The Butterfly Effect of Model Editing: Few Edits Can Trigger Large Language Models Collapse
CAS Key Laboratory of AI Safety, Institute of Computing Technology, Chinese Academy of Sciences（中国科学院计算技术研究所人工智能安全重点实验室）；University of Chinese Academy of Sciences（中国科学院大学）；Baidu Inc.（百度公司）
Although model editing has shown promise in revising knowledge in Large Language Models (LLMs), its impact on the inherent capabilities of LLMs is often overlooked. In this work, we reveal a critical phenomenon: even a single edit can trigger model collapse, manifesting as significant performance degradation in various benchmark tasks. However, benchmarking LLMs after each edit, while necessary to prevent such collapses, is impractically time-consuming and resource-intensive. To mitigate this, we propose using perplexity as a surrogate metric, validated by extensive experiments demonstrating changes in an edited model's perplexity are strongly correlated with its downstream task performances. We further conduct an in-depth study on sequential editing, a practical setting for real-world scenarios, across various editing methods and LLMs, focusing on hard cases from our previous single edit studies. The results indicate that nearly all examined editing methods result in model collapse after only few edits. To facilitate further research, we have utilized GPT-3.5 to develop a new dataset, HardEdit, based on those hard cases. This dataset aims to establish the foundation for pioneering research in reliable model editing and the mechanisms underlying editing-induced model collapse. We hope this work can draw the community's attention to the potential risks inherent in model editing practices.
Orthogonal Representation Editing: Decoupling Semantic Entanglement in Batch Knowledge Editing of LLMs
School of Computer Science and Technology, Tianjin University（天津大学计算机科学与技术学院）；Kexin Technology（可心科技）；University of Chinese Academy of Sciences（中国科学院大学）；Tencent Hunyuan（腾讯混元）；College of Computer Science, Chongqing University（重庆大学计算机学院）
Potential and Challenges of Model Editing for Social Debiasing
Zhejiang University School of Engineering（浙江大学工程学院）；Westlake University（西湖大学）；Institute of Advanced Technology, Westlake Institute for Advanced Study（西湖高等研究院先进技术研究所）
Visual-Oriented Fine-Grained Knowledge Editing for MultiModal Large Language Models
Hefei University of Technology（合肥工业大学）；University of Science and Technology of China（中国科学技术大学）；Hefei Comprehensive National Science Center（合肥综合性国家科学中心）
Constraining Sequential Model Editing with Editing Anchor Compression
National Engineering Research Center of Speech and Language Information Processing, University of Science and Technology of China（中国科学技术大学语音及语言信息处理国家工程研究中心）；Zhejiang University（浙江大学）；University of California, Los Angeles（加利福尼亚大学洛杉矶分校）
KEditVis: A Visual Analytics System for Knowledge Editing of Large Language Models
Zhejiang University（浙江大学）；State Key Lab of CAD&CG, Zhejiang University（浙江大学CAD&CG国家重点实验室）；Hong Kong University of Science and Technology（香港科技大学）；School of Software Engineering, South China University of Technology（华南理工大学软件学院）；School of Software Technology, Zhejiang University（浙江大学软件学院）
NVIDIA（英伟达）；University of Toronto（多伦多大学）；Vector Institute（矢量研究所）；MIT（麻省理工学院）
Generative adversarial networks (GANs) have recently found applications in image editing. However, most GAN based image editing methods often require large scale datasets with semantic segmentation annotations for training, only provide high level control, or merely interpolate between different images. Here, we propose EditGAN, a novel method for high quality, high precision semantic image editing, allowing users to edit images by modifying their highly detailed part segmentation masks, e.g., drawing a new mask for the headlight of a car. EditGAN builds on a GAN framework that jointly models …