Editing Large Language Models: Problems, Methods, and Opportunities

By Yunzhi Yao, Peng Wang, Bo Tian, Siyuan Cheng, Zhoubo Li, Shumin Deng, Huajun Chen, Ningyu Zhang. Conference on Empirical Methods in Natural Language Processing, 2023

475 citations recorded by the source corpus.

Despite the ability to train capable LLMs, the methodology for maintaining their relevancy and rectifying errors remains elusive. To this end, the past few years have witnessed a surge in techniques for editing LLMs, the objective of which is to efficiently alter the behavior of LLMs within a specific domain without negatively impacting performance across other inputs. This paper embarks on a deep exploration of the problems, methods, and opportunities related to model editing for LLMs. In particular, we provide an exhaustive overview of the task definition and challenges associated with model editing, along with an in-depth empirical analysis of the most progressive methods currently at our disposal. We also build a new benchmark dataset to facilitate a more robust evaluation and pinpoint enduring issues intrinsic to existing techniques. Our objective is to provide valuable insights into the effectiveness and feasibility of each editing technique, thereby assisting the community in making informed decisions on the selection of the most appropriate method for a specific task or context. Code and datasets are available at https://github.com/zjunlp/EasyEdit.

View Paper: https://www.semanticscholar.org/paper/f5c73d9e6641b018b633690102121f5605d34fb0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Shirui Pan
2. Ce Zheng
3. Yufei Wang
4. Jiashuo Sun
5. Lei Li
6. Jingjing Xu
7. Linhao Luo
8. Qingxiu Dong

Citations in the corpus

• Unifying Large Language Models and Knowledge Graphs: A Roadmap (2023, background): “There are some research efforts proposed for editing knowledge in LLMs [252] without re-training the whole LLMs.”
• Think-on-Graph: Deep and Responsible Reasoning of Large Language Model with Knowledge Graph (2023, background): “Moreover, prior studies Yao et al. (2023); Sun et al. (2023a); Xie et al. (2023) have demonstrated that LLMs can enhance their reasoning performance by employing algorithms based on limited search methods.”
• A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions (2023, unknown): “…editing techniques can fix factual errors and refresh outdated information to mitigate the knowledge gap, which can be categorized into two classes: changing the model’s behavior by modifying the model parameters or using an external model plug-in with the original model frozen (Yao et al., 2023c).”
• Knowledge Conflicts for LLMs: A Survey (2024, unknown): “This can result in LLMs producing inconsistent responses when dealing with the same piece of knowledge in varying situations (Li et al., 2023f; Yao et al., 2023).”
• Can We Edit Factual Knowledge by In-Context Learning? (2023, background): “Knowledge Editing Methods Recent studies on knowledge editing are mostly hype-network-based or attribution-based (Yao et al., 2023).”
• A Survey on the Memory Mechanism of Large Language Model-based Agents (2024, unknown): “In addition, cognitive psychology has been studied for a long time, so many effective human memory theories and architectures have been accumulated, which can support more advanced capabilities of the agents [147].”