---
title: "RKLD — paper by Wang et al. (2024) — summary and ideas"
description: "The authors propose RKLD, a two-stage method. 39 references, cited by 12 papers here."
authors: ["Bichen Wang", "Yuzhe Zi", "Yixin Sun", "Yanyan Zhao", "Bing Qin"]
year: 2024
venue: "arXiv:2406.01983"
source: "https://arxiv.org/abs/2406.01983"
doi: "10.48550/arxiv.2406.01983"
arxiv: "2406.01983"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/wang-2024-rkld-reverse-kl-divergence-based/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/wang-2024-rkld-reverse-kl-divergence-based/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2024](https://md.research.elamur.ai/free/en/papers/2024/page/9/) / RKLD: Reverse KL-Divergence-based Knowledge Distillation for Unlearning Personal Information in Large Language Models

# RKLD: Reverse KL-Divergence-based Knowledge Distillation for Unlearning Personal Information in Large Language Models

**Authors:** [Bichen Wang](https://md.research.elamur.ai/free/en/authors/bichen-wang/), Yuzhe Zi, [Yixin Sun](https://md.research.elamur.ai/free/en/authors/yixin-sun/), [Yanyan Zhao](https://md.research.elamur.ai/free/en/authors/yanyan-zhao/) and [Bing Qin](https://md.research.elamur.ai/free/en/authors/bing-qin/) · **Year:** 2024 · **Venue:** arXiv:2406.01983 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2406.01983) · [PDF](https://arxiv.org/pdf/2406.01983) · [DOI 10.48550/arxiv.2406.01983](https://doi.org/10.48550/arxiv.2406.01983) · [arXiv 2406.01983](https://arxiv.org/abs/2406.01983) · [OpenAlex](https://openalex.org/W4399413134)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · [12](https://md.research.elamur.ai/free/en/papers/wang-2024-rkld-reverse-kl-divergence-based/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Large language models can memorize personal details from their training data, and right-to-be-forgotten rules such as the GDPR require that this data be removable. Current unlearning methods based on gradient ascent lower the likelihood of target answers but also distort unrelated token probabilities. As a result, models either lose fluency or forget only partially.

The authors propose RKLD, a two-stage method. First, the original model is fine-tuned further on the forget set. Tokens whose logits grow during this step are treated as linked to the information to erase. A teacher is then built by subtracting a scaled, ReLU-clipped version of that growth from the original logits. Second, the original model is distilled toward this teacher on forget-set prefixes using reverse KL divergence, which penalizes probability mass that the teacher does not support.

On the TOFU benchmark of fictitious author profiles, RKLD outperforms the baselines in forget quality as well as model utility, with the advantage most pronounced when the forget set is larger, and it retains general abilities. This offers a practical route to privacy-compliant LLMs without full retraining.

## Key points
- **Setting:** TOFU contains 200 synthetic people with 20 QA pairs each, 4,000 in total. A subset forms the forget set and the remainder is the retain set. After unlearning, the model should behave as if it had never seen the forgotten data.
- **Teacher construction:** l_tea = l_ori − α·ReLU(l_str − l_ori), where α sets forgetting strength. Used directly as an unlearned model, the teacher performs poorly: it is weak on paraphrased queries and needs two models at inference time. It therefore serves only as a guide.
- **Reverse vs. forward KL:** The ablation favors reverse KL. A forward-KL variant (FKLD) yields low ROUGE-L, yet it still assigns high probability to the private answers, so its forget quality is poor.
- **Case study:** Baselines are GA, IDK, NPO and TA. In a Forget05 example (Hina Ameen, born in Karachi), the top-5 fill-in-the-blank completions of most methods still reveal the true city. Only RKLD and GA avoid this leak, and GA severely damages fluency. The IDK approach leaks especially easily.
- **Limitations:** Experiments use only controlled synthetic data. Behavior after unlearning remains uncertain, including possible hallucinations. Current metrics depend on comparison with a retrained model, which is not available in real deployments.

