Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat
Sign Up
Cited by 16

The DOI matches its Crossref record; 9 of the 9 references it cites were found in a registry or an index.

How to Protect Copyright Data in Optimization of Large Language Models?

Chu, Timothy, Song, Zhao, Yang, Chiwun·2024·Proceedings of the AAAI Conference on Artificial Intelligence
DOI 10.1609/aaai.v38i16.29741
Open this paper in CoChat — free account

Abstract

Large language models (LLMs) and generative AI have played a transformative role in computer research and applications. Controversy has arisen as to whether these models output copyrighted data, which can occur if the data the models are trained on is copyrighted. LLMs are built on the transformer neural network architecture, which in turn relies on a mathematical computation called Attention that uses the softmax function. In this paper, we observe that large language model training and optimization can be seen as a softmax regression problem. We then establish a method of efficiently performing softmax regression, in a way that prevents the regression function from generating copyright data. This establishes a theoretical method of training large language models in a way that avoids generating copyright data.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1609/aaai.v38i16.29741

Cites

9 of 9 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 9 references

Research works printed but not linked: 66

  1. Allen-Zhu, Z.; Li, Y.; and Song, Z. 2019a. A convergence theory for deep learning via over-parameterization. In ICML.
  2. Allen-Zhu, Z.; Li, Y.; and Song, Z. 2019b. On the convergence rate of training recurrent neural networks. NeurIPS, 32.
  3. Alman, J.; Liang, J.; Song, Z.; Zhang, R.; and Zhuo, D. 2023. By- pass exponential time preprocessing: Fast neural network training via weight-data correlation preprocessing. In NeurIPS.
  4. Alman, J.; and Song, Z. 2023. Fast Attention Requires Bounded Entries. arXiv preprint arXiv:2302.13214. · arXiv
  5. Anil, R.; Dai, A. M.; Firat, O.; Johnson, M.; Lepikhin, D.; Passos, A.; Shakeri, S.; Taropa, E.; Bailey, P.; Chen, Z.; et al. 2023. Palm 2 technical report. arXiv preprint arXiv:2305.10403. · arXiv
  6. Bao, F.; Nie, S.; Xue, K.; Cao, Y.; Li, C.; Su, H.; and Zhu, J. 2023. All are worth words: A vit backbone for diffusion models. In CVPR, 22669-22679.
  7. Brand, J. v. d.; Peng, B.; Song, Z.; and Weinstein, O. 2020. Train- ing (overparametrized) neural networks in near-linear time. arXiv preprint arXiv:2006.11648. · arXiv
  8. Brand, J. v. d.; Song, Z.; and Zhou, T. 2023. Algorithm and Hard- ness for Dynamic Attention Maintenance in Large Language Mod- els. arXiv preprint arXiv:2304.02207. · arXiv
  9. Bubeck, S.; Chandrasekaran, V.; Eldan, R.; Gehrke, J.; Horvitz, E.; Kamar, E.; Lee, P.; Lee, Y. T.; Li, Y.; Lundberg, S.; et al. 2023. Sparks of artificial general intelligence: Early experiments with gpt-4. arXiv preprint arXiv:2303.12712. · arXiv
  10. Cai, T.; Gao, R.; Hou, J.; Chen, S.; Wang, D.; He, D.; Zhang, Z.; and Wang, L. 2019. Gram-gauss-newton method: Learning overparameterized neural networks for regression problems. arXiv preprint arXiv:1905.11675. · arXiv
  11. Cao, H.; Wang, J.; Ren, T.; Qi, X.; Chen, Y.; Yao, Y.; and Zhang, L. 2022. Exploring vision transformers as diffusion learners. arXiv preprint arXiv:2212.13771. · arXiv
