natural-language-processing-9

knowledge editing

3260 papers • 126 benchmarks • 313 datasets

This task has no description! Would you like to contribute one?

(Image credit: Papersgraph)

Benchmarks

These leaderboards are used to track progress in knowledge-editing-9

Trend

Dataset

Best Model

Actions

zsRE

Libraries

i

Use these libraries to find knowledge-editing-9 models and implementations

zjunlp/easyedit

5 papers 1,343

Datasets

KnowEdit

WikiFactDiff

Subtasks

No subtasks available.

Most implemented papers

Editing Language Model-based Knowledge Graph Embeddings

Huajun Chen, Ningyu Zhang, Bo Tian, Siyuan Cheng, Zelin Dai, Feiyu Xiong, Wei Guo•Tue Jan 24 2023

This paper proposes a new task of editing language model-based KG embeddings, and proposes a simple yet strong baseline dubbed KGEditor, which utilizes additional parametric layers of the hypernetwork to edit/add facts.

33

Content

0

Paper Graph

MQuAKE: Assessing Knowledge Editing in Language Models via Multi-Hop Questions

Christopher D. Manning, Christopher Potts, Zexuan Zhong, Danqi Chen, Zhengxuan Wu•Tue May 23 2023

This work presents a benchmark, MQuAKE (Multi-hop Question Answering for Knowledge Editing), comprising multi-hop questions that assess whether edited models correctly answer questions where the answer should change as an entailed consequence of edited facts, and proposes a simple memory-based approach, MeLLo.

283 0

Paper Graph

Can We Edit Factual Knowledge by In-Context Learning?

Jingjing Xu, Baobao Chang, Lei Li, Zhiyong Wu, Qingxiu Dong, Ce Zheng, Yuxuan Fan•Sun May 21 2023

Experiments show that in-context knowledge editing (IKE), without any gradient and parameter updating, achieves a competitive success rate compared to gradient-based methods on GPT-J (6B) but with much fewer side effects, including less over-editing on similar but unrelated facts and less knowledge forgetting on previously stored knowledge.

293 0

Paper Graph

EasyEdit: An Easy-to-use Knowledge Editing Framework for Large Language Models

Zekun Xi, Huajun Chen, Yunzhi Yao, Bo Tian, Peng Wang, Mengru Wang, Yuansheng Ni, Siyuan Cheng, Ningyu Zhang, Kangwei Liu, Guozhou Zheng•Sun Aug 13 2023

EasyEdit is proposed, an easy-to-use knowledge editing framework for LLMs that supports various cutting-edge knowledge editing approaches and can be readily applied to many well-known LLMs such as T5, GPT-J, LlaMA, etc.

84 0

Paper Graph

Cross-Lingual Knowledge Editing in Large Language Models

Yunlong Liang, Jiaan Wang, Zengkui Sun, Yu Cao, Jiarong Xu•Fri Sep 15 2023

This paper collects a large-scale cross-lingual synthetic dataset, conducts English editing on various knowledge editing methods covering different paradigms, and evaluates their performance in Chinese, and vice versa to figure out this cross-lingual effect in knowledge editing.

17 0

Paper Graph

MLaKE: Multilingual Knowledge Editing Benchmark for Large Language Models

Liang Pang, Huawei Shen, Xueqi Cheng, Hanxing Ding, Zihao Wei, Jingcheng Deng•Sat Apr 06 2024

This study introduces MLaKE (Multilingual Language Knowledge Editing), a novel benchmark comprising 4072 multi-hop and 5360 single-hop questions designed to evaluate the adaptability of knowledge editing methods across five languages: English, Chinese, Japanese, French, and German.

10 0

Paper Graph

Assessing Knowledge Editing in Language Models via Relation Perspective

Kang Liu, Fangyu Lei, Yifan Wei, Xiaoyan Yu, Huanhuan Ma, Yixuan Weng, Ran Song•Tue Nov 14 2023

A new benchmark named RaKE is constructed, which focuses on Relation based Knowledge Editing, and confirms that knowledge related to relations is not only stored in the FFN network but also in the attention layers, which provides experimental support for future relation-based knowledge editing methods.

16 0

Paper Graph

A Comprehensive Study of Knowledge Editing for Large Language Models

Yong Jiang, Fei Huang, Zekun Xi, Jia-Chen Gu, Jun Zhou, Pengjun Xie, Huajun Chen, Ningyu Zhang, Yunzhi Yao, Bo Tian, Peng Wang, Shumin Deng, Mengru Wang, Shengyu Mao, Jintian Zhang, Yuansheng Ni, Siyuan Cheng, Ziwen Xu, Xin Xu, Lei Liang, Zhiqiang Zhang, Xiaowei Zhu•Mon Jan 01 2024

A unified categorization criterion is proposed that classifies knowledge editing methods into three groups: resorting to external knowledge, merging knowledge into the model, and editing intrinsic knowledge and an in-depth analysis of knowledge location is provided, which can give a deeper understanding of the knowledge structures inherent within LLMs.

126 0

Paper Graph

Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Models

Mohit Bansal, Asma Ghandeharioun, Been Kim, Peter Hase•Mon Jan 09 2023

This paper shows that localization conclusions from representation denoising do not provide any insight into which model MLP layer would be best to edit in order to override an existing stored fact with a new one, and finds that which layer the authors edit is a far better predictor of performance.

235 0

Paper Graph

Propagating Knowledge Updates to LMs Through Distillation

Eunsol Choi, Greg Durrett, Yasumasa Onoe, Michael J.Q. Zhang, Shankar Padmanabhan•Wed Jun 14 2023

It is demonstrated that a context distillation-based approach can both impart knowledge about entities and propagate that knowledge to enable broader inferences and is more effective at propagating knowledge updates than fine-tuning and other gradient-based knowledge-editing methods.

22 0

Paper Graph

Adding a benchmark result helps the community track progress.