*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

- [Harbin Institute of Technology](https://md.research.elamur.ai/free/en/universities/harbin-institute-of-technology/) — University, Harbin, China · 158 works · 28,841 citations

📚 Shelves (2): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works

🧭 Topics (2): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works · [🪶 Efficient models & compression](https://md.research.elamur.ai/free/en/topics/efficient-models/) · 242 works

## Cite

Cite as: Bichen Wang, Yuzhe Zi, Yixin Sun, Yanyan Zhao and Bing Qin. 2024. RKLD: Reverse KL-Divergence-based Knowledge Distillation for Unlearning Personal Information in Large Language Models. arXiv:2406.01983.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{wang2024rkld,
  title = {{RKLD: Reverse KL-Divergence-based Knowledge Distillation for Unlearning Personal Information in Large Language Models}},
  author = {Bichen Wang, Yuzhe Zi, Yixin Sun, Yanyan Zhao and Bing Qin},
  year = {2024},
  doi = {10.48550/arxiv.2406.01983},
  eprint = {2406.01983},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2406.01983}
}
```

- [← previous Psychometric Alignment: Capturing Human Knowledge Distributions via Language Models](https://md.research.elamur.ai/free/en/papers/he-yueya-2024-psychometric-alignment-capturing-human-knowledge/)
- [next → Synthetic Arabic Medical Dialogues Using Advanced Multi-Agent LLM Techniques](https://md.research.elamur.ai/free/en/papers/almutairi-2024-synthetic-arabic-medical-dialogues-advanced/)
- [↑ back to papers of 2024](https://md.research.elamur.ai/free/en/papers/2024/page/9/)

## References

39 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. George-Octavian Barbulescu and Peter Triantafillou. 2024. To each (textual sequence) its own: Improving memorized-data unlearning in large language models. arXiv preprint arXiv:2405.03097. — [arXiv:2405.03097](https://arxiv.org/abs/2405.03097) [find](https://scholar.google.com/scholar?q=George%2DOctavian+Barbulescu+and+Peter+Triantafillou%2E+2024%2E+To+each+%28textual+sequence%29+its+own%3A+Improving+memorized%2Ddata+unlearning+in+large+language+models%2E+arXiv+preprint+arXiv%3A2405%2E03097%2E+%E2%80%94+%5BarXiv%3A24)
2. Lucas Bourtoule, Varun Chandrasekaran, Christopher A Choquette-Choo, Hengrui Jia, Adelin Travers, Baiwu Zhang, David Lie, and Nicolas Papernot. 2021. Machine unlearning. In 2021 IEEE Symposium on Security and Privacy (SP), pages 141–159. IEEE. [find](https://scholar.google.com/scholar?q=Lucas+Bourtoule%2C+Varun+Chandrasekaran%2C+Christopher+A+Choquette%2DChoo%2C+Hengrui+Jia%2C+Adelin+Travers%2C+Baiwu+Zhang%2C+David+Lie%2C+and+Nicolas+Papernot%2E+2021%2E+Machine+unlearning%2E+In+2021+IEEE+Symposium+on+Secu)
3. Nicholas Carlini, Daphne Ippolito, Matthew Jagielski, Katherine Lee, Florian Tramer, and Chiyuan Zhang. 2022. Quantifying memorization across neural language models. arXiv preprint arXiv:2202.07646. [find](https://scholar.google.com/scholar?q=Nicholas+Carlini%2C+Daphne+Ippolito%2C+Matthew+Jagielski%2C+Katherine+Lee%2C+Florian+Tramer%2C+and+Chiyuan+Zhang%2E+2022%2E+Quantifying+memorization+across+neural+language+models%2E+arXiv+preprint+arXiv%3A2202%2E07646%2E)
4. Nicholas Carlini, Florian Tramer, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Ulfar Erlingsson, et al. 2021. [Extracting training data from large language models](https://md.research.elamur.ai/free/en/papers/carlini-2020-extracting-training-data-large-language/). In 30th USENIX Security Symposium (USENIX Security 21), pages 2633–2650.