  12. Cao, Y.; and Gu, Q. 2019. Generalization bounds of stochastic gradient descent for wide and deep neural networks. NeurIPS, 32.
  13. Chen, B.; Dao, T.; Winsor, E.; Song, Z.; Rudra, A.; and Ré, C. 2021. Scatterbrain: Unifying sparse and low-rank attention. NeurIPS, 34: 17413-17426.
  14. Chu, T.; Song, Z.; and Yang, C. 2023a. Fine-tune language mod- els to approximate unbiased in-context learning. arXiv preprint arXiv:2310.03331. · arXiv
  15. Chu, T.; Song, Z.; and Yang, C. 2023b. How to Protect Copyright Data in Optimization of Large Language Models? arXiv preprint arxiv:2308.12247. · arXiv
  16. Deng, Y.; Li, Z.; Mahadevan, S.; and Song, Z. 2023. Zero-th Or- der Algorithm for Softmax Attention Optimization. arXiv preprint arXiv:2307.08352. · arXiv
  17. Deng, Y.; Li, Z.; and Song, Z. 2023. Attention scheme inspired softmax regression. arXiv preprint arXiv:2304.10411. · arXiv
  18. Du, S. S.; Zhai, X.; Poczos, B.; and Singh, A. 2018. Gradient descent provably optimizes over-parameterized neural networks. arXiv preprint arXiv:1810.02054. for the 9th circuits, U. S. C. 2022. Copying-Access and Sub- stantial Similarity. https://www.ce9.uscourts.gov/jury-instructions/ node/274/. Accessed: 2022-12-31. · arXiv
  19. Gao, Y.; Mahadevan, S.; and Song, Z. 2023. An over-parameterized exponential regression. arXiv preprint arXiv:2303.16504. · arXiv
  20. Gao, Y.; Song, Z.; and Yang, X. 2023. Differentially private atten- tion computation. arXiv preprint arXiv:2305.04701. · arXiv
  21. Gao, Y.; Song, Z.; Yang, X.; and Zhang, R. 2023. Fast Quantum Algorithm for Attention Computation. arXiv preprint arXiv:2307.08045. · arXiv
  22. Gao, Y.; Song, Z.; and Yin, J. 2023. GradientCoin: A Peer- to-Peer Decentralized Large Language Models. arXiv preprint arXiv:2308.10502. · arXiv
  23. Han, K.; Wang, Y.; Chen, H.; Chen, X.; Guo, J.; Liu, Z.; Tang, Y.; Xiao, A.; Xu, C.; Xu, Y.; et al. 2022. A survey on vision trans- former. PAMI, 45(1): 87-110.
  24. He, X.; Xu, Q.; Lyu, L.; Wu, F.; and Wang, C. 2022a. Protect- ing intellectual property of language generation apis with lexical watermark. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, 10758-10766.
  25. He, X.; Xu, Q.; Zeng, Y.; Lyu, L.; Wu, F.; Li, J.; and Jia, R. 2022b. Cater: Intellectual property protection on text generation apis via conditional watermarks. NeurIPS, 35: 5431-5445.
  26. Huang, B.; Li, X.; Song, Z.; and Yang, X. 2021. Fl-ntk: A neural tangent kernel-based framework for federated learning analysis. In ICML.
  27. Ignat, O.; Jin, Z.; Abzaliev, A.; Biester, L.; Castro, S.; Deng, N.; Gao, X.; Gunal, A.; He, J.; Kazemi, A.; et al. 2023. A PhD Stu- dent's Perspective on Research in NLP in the Era of Very Large Language Models. arXiv preprint arXiv:2305.12544. · arXiv
  28. Ji, Z.; and Telgarsky, M. 2019. Polylogarithmic width suffices for gradient descent to achieve arbitrarily small test error with shallow relu networks. arXiv preprint arXiv:1909.12292. · arXiv
  29. Jiang, D.; Ren, X.; and Lin, B. Y. 2023. LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fu- sion. arXiv preprint arXiv:2306.02561. · arXiv
  30. Kitaev, N.; Kaiser, Ł.; and Levskaya, A. 2020. Reformer: The effi- cient transformer. arXiv preprint arXiv:2001.04451. · arXiv
  31. Lee, J. D.; Shen, R.; Song, Z.; Wang, M.; and Yu, Z. 2020. Gener- alized leverage score sampling for neural networks. NeurIPS.
  32. Li, Y.; and Liang, Y. 2018. Learning overparameterized neural net- works via stochastic gradient descent on structured data. NeurIPS.
