational Linguistics: ACL 2023, Toronto, Canada, July 9-14, 2023, pages 3…
  9. Akshay Chawla, Hongxu Yin, Pavlo Molchanov, and Jose Alvarez. 2021. Data-free knowledge distillation for object detection. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 3289–3298. find
  10. Yen-Chun Chen, Zhe Gan, Yu Cheng, Jingzhou Liu, and Jingjing Liu. 2019. Distilling knowledge learned in bert for text generation. arXiv preprint arXiv:1911.03829. find
  11. Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E. Gonzalez, Ion Stoica, and Eric P. Xing. 2023. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality. find
  12. Gongfan Fang, Kanya Mo, Xinchao Wang, Jie Song, Shitao Bei, Haofei Zhang, and Mingli Song. 2022. Up to 100x faster data-free knowledge distillation. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 6597–6604. find
  13. Gongfan Fang, Jie Song, Chengchao Shen, Xinchao Wang, Da Chen, and Mingli Song. 2019. Data-free adversarial distillation. CoRR, abs/1912.11006. — [arXiv:1912.11006](https://arxiv.org/abs/1912.11006) find
  14. Arnav Gudibande, Eric Wallace, Charlie Snell, Xinyang Geng, Hao Liu, Pieter Abbeel, Sergey Levine, and Dawn Song. 2023. The false promise of imitating proprietary llms. CoRR, abs/2305.15717.
  15. Byeongho Heo, Minsik Lee, Sangdoo Yun, and Jin Young Choi. 2019. Knowledge distillation with adversarial samples supporting decision boundary. In The Thirty-Third AAAI Conference on Artificial Intelligence, AAAI 2019, The Thirty-First Innovative Applications of Artificial Intelligence Conference, IA… find
  16. Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531.
  17. Yuxin Jiang, Linhan Zhang, and Wei Wang. 2022. Improved universal sentence embeddings with prompt-based contrastive learning and energy-based learning. In Findings of the Association for Computational Linguistics: EMNLP 2022, Abu Dhabi, United Arab Emirates, December 7-11, 2022, pages 3021–3035. Ass… find
  18. Sanjay Kariyappa, Atul Prakash, and Moinuddin K Qureshi. 2021. Maze: Data-free model stealing attack using zeroth-order gradient estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13814–13823. find
  19. Jan Kocon, Igor Cichecki, Oliwier Kaszyca, Mateusz Kochanek, Dominika Szydlo, Joanna Baran, Julita Bielaniewicz, Marcin Gruza, Arkadiusz Janz, Kamil Kanclerz, Anna Kocon, Bartlomiej Koptyra, Wiktoria Mieleszczenko-Kowszewicz, Piotr Milkowski, Marcin Oleksy, Maciej Piasecki, Lukasz Radlinski, Konrad… — ChatGPT: Jack of all trades, master of none
  20. Haoran Li, Dadi Guo, Wei Fan, Mingshi Xu, and Yangqiu Song. 2023. Multi-step jailbreaking privacy attacks on chatgpt. CoRR, abs/2304.05197.
  21. Paul Micaelli and Amos J. Storkey. 2019a. Zero-shot knowledge transfer via adversarial belief matching. In Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vancouver, BC, Canada, pages 9547– 9557… find
  22. Paul Micaelli and Amos J Storkey. 2019b. Zero-shot knowledge transfer via adversarial belief matching. Advances in Neural Information Processing Systems, 32. find
  23. 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), ACL 2022, Dublin, Ireland, May… find
  24. Subhabrata Mukherjee, Arindam Mitra, Ganesh Jawahar, Sahaj Agarwal, Hamid Palangi, and Ahmed Hassan Awadallah. 2023. Orca: Progressive learning from complex explanation traces of GPT-4. CoRR, abs/2306.02707.