5. Jiaao Chen and Diyi Yang. 2023. [Unlearn what you want to forget: Efficient unlearning for llms](https://md.research.elamur.ai/free/en/papers/chen-2023-unlearn-what-you-want-forget/). In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 12041– 12052.
6. Eli Chien, Chao Pan, and Olgica Milenkovic. 2022. Efficient model updates for approximate unlearning of graph-structured data. In The Eleventh International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Eli+Chien%2C+Chao+Pan%2C+and+Olgica+Milenkovic%2E+2022%2E+Efficient+model+updates+for+approximate+unlearning+of+graph%2Dstructured+data%2E+In+The+Eleventh+International+Conference+on+Learning+Representations%2E)
7. Ronen Eldan and Mark Russinovich. 2023. Who’s harry potter? approximate unlearning in llms. arXiv preprint arXiv:2310.02238. [find](https://scholar.google.com/scholar?q=Ronen+Eldan+and+Mark+Russinovich%2E+2023%2E+Who%E2%80%99s+harry+potter%3F+approximate+unlearning+in+llms%2E+arXiv+preprint+arXiv%3A2310%2E02238%2E)
8. Kawin Ethayarajh, Winnie Xu, Niklas Muennighoff, Dan Jurafsky, and Douwe Kiela. 2024. Kto: Model alignment as prospect theoretic optimization. arXiv e-prints, pages arXiv–2402. [find](https://scholar.google.com/scholar?q=Kawin+Ethayarajh%2C+Winnie+Xu%2C+Niklas+Muennighoff%2C+Dan+Jurafsky%2C+and+Douwe+Kiela%2E+2024%2E+Kto%3A+Model+alignment+as+prospect+theoretic+optimization%2E+arXiv+e%2Dprints%2C+pages+arXiv%E2%80%932402%2E)
9. Rohit Gandikota, Joanna Materzynska, Jaden Fiotto-Kaufman, and David Bau. 2023. Erasing concepts from diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 2426–2436. [find](https://scholar.google.com/scholar?q=Rohit+Gandikota%2C+Joanna+Materzynska%2C+Jaden+Fiotto%2DKaufman%2C+and+David+Bau%2E+2023%2E+Erasing+concepts+from+diffusion+models%2E+In+Proceedings+of+the+IEEE%2FCVF+International+Conference+on+Computer+Vision%2C+page)
10. Aditya Golatkar, Alessandro Achille, and Stefano Soatto. 2020. Eternal sunshine of the spotless net: Selective forgetting in deep networks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9304–9312. [find](https://scholar.google.com/scholar?q=Aditya+Golatkar%2C+Alessandro+Achille%2C+and+Stefano+Soatto%2E+2020%2E+Eternal+sunshine+of+the+spotless+net%3A+Selective+forgetting+in+deep+networks%2E+In+Proceedings+of+the+IEEE%2FCVF+Conference+on+Computer+Vision)
11. Anisa Halimi, Swanand Kadhe, Ambrish Rawat, and Nathalie Baracaldo. 2022. Federated unlearning: How to efficiently erase a client in fl? arXiv preprint arXiv:2207.05521. — [arXiv:2207.05521](https://arxiv.org/abs/2207.05521) [find](https://scholar.google.com/scholar?q=Anisa+Halimi%2C+Swanand+Kadhe%2C+Ambrish+Rawat%2C+and+Nathalie+Baracaldo%2E+2022%2E+Federated+unlearning%3A+How+to+efficiently+erase+a+client+in+fl%3F+arXiv+preprint+arXiv%3A2207%2E05521%2E+%E2%80%94+%5BarXiv%3A2207%2E05521%5D%28https%3A%2F%2Fa)
12. Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. 2015. [Distilling the knowledge in a neural network](https://md.research.elamur.ai/free/en/papers/hinton-2015-distilling-knowledge-neural-network/). arXiv preprint arXiv:1503.02531.
13. Gabriel Ilharco, Marco Tulio Ribeiro, Mitchell Wortsman, Ludwig Schmidt, Hannaneh Hajishirzi, and Ali Farhadi. 2022. [Editing models with task arithmetic](https://md.research.elamur.ai/free/en/papers/ilharco-2022-editing-models-task-arithmetic/). In The Eleventh International Conference on Learning Representations.