  33. Li, Z.; Song, Z.; and Zhou, T. 2023. Solving regularized exp, cosh and sinh regression problems. arXiv preprint arXiv:2303.15725. · arXiv
  34. Liu, H.; Li, Z.; Hall, D.; Liang, P.; and Ma, T. 2023a. Sophia: A Scalable Stochastic Second-order Optimizer for Language Model Pre-training. arXiv preprint arXiv:2305.14342. · arXiv
  35. Liu, Z.; Wang, J.; Dao, T.; Zhou, T.; Yuan, B.; Song, Z.; Shrivas- tava, A.; Zhang, C.; Tian, Y.; Re, C.; et al. 2023b. Deja vu: Con- textual sparsity for efficient llms at inference time. In ICML.
  36. Malladi, S.; Gao, T.; Nichani, E.; Damian, A.; Lee, J. D.; Chen, D.; and Arora, S. 2023. Fine-Tuning Language Models with Just Forward Passes. arXiv preprint arXiv:2305.17333. · arXiv
  37. Merity, S.; Xiong, C.; Bradbury, J.; and Socher, R. 2016. Pointer sentinel mixture models. arXiv preprint arXiv:1609.07843. · arXiv
  38. Munteanu, A.; Omlor, S.; Song, Z.; and Woodruff, D. 2022. Bounding the width of neural networks via coupled initialization a worst case analysis. In ICML, 16083-16122.
  39. Noci, L.; Anagnostidis, S.; Biggio, L.; Orvieto, A.; Singh, S. P.; and Lucchi, A. 2022. Signal propagation in transformers: Theoretical perspectives and the role of rank collapse. NeurIPS, 35.
  40. Oymak, S.; and Soltanolkotabi, M. 2020. Toward moderate overpa- rameterization: Global convergence guarantees for training shallow neural networks. IEEE Journal on Selected Areas in Information Theory, 1(1): 84-105.
  41. Panigrahi, A.; Malladi, S.; Xia, M.; and Arora, S. 2023. Trainable Transformer in Transformer. arXiv preprint arXiv:2307.01189. · arXiv
  42. Qin, L.; Song, Z.; and Yang, Y. 2023. Efficient SGD Neural Net- work Training via Sublinear Activated Neuron Identification. arXiv preprint arXiv:2307.06565. · arXiv
  43. Rücklé, A.; Geigle, G.; Glockner, M.; Beck, T.; Pfeiffer, J.; Reimers, N.; and Gurevych, I. 2020. Adapterdrop: On the efficiency of adapters in transformers. arXiv preprint arXiv:2010.11918. · arXiv
  44. Sanford, C.; Hsu, D.; and Telgarsky, M. 2023. Representa- tional Strengths and Limitations of Transformers. arXiv preprint arXiv:2306.02896. · arXiv
  45. Song, Z.; and Yang, X. 2019. Quadratic suffices for over- parametrization via matrix chernoff bound. arXiv preprint arXiv:1906.03593. · arXiv
  46. Song, Z.; Zhang, L.; and Zhang, R. 2021. Training multi-layer over-parametrized neural network in subquadratic time. arXiv preprint arXiv:2112.07628. · arXiv
  47. Sun, A. Y.; Zemour, E.; Saxena, A.; Vaidyanathan, U.; Lin, E.; Lau, C.; and Mugunthan, V. 2023. Does fine-tuning GPT-3 with the OpenAI API leak personally-identifiable information? arXiv preprint arXiv:2307.16382. · arXiv
  48. Tay, Y.; Bahri, D.; Metzler, D.; Juan, D.-C.; Zhao, Z.; and Zheng, C. 2021. Synthesizer: Rethinking self-attention for transformer mod- els. In ICML, 10183-10192.
  49. PMLR.
  50. Tay, Y.; Dehghani, M.; Abnar, S.; Shen, Y.; Bahri, D.; Pham, P.; Rao, J.; Yang, L.; Ruder, S.; and Metzler, D. 2020. Long range arena: A benchmark for efficient transformers. arXiv preprint arXiv:2011.04006. · arXiv
  51. Touvron, H.; Martin, L.; Stone, K.; Albert, P.; Almahairi, A.; Babaei, Y.; Bashlykov, N.; Batra, S.; Bhargava, P.; Bhosale, S.; et al. 2023b. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288. · arXiv
  52. Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, Ł.; and Polosukhin, I. 2017. Attention is all you need. NeurIPS, 30.
