Elamur Research
☰
home / papers / 2023 / LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset

LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive license to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/2309.11998

Summary

Research into how people use large language models has been held back by the lack of public logs of genuine user interactions. Serving models is expensive, commercial vendors keep their query data private, and open models struggle to attract users on their own. The authors fill this gap with LMSYS-Chat-1M, a release of one million conversations collected between April and August 2023. The data comes from their free Vicuna demo and the gamified Chatbot Arena site, where visitors talked to 25 open and proprietary LLMs. Users agreed to terms of use that permit publication. Conversations with personal data were removed, while unsafe dialogues were deliberately kept and labeled with OpenAI moderation API outputs to support safety research. The paper reports dataset statistics and a clustering-based topic analysis, which shows that programming requests dominate and that unsafe requests make up a notable share. Four applications demonstrate the data's usefulness:

  • small moderation models fine-tuned to rival GPT-4;
  • a jailbreak-based safety benchmark;
  • Llama-2 instruction tuning that matches Vicuna on MMLU and MT-bench;
  • the Arena-Hard-200 benchmark.

The resource offers a rare open view of real-world, multilingual LLM usage.

Key points

  • Scale: 1,000,000 conversations with 25 models from 210,479 unique IPs in over 150 languages. Ranked by frequency, English comes first, then Portuguese, then Russian, with Chinese fourth and Spanish fifth (identified using Polyglot). Vicuna, the site's default model, has the most conversations.
  • Statistics (Llama2 tokenizer): averages of 2.0 turns per sample, 69.5 tokens per prompt and 214.5 tokens per response. Unlike Anthropic HH, OpenAssistant and Chatbot Arena Conversations, the dataset has no human preference labels.
  • Topic analysis: 100K English prompts were embedded with all-mpnet-base-v2 and grouped into 20 k-means clusters, which GPT-4 then summarized. Coding and software clusters lead, explicit or toxic content forms three clusters, and some clusters appear to contain scripted, templated submissions.
  • Arena-Hard-200: 200 hard, high-quality user prompts selected by having GPT-3.5-Turbo rate each prompt from 1 to 10. The benchmark exposes the gap between proprietary and open models.
  • Access: the dataset is on Hugging Face as lmsys/lmsys-chat-1m, and the serving code is in FastChat.

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

208 papers of our library cite it (of the 140673 whose bibliographies we read); 5 of them have a page here, 203 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2309.11998 · arXiv 2309.11998

@misc{zheng2023lmsys,
  title = {{LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset}},
  author = {Zheng, Lianmin and Chiang, Wei-Lin and Sheng, Ying and Li, Tianle and Zhuang, Siyuan and Wu, Zhanghao and Zhuang, Yonghao and Li, Zhuohan and Lin, Zi and Xing, Eric P. and Gonzalez, Joseph E. and Stoica, Ion and Zhang, Hao},
  year = {2023},
  doi = {10.48550/arxiv.2309.11998},
  eprint = {2309.11998},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2309.11998}
}