14. Joel Jang, Dongkeun Yoon, Sohee Yang, Sungmin Cha, Moontae Lee, Lajanugen Logeswaran, and Minjoon Seo. 2022. [Knowledge unlearning for mitigating privacy risks in language models](https://md.research.elamur.ai/free/en/papers/jang-2022-knowledge-unlearning-mitigating-privacy-risks/). arXiv preprint arXiv:2210.01504.
15. Joel Jang, Dongkeun Yoon, Sohee Yang, Sungmin Cha, Moontae Lee, Lajanugen Logeswaran, and Minjoon Seo. 2023. [Knowledge unlearning for mitigating privacy risks in language models](https://md.research.elamur.ai/free/en/papers/jang-2022-knowledge-unlearning-mitigating-privacy-risks/). In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages…
16. Piji Li. 2022. uChecker: Masked pretrained language models as unsupervised Chinese spelling checkers. In Proceedings of the 29th International Conference on Computational Linguistics, pages 2812–2822, Gyeongju, Republic of Korea. International Committee on Computational Linguistics. [find](https://scholar.google.com/scholar?q=Piji+Li%2E+2022%2E+uChecker%3A+Masked+pretrained+language+models+as+unsupervised+Chinese+spelling+checkers%2E+In+Proceedings+of+the+29th+International+Conference+on+Computational+Linguistics%2C+pages+2812%E2%80%932822%2C)
17. Sijia Liu, Yuanshun Yao, Jinghan Jia, Stephen Casper, Nathalie Baracaldo, Peter Hase, Xiaojun Xu, Yuguang Yao, Hang Li, Kush R Varshney, et al. 2024. Rethinking machine unlearning for large language models. arXiv preprint arXiv:2402.08787. [find](https://scholar.google.com/scholar?q=Sijia+Liu%2C+Yuanshun+Yao%2C+Jinghan+Jia%2C+Stephen+Casper%2C+Nathalie+Baracaldo%2C+Peter+Hase%2C+Xiaojun+Xu%2C+Yuguang+Yao%2C+Hang+Li%2C+Kush+R+Varshney%2C+et+al%2E+2024%2E+Rethinking+machine+unlearning+for+large+language+m)
18. Ziyao Liu, Yu Jiang, Jiyuan Shen, Minyi Peng, Kwok-Yan Lam, and Xingliang Yuan. 2023. A survey on federated unlearning: Challenges, methods, and future directions. arXiv preprint arXiv:2310.20448. — [arXiv:2310.20448](https://arxiv.org/abs/2310.20448) [find](https://scholar.google.com/scholar?q=Ziyao+Liu%2C+Yu+Jiang%2C+Jiyuan+Shen%2C+Minyi+Peng%2C+Kwok%2DYan+Lam%2C+and+Xingliang+Yuan%2E+2023%2E+A+survey+on+federated+unlearning%3A+Challenges%2C+methods%2C+and+future+directions%2E+arXiv+preprint+arXiv%3A2310%2E20448%2E+%E2%80%94+%5B)
19. Ximing Lu, Sean Welleck, Jack Hessel, Liwei Jiang, Lianhui Qin, Peter West, Prithviraj Ammanabrolu, and Yejin Choi. 2022. [Quark: Controllable text generation with reinforced unlearning](https://md.research.elamur.ai/free/en/papers/lu-2022-quark-controllable-text-generation-reinforced/). Advances in neural information processing systems, 35:27591– 27609.
20. Pratyush Maini, Zhili Feng, Avi Schwarzschild, Zachary C Lipton, and J Zico Kolter. 2024. [Tofu: A task of fictitious unlearning for llms](https://md.research.elamur.ai/free/en/papers/maini-2024-tofu-task-fictitious-unlearning-llms/). arXiv preprint arXiv:2401.06121.
21. Pratyush Maini, Mohammad Yaghini, and Nicolas Papernot. 2020. Dataset inference: Ownership resolution in machine learning. In International Conference on Learning Representations. [find](https://scholar.google.com/scholar?q=Pratyush+Maini%2C+Mohammad+Yaghini%2C+and+Nicolas+Papernot%2E+2020%2E+Dataset+inference%3A+Ownership+resolution+in+machine+learning%2E+In+International+Conference+on+Learning+Representations%2E)
