Large Scale Knowledge Washing
Abstract: LLMs show impressive abilities in memorizing world knowledge, which leads to concerns regarding memorization of private information, toxic or sensitive knowledge, and copyrighted content. We introduce the problem of Large Scale Knowledge Washing, focusing on unlearning an extensive amount of factual knowledge. Previous unlearning methods usually define the reverse loss and update the model via backpropagation, which may affect the model's fluency and reasoning ability or even destroy the model due to extensive training with the reverse loss. Existing works introduce additional data from downstream tasks to prevent the model from losing capabilities, which requires downstream task awareness. Controlling the tradeoff of unlearning and maintaining existing capabilities is also challenging. To this end, we propose LAW (Large Scale Washing) to update the MLP layers in decoder-only LLMs to perform knowledge washing, as inspired by model editing methods and based on the hypothesis that knowledge and reasoning are disentanglable. We derive a new objective with the knowledge to be unlearned to update the weights of certain MLP layers. Experimental results demonstrate the effectiveness of LAW in forgetting target knowledge while maintaining reasoning ability. The code will be open-sourced at https://github.com/wangyu-ustc/LargeScaleWashing.
- A. Act. Health insurance portability and accountability act of 1996. Public law, 104:191, 1996.
- Prompting as probing: Using language models for knowledge base construction. In LM-KBC@ISWC, volume 3274 of CEUR Workshop Proceedings, pages 11–34. CEUR-WS.org, 2022.
- Longbench: A bilingual, multitask benchmark for long context understanding. arXiv preprint arXiv:2308.14508, 2023.
- Gpt-neo: Large scale autoregressive language modeling with mesh-tensorflow. If you use this software, please cite it using these metadata, 58:2, 2021.
- J. Chen and D. Yang. Unlearn what you want to forget: Efficient unlearning for llms. arXiv preprint arXiv:2310.20150, 2023.
- Probing simile knowledge from pre-trained language models. In ACL (1), pages 5875–5887. Association for Computational Linguistics, 2022.
- Think you have solved question answering? try arc, the ai2 reasoning challenge, 2018.
- R. Eldan and M. Russinovich. Who’s harry potter? approximate unlearning in llms. arXiv preprint arXiv:2310.02238, 2023.
- G. Gangadhar and K. Stratos. Model editing by pure fine-tuning. CoRR, abs/2402.11078, 2024.
- A framework for few-shot language model evaluation, 12 2023. URL https://zenodo.org/records/10256836.
- Transformer feed-forward layers are key-value memories. arXiv preprint arXiv:2012.14913, 2020.
- Aging with grace: Lifelong model editing with discrete key-value adaptors. arXiv preprint arXiv:2211.11031, 2022.
- Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021.
- Transformer-patcher: One mistake worth one neuron. arXiv preprint arXiv:2301.09785, 2023.
- M. Isonuma and I. Titov. Unlearning reveals the influential training data of language models. arXiv preprint arXiv:2401.15241, 2024.
- C. S. Legislature. California consumer privacy act (CCPA). "https://oag.ca.gov/privacy/ccpa", 2018.
- Zero-shot relation extraction via reading comprehension. arXiv preprint arXiv:1706.04115, 2017.
- Rethinking machine unlearning for large language models. arXiv preprint arXiv:2402.08787, 2024a.
- Towards safer large language models through machine unlearning. arXiv preprint arXiv:2402.10058, 2024b.
- Quark: Controllable text generation with reinforced unlearning. Advances in neural information processing systems, 35:27591–27609, 2022.
- Locating and editing factual associations in gpt. Advances in Neural Information Processing Systems, 35:17359–17372, 2022.
- Mass-editing memory in a transformer. In ICLR. OpenReview.net, 2023.
- Fast model editing at scale. In ICLR. OpenReview.net, 2022a.
- Memory-based model editing at scale. In ICML, volume 162 of Proceedings of Machine Learning Research, pages 15817–15831. PMLR, 2022b.
- The lambada dataset, Aug 2016. URL https://doi.org/10.5281/zenodo.2630551.
- E. Parliament and C. of the European Union. General data protection regulation (GDPR), 2016.
- In-context unlearning: Language models as few shot unlearners. arXiv preprint arXiv:2310.07579, 2023.
- Language models are unsupervised multitask learners. 2019.
- Knowledge unlearning for llms: Tasks, methods, and challenges. arXiv preprint arXiv:2311.15766, 2023.
- Selective forgetting: Advancing machine unlearning techniques and evaluation in language models. arXiv preprint arXiv:2402.05813, 2024.
- Machine unlearning of pre-trained large language models. arXiv preprint arXiv:2402.15159, 2024.
- Editing large language models: Problems, methods, and opportunities. CoRR, abs/2305.13172, 2023a.
- Large language model unlearning. arXiv preprint arXiv:2310.10683, 2023b.
- Give me the facts! A survey on factual knowledge probing in pre-trained language models. In EMNLP (Findings), pages 15588–15605. Association for Computational Linguistics, 2023.
- Hellaswag: Can a machine really finish your sentence? In ACL (1), pages 4791–4800. Association for Computational Linguistics, 2019.
- Right to be forgotten in the era of large language models: Implications, challenges, and solutions. arXiv preprint arXiv:2307.03941, 2023.
- Deciphering the lmpact of pretraining data on large language models through machine unlearning. arXiv preprint arXiv:2402.11537, 2024.
- Can we edit factual knowledge by in-context learning? arXiv preprint arXiv:2305.12740, 2023.
- Modifying memories in transformer models. CoRR, abs/2012.00363, 2020.
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.