Summary of Llm Unlearning Via Loss Adjustment with Only Forget Data, by Yaxuan Wang et al.
LLM Unlearning via Loss Adjustment with Only Forget Data
by Yaxuan Wang, Jiaheng Wei, Chris Yuhao Liu, Jinlong Pang, Quan Liu, Ankit Parag Shah, Yujia Bao, Yang Liu, Wei Wei
First submitted to arxiv on: 14 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed approach, Forget data only Loss AjustmenT (FLAT), eliminates the need for retain data or a reference LLM in unlearning large language models. This method guides the model on what not to respond to and how to respond based on forget data. FLAT maximizes f-divergence between template answers and forget answers with respect to forget data, allowing for loss adjustment by assigning importance weights. Empirical results show that FLAT achieves superior unlearning performance while minimizing impact on retained capabilities across various tasks, including copyrighted content unlearning and entity unlearning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models can be trained to “forget” specific information to ensure responsible AI use. The current approach relies on retain data or a reference LLM, but these methods struggle to balance unlearning performance with overall model utility. This new method eliminates the need for retain data or a reference LLM and guides the model on what not to respond to and how to respond based on forget data. The results show that this method is effective in achieving superior unlearning performance while minimizing impact on retained capabilities. |