22. Jordan Meadows, Zili Zhou, and André Freitas. 2022. PhysNLU: A language resource for evaluating natural language understanding and explanation coherence in physics. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 4611–4619, Marseille, France. European Language Re… [find](https://scholar.google.com/scholar?q=Jordan+Meadows%2C+Zili+Zhou%2C+and+Andr%C3%A9+Freitas%2E+2022%2E+PhysNLU%3A+A+language+resource+for+evaluating+natural+language+understanding+and+explanation+coherence+in+physics%2E+In+Proceedings+of+the+Thirteenth+La)
23. Seth Neel and Peter Chang. 2023. Privacy issues in [large language models: A survey](https://md.research.elamur.ai/free/en/papers/minaee-2024-large-language-models-survey/). arXiv preprint arXiv:2312.06717.
24. Martin Pawelczyk, Seth Neel, and Himabindu Lakkaraju. 2023. In-context unlearning: Language models as few shot unlearners. arXiv preprint arXiv:2310.07579. [find](https://scholar.google.com/scholar?q=Martin+Pawelczyk%2C+Seth+Neel%2C+and+Himabindu+Lakkaraju%2E+2023%2E+In%2Dcontext+unlearning%3A+Language+models+as+few+shot+unlearners%2E+arXiv+preprint+arXiv%3A2310%2E07579%2E)
25. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2024. [Direct preference optimization: Your language model is secretly a reward model](https://md.research.elamur.ai/free/en/papers/rafailov-2023-direct-preference-optimization-language-model/). Advances in Neural Information Processing Systems, 36.
26. Igal Sason and Sergio Verdú. 2016. f-divergence inequalities via functional domination. In 2016 IEEE International Conference on the Science of Electrical Engineering (ICSEE), pages 1–5. IEEE. [find](https://scholar.google.com/scholar?q=Igal+Sason+and+Sergio+Verd%C3%BA%2E+2016%2E+f%2Ddivergence+inequalities+via+functional+domination%2E+In+2016+IEEE+International+Conference+on+the+Science+of+Electrical+Engineering+%28ICSEE%29%2C+pages+1%E2%80%935%2E+IEEE%2E)
27. Ayush Sekhari, Jayadev Acharya, Gautam Kamath, and Ananda Theertha Suresh. 2021. Remember what you want to forget: Algorithms for machine unlearning. [find](https://scholar.google.com/scholar?q=Ayush+Sekhari%2C+Jayadev+Acharya%2C+Gautam+Kamath%2C+and+Ananda+Theertha+Suresh%2E+2021%2E+Remember+what+you+want+to+forget%3A+Algorithms+for+machine+unlearning%2E)
28. Advances in Neural Information Processing Systems, 34:18075–18086. [find](https://scholar.google.com/scholar?q=Advances+in+Neural+Information+Processing+Systems%2C+34%3A18075%E2%80%9318086%2E)
29. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, et al. 2023. [Llama 2: Open foundation and fine-tuned chat models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-2-open-foundation-fine/). arXiv preprint arXiv:2307.09288.
30. Paul Voigt and Axel Von dem Bussche. 2017. The eu general data protection regulation (gdpr). A Practical Guide, 1st Ed., Cham: Springer International Publishing, 10(3152676):10–5555. [find](https://scholar.google.com/scholar?q=Paul+Voigt+and+Axel+Von+dem+Bussche%2E+2017%2E+The+eu+general+data+protection+regulation+%28gdpr%29%2E+A+Practical+Guide%2C+1st+Ed%2E%2C+Cham%3A+Springer+International+Publishing%2C+10%283152676%29%3A10%E2%80%935555%2E)
31. Lingzhi Wang, Tong Chen, Wei Yuan, Xingshan Zeng, Kam-Fai Wong, and Hongzhi Yin. 2023. Kga: A general machine unlearning framework based on knowledge gap alignment. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13264– 13276. [find](https://scholar.google.com/scholar?q=Lingzhi+Wang%2C+Tong+Chen%2C+Wei+Yuan%2C+Xingshan+Zeng%2C+Kam%2DFai+Wong%2C+and+Hongzhi+Yin%2E+2023%2E+Kga%3A+A+general+machine+unlearning+framework+based+on+knowledge+gap+alignment%2E+In+Proceedings+of+the+61st+Annual+M)