  25. OpenAI. 2023. GPT-4 technical report. abs/2303.08774.
  26. CoRR, TB OpenAI. 2022. Chatgpt: Optimizing language models for dialogue. OpenAI. find
  27. Tribhuvanesh Orekondy, Bernt Schiele, and Mario Fritz. 2019. Knockoff nets: Stealing functionality of black-box models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4954–4963. find
  28. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul F. Christiano, Jan Leike, and Ryan Lowe. 2022. Tr… — Training language models to follow instructions with human feedback
  29. Nicolas Papernot, Patrick McDaniel, Ian Goodfellow, Somesh Jha, Z Berkay Celik, and Ananthram Swami. 2017. Practical black-box attacks against machine learning. In Proceedings of the 2017 ACM on Asia conference on computer and communications security, pages 506–519. find
  30. Ilija Radosavovic, Piotr Dollár, Ross Girshick, Georgia Gkioxari, and Kaiming He. 2018. Data distillation: Towards omni-supervised learning. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4119–4128. find
  31. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. J. Mach. Learn. Res., 21:140:1–140:67.
  32. Partha Pratim Ray. 2023. Chatgpt: A comprehensive review on background, applications, key challenges, bias, ethics, limitations and future scope. Internet of Things and Cyber-Physical Systems. find
  33. Victor Sanh, Albert Webson, Colin Raffel, Stephen H. Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Arun Raja, Manan Dey, M Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Nihal V. Nayak, Debajyoti Datta, Jonatha… — Multitask Prompted Training Enables Zero-Shot Task Generalization
  34. Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, et al. 2022. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. arXiv preprint arXiv:2206.04615.
  35. Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc V. Le, Ed H. Chi, Denny Zhou, and Jason Wei. 2022. Challenging big-bench tasks and whether chain-of-thought can solve them. CoRR, abs/2210.09261.
  36. Rohan Taori, Ishaan Gulrajani, Tianyi Zhang, Yann Dubois, Xuechen Li, Carlos Guestrin, Percy Liang, and Tatsunori B. Hashimoto. 2023. Stanford alpaca: An instruction-following llama model. https://github.com/tatsu-lab/stanford_alpaca. ↗
  37. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation language models. arXiv pr…
  38. Jean-Baptiste Truong, Pratyush Maini, Robert J Walls, and Nicolas Papernot. 2021. Data-free model extraction. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4771–4780. find
  39. 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. CoRR, abs/2305.17926. find
  40. Yizhong Wang, Yeganeh Kordi, Swaroop Mishra, Alisa Liu, Noah A. Smith, Daniel Khashabi, and Hannaneh Hajishirzi. 2022. Self-instruct: Aligning language model with self generated instructions. CoRR, abs/2212.10560. find
  41. Jason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, and Quoc V. Le. 2022a. Finetuned language models are zero-shot learners. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenRevie…
  42. Jason Wei, Maarten Bosma, Vincent Y. Zhao, Kelvin Guu, Adams Wei Yu, Brian Lester, Nan Du, Andrew M. Dai, and Quoc V. Le. 2022b. Finetuned language models are zero-shot learners. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenRevie…
  43. 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. 2022c. Emergent abilities of large language models. CoRR, abs/2206.07…
  44. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, and Denny Zhou. 2022d. Chain-of-thought prompting elicits reasoning in large language models. In NeurIPS.
  45. Can Xu, Qingfeng Sun, Kai Zheng, Xiubo Geng, Pu Zhao, Jiazhan Feng, Chongyang Tao, and Daxin Jiang. 2023. Wizardlm: Empowering large language models to follow complex instructions. find
  46. Hongxu Yin, Pavlo Molchanov, Jose M Alvarez, Zhizhong Li, Arun Mallya, Derek Hoiem, Niraj K Jha, and Jan Kautz. 2020. Dreaming to distill: Data-free knowledge transfer via deepinversion. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8715–8724. find
  47. Wanjun Zhong, Ruixiang Cui, Yiduo Guo, Yaobo Liang, Shuai Lu, Yanlin Wang, Amin Saied, Weizhu Chen, and Nan Duan. 2023. Agieval: A human-centric benchmark for evaluating foundation models. CoRR, abs/2304.06364.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/jiang-2023-lion-adversarial-distillation-proprietary-large/