  53. Vyas, N.; Kakade, S.; and Barak, B. 2023. Provable copyright pro- tection for generative models. arXiv preprint arXiv:2302.10870. · arXiv
  54. Wei, J.; Wang, X.; Schuurmans, D.; Bosma, M.; Xia, F.; Chi, E.; Le, Q. V.; Zhou, D.; et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. NeurIPS, 35: 24824-24837.
  55. Wu, J.; Fu, R.; Fang, H.; Zhang, Y.; and Xu, Y. 2023a. Medsegdiff- v2: Diffusion based medical image segmentation with transformer. arXiv preprint arXiv:2301.11798. · arXiv
  56. Wu, J.; Yu, T.; Wang, R.; Song, Z.; Zhang, R.; Zhao, H.; Lu, C.; Li, S.; and Henao, R. 2023b. InfoPrompt: Information-Theoretic Soft Prompt Tuning for Natural Language Understanding. In NeurIPS.
  57. Wu, Y.; Schuster, M.; Chen, Z.; Le, Q. V.; Norouzi, M.; Macherey, W.; Krikun, M.; Cao, Y.; Gao, Q.; Macherey, K.; et al. 2016. Google's neural machine translation system: Bridging the gap between human and machine translation. arXiv preprint arXiv:1609.08144. · arXiv
  58. Xu, Z.; Zhang, Y.; Andrew, G.; Choquette-Choo, C. A.; Kairouz, P.; McMahan, H. B.; Rosenstock, J.; and Zhang, Y. 2023. Federated Learning of Gboard Language Models with Differential Privacy. arXiv preprint arXiv:2305.18465. · arXiv
  59. Zelikman, E.; Huang, Q.; Liang, P.; Haber, N.; and Goodman, N. D. 2023. Just One Byte (per gradient): A Note on Low-Bandwidth Decentralized Language Model Finetuning Using Shared Random- ness. arXiv preprint arXiv:2306.10015. · arXiv
  60. Zhang, G.; Martens, J.; and Grosse, R. B. 2019. Fast convergence of natural gradient descent for over-parameterized neural networks. NeurIPS, 32.
  61. Zhang, J.; Karimireddy, S. P.; Veit, A.; Kim, S.; Reddi, S.; Kumar, S.; and Sra, S. 2020a. Why are adaptive methods good for attention models? NeurIPS, 33: 15383-15393.
  62. Zhang, L. 2022. Speeding up optimizations via data structures: Faster search, sample and maintenance. Ph.D. thesis, Master's the- sis, Carnegie Mellon University.
  63. Zhang, S.; Roller, S.; Goyal, N.; Artetxe, M.; Chen, M.; Chen, S.; Dewan, C.; Diab, M.; Li, X.; Lin, X. V.; et al. 2022. Opt: Open pre-trained transformer language models. arXiv preprint arXiv:2205.01068. · arXiv
  64. Zhang, Y.; Plevrakis, O.; Du, S. S.; Li, X.; Song, Z.; and Arora, S. 2020b. Over-parameterized adversarial training: An analysis over- coming the curse of dimensionality. NeurIPS, 33: 679-688.
  65. Zhao, H.; Panigrahi, A.; Ge, R.; and Arora, S. 2023. Do Trans- formers Parse while Predicting the Masked Word? arXiv preprint arXiv:2303.08117. · arXiv
  66. Zou, D.; and Gu, Q. 2019. An improved analysis of training over- parameterized deep neural networks. NeurIPS, 32.

Cited by

See all 16

Metadata

Authors
Chu, T., Song, Z., & Yang, C.
Venue
Proceedings of the AAAI Conference on Artificial Intelligence
Publisher
Association for the Advancement of Artificial Intelligence (AAAI)
Published
2024-03-24
Volume
38
Issue
16
Pages
17871-17879
ISSN
  • 2374-3468
  • 2159-5399
DOI
10.1609/aaai.v38i16.29741
Cited by
16
References
9

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: How to Protect Copyright Data in Optimization of Large Language Models? https://app.cochat.ai/papers/10.1609/aaai.v38i16.29741
The paper
Chu, T., Song, Z., & Yang, C. (2024). How to Protect Copyright Data in Optimization of Large Language Models? Proceedings of the AAAI Conference on Artificial Intelligence. https://doi.org/10.1609/aaai.v38i16.29741