32. Kun Wu, Jie Shen, Yue Ning, Ting Wang, and Wendy Hui Wang. 2023. Certified edge unlearning for graph neural networks. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 2606–2617. [find](https://scholar.google.com/scholar?q=Kun+Wu%2C+Jie+Shen%2C+Yue+Ning%2C+Ting+Wang%2C+and+Wendy+Hui+Wang%2E+2023%2E+Certified+edge+unlearning+for+graph+neural+networks%2E+In+Proceedings+of+the+29th+ACM+SIGKDD+Conference+on+Knowledge+Discovery+and+Data+M)
33. Taiqiang Wu, Chaofan Tao, Jiahao Wang, Zhe Zhao, and Ngai Wong. 2024. [Rethinking kullback-leibler divergence in knowledge distillation for large language models](https://md.research.elamur.ai/free/en/papers/wu-2025-rethinking-kullback-leibler-divergence-knowledge/). arXiv preprint arXiv:2404.02657.
34. Yuanshun Yao, Xiaojun Xu, and Yang Liu. 2023. [Large language model unlearning](https://md.research.elamur.ai/free/en/papers/yao-2023-large-language-model-unlearning/). arXiv preprint arXiv:2310.10683.
35. Dongkeun Yoon, Joel Jang, Sungdong Kim, and Minjoon Seo. 2023. Gradient ascent post-training enhances language model generalization. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 851–864. [find](https://scholar.google.com/scholar?q=Dongkeun+Yoon%2C+Joel+Jang%2C+Sungdong+Kim%2C+and+Minjoon+Seo%2E+2023%2E+Gradient+ascent+post%2Dtraining+enhances+language+model+generalization%2E+In+Proceedings+of+the+61st+Annual+Meeting+of+the+Association+for+Co)
36. Charles Yu, Sullam Jeoung, Anish Kasi, Pengfei Yu, and Heng Ji. 2023. Unlearning bias in language models by partitioning gradients. In Findings of the Association for Computational Linguistics: ACL 2023, pages 6032–6048. [find](https://scholar.google.com/scholar?q=Charles+Yu%2C+Sullam+Jeoung%2C+Anish+Kasi%2C+Pengfei+Yu%2C+and+Heng+Ji%2E+2023%2E+Unlearning+bias+in+language+models+by+partitioning+gradients%2E+In+Findings+of+the+Association+for+Computational+Linguistics%3A+ACL+20)
37. Eric Zhang, Kai Wang, Xingqian Xu, Zhangyang Wang, and Humphrey Shi. 2023a. Forget-me-not: Learning to forget in text-to-image diffusion models. arXiv preprint arXiv:2303.17591. — [arXiv:2303.17591](https://arxiv.org/abs/2303.17591) [find](https://scholar.google.com/scholar?q=Eric+Zhang%2C+Kai+Wang%2C+Xingqian+Xu%2C+Zhangyang+Wang%2C+and+Humphrey+Shi%2E+2023a%2E+Forget%2Dme%2Dnot%3A+Learning+to+forget+in+text%2Dto%2Dimage+diffusion+models%2E+arXiv+preprint+arXiv%3A2303%2E17591%2E+%E2%80%94+%5BarXiv%3A2303%2E17591%5D%28h)
38. Jinghan Zhang, Junteng Liu, Junxian He, et al. 2023b. Composing parameter-efficient modules with arithmetic operation. Advances in Neural Information Processing Systems, 36:12589–12610. [find](https://scholar.google.com/scholar?q=Jinghan+Zhang%2C+Junteng+Liu%2C+Junxian+He%2C+et+al%2E+2023b%2E+Composing+parameter%2Defficient+modules+with+arithmetic+operation%2E+Advances+in+Neural+Information+Processing+Systems%2C+36%3A12589%E2%80%9312610%2E)
39. Ruiqi Zhang, Licong Lin, Yu Bai, and Song Mei. 2024. [Negative preference optimization: From catastrophic collapse to effective unlearning](https://md.research.elamur.ai/free/en/papers/zhang-2024-negative-preference-optimization-catastrophic-collapse/). arXiv preprint arXiv:2404.05868.

*Updated 31 October 2026*