References

46 references, 26 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. Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, et al. Training a helpful and harmless assistant with reinforcement learning from human feedback. arXiv preprint arXiv:2204.05862, 2022a. find
  3. Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, et al. Constitutional ai: Harmlessness from ai feedback. arXiv preprint arXiv:2212.08073, 2022b. find our card
  4. Clark Barrett, Brad Boyd, Ellie Burzstein, Nicholas Carlini, Brad Chen, Jihye Choi, Amrita Roy Chowdhury, Mihai Christodorescu, Anupam Datta, Soheil Feizi, et al. Identifying and mitigating the security risks of generative ai. arXiv preprint arXiv:2308.14840, 2023. — [arXiv:2308.14840](https://arxiv.org/abs/2308.14840) find
  5. Rishabh Bhardwaj and Soujanya Poria. Red-teaming large language models using chain of utterances for safety-alignment, 2023. find
  6. Paweł Budzianowski, Tsung-Hsien Wen, Bo-Hsiang Tseng, Iñigo Casanueva, Stefan Ultes, Osman Ramadan, and Milica Gasic. Multiwoz-a large-scale multi-domain wizard-of-oz dataset for task-oriented dialogue modelling. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Process… find
  7. Nicholas Carlini, Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, et al. Extracting training data from large language models. In 30th USENIX Security Symposium (USENIX Security 21), pp. 2633–2650, 2021. find our card
  8. 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
  9. Lingjiao Chen, Matei Zaharia, and James Zou. Frugalgpt: How to use large language models while reducing cost and improving performance. arXiv preprint arXiv:2305.05176, 2023. find our card
  10. 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
  11. Yew Ken Chia, Pengfei Hong, Lidong Bing, and Soujanya Poria. Instructeval: Towards holistic evaluation of instruction-tuned large language models. arXiv preprint arXiv:2306.04757, 2023. find
  12. 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. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality, March 2023. URL https: //lmsys.org/blog/2023-03-30-vicuna… find
  13. Ning Ding, Yulin Chen, Bokai Xu, Yujia Qin, Zhi Zheng, Shengding Hu, Zhiyuan Liu, Maosong Sun, and Bowen Zhou. Enhancing chat language models by scaling high-quality instructional conversations. arXiv preprint arXiv:2305.14233, 2023. find our card
  14. Zhengxiao Du, Yujie Qian, Xiao Liu, Ming Ding, Jiezhong Qiu, Zhilin Yang, and Jie Tang. Glm: General language model pretraining with autoregressive blank infilling. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 320–335, 2022. find our card
  15. Kawin Ethayarajh, Yejin Choi, and Swabha Swayamdipta. Understanding dataset difficulty with V-usable information. In Kamalika Chaudhuri, Stefanie Jegelka, Le Song, Csaba Szepesvari, Gang Niu, and Sivan Sabato (eds.), Proceedings of the 39th International Conference on Machine Learning, volume 162 of… find
  16. Rand Fishkin. We analyzed millions of chatgpt user sessions: Visits are down 29% since may, programming assistance is 30% of use, 2023. URL https://sparktoro.com/blog. Accessed: 2023-09-18. ↗
  17. Deep Ganguli, Liane Lovitt, Jackson Kernion, Amanda Askell, Yuntao Bai, Saurav Kadavath, Ben Mann, Ethan Perez, Nicholas Schiefer, Kamal Ndousse, et al. Red teaming language models to reduce harms: Methods, scaling behaviors, and lessons learned. arXiv preprint arXiv:2209.07858, 2022. find our card
  18. Xinyang Geng, Arnav Gudibande, Hao Liu, Eric Wallace, Pieter Abbeel, Sergey Levine, and Dawn Song. Koala: A dialogue model for academic research. Blog post, April 2023. URL https: //bair.berkeley.edu/blog/2023/04/03/koala/. find
  19. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. Measuring massive multitask language understanding. In International Conference on Learning Representations, 2020. find our card
  20. Hyunwoo Kim, Jack Hessel, Liwei Jiang, Peter West, Ximing Lu, Youngjae Yu, Pei Zhou, Ronan Le Bras, Malihe Alikhani, Gunhee Kim, Maarten Sap, and Yejin Choi. Soda: Million-scale dialogue distillation with social commonsense contextualization. ArXiv, abs/2212.10465, 2022. find our card
  21. Andreas Köpf, Yannic Kilcher, Dimitri von Rütte, Sotiris Anagnostidis, Zhi-Rui Tam, Keith Stevens, Abdullah Barhoum, Nguyen Minh Duc, Oliver Stanley, Richárd Nagyfi, et al. Openassistant conversations–democratizing large language model alignment. arXiv preprint arXiv:2304.07327, 2023. find our card
  22. Nathan Lambert, Lewis Tunstall, Nazneen Rajani, and Tristan Thrush. Huggingface h4 stack exchange preference dataset, 2023. URL https://huggingface.co/datasets/HuggingFaceH4/stack-exchange-preferences. ↗
  23. Guohao Li, Hasan Abed Al Kader Hammoud, Hani Itani, Dmitrii Khizbullin, and Bernard Ghanem. Camel: Communicative agents for ”mind” exploration of large scale language model society, 2023. find
  24. Yanran Li, Hui Su, Xiaoyu Shen, Wenjie Li, Ziqiang Cao, and Shuzi Niu. Dailydialog: A manually labelled multi-turn dialogue dataset. In Proceedings of the Eighth International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pp. 986–995, 2017. find our card
  25. Yujia Li, David Choi, Junyoung Chung, Nate Kushman, Julian Schrittwieser, Rémi Leblond, Tom Eccles, James Keeling, Felix Gimeno, Agustin Dal Lago, et al. Competition-level code generation with alphacode. Science, 378(6624):1092–1097, 2022. find our card
  26. Zi Lin, Zihan Wang, Yongqi Tong, Yangkun Wang, Yuxin Guo, Yujia Wang, and Jingbo Shang. Toxicchat: Unveiling hidden challenges of toxicity detection in real-world user-ai conversation, 2023. find
  27. Ryan Lowe, Nissan Pow, Iulian Vlad Serban, and Joelle Pineau. The ubuntu dialogue corpus: A large dataset for research in unstructured multi-turn dialogue systems. In Proceedings of the 16th Annual Meeting of the Special Interest Group on Discourse and Dialogue, pp. 285–294, 2015. find our card
  28. Todor Markov, Chong Zhang, Sandhini Agarwal, Florentine Eloundou Nekoul, Theodore Lee, Steven Adler, Angela Jiang, and Lilian Weng. A holistic approach to undesired content detection in the real world. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pp. 15009–15018, 2023… find our card
  29. OpenAI. Gpt-4 technical report. arXiv preprint arXiv:2303.08774, 2023a. find our card
  30. OpenAI. Openai moderation api, 2023b. URL https://platform.openai.com/docs/guides/moderation. ↗
  31. Hannah Rashkin, Eric Michael Smith, Margaret Li, and Y-Lan Boureau. Towards empathetic open-domain conversation models: A new benchmark and dataset. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 5370–5381, 2019. find
  32. Siva Reddy, Danqi Chen, and Christopher D Manning. Coqa: A conversational question answering challenge. Transactions of the Association for Computational Linguistics, 7:249–266, 2019. find our card
  33. Nils Reimers and Iryna Gurevych. Sentence-bert: Sentence embeddings using siamese bert-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 11 2019. URL https://arxiv. org/abs/1908.10084. find
  34. Paul Röttger, Hannah Rose Kirk, Bertie Vidgen, Giuseppe Attanasio, Federico Bianchi, and Dirk Hovy. Xstest: A test suite for identifying exaggerated safety behaviours in large language models. arXiv preprint arXiv:2308.01263, 2023. find our card
  35. Baptiste Rozière, Jonas Gehring, Fabian Gloeckle, Sten Sootla, Itai Gat, Xiaoqing Ellen Tan, Yossi Adi, Jingyu Liu, Tal Remez, Jérémy Rapin, et al. Code llama: Open foundation models for code. arXiv preprint arXiv:2308.12950, 2023. find our card
  36. Xinyue Shen, Zeyuan Chen, Michael Backes, Yun Shen, and Yang Zhang. “do anything now”: Characterizing and evaluating in-the-wild jailbreak prompts on large language models. arXiv preprint arXiv:2308.03825, 2023. find our card
  37. 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. ↗
  38. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023a. find
  39. 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, 2023b. find
  40. Alexander Wei, Nika Haghtalab, and Jacob Steinhardt. Jailbroken: How does llm safety training fail? arXiv preprint arXiv:2307.02483, 2023. find our card
  41. Can Xu, Qingfeng Sun, Kai Zheng, Xiubo Geng, Pu Zhao, Jiazhan Feng, Chongyang Tao, and Daxin Jiang. Wizardlm: Empowering large language models to follow complex instructions. arXiv preprint arXiv:2304.12244, 2023a. find our card
  42. Canwen Xu, Daya Guo, Nan Duan, and Julian McAuley. Baize: An open-source chat model with parameter-efficient tuning on self-chat data. arXiv preprint arXiv:2304.01196, 2023b. find our card
  43. Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. Personalizing dialogue agents: I have a dog, do you have pets too? In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 2204–2213, 2018. find our card
  44. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, et al. Judging llm-as-a-judge with mt-bench and chatbot arena. arXiv preprint arXiv:2306.05685, 2023. find
  45. Banghua Zhu, Ying Sheng, Lianmin Zheng, Clark Barrett, Michael I Jordan, and Jiantao Jiao. On optimal caching and model multiplexing for large model inference. arXiv preprint arXiv:2306.02003, 2023. — [arXiv:2306.02003](https://arxiv.org/abs/2306.02003) find
  46. Andy Zou, Zifan Wang, J Zico Kolter, and Matt Fredrikson. Universal and transferable adversarial attacks on aligned language models. arXiv preprint arXiv:2307.15043, 2023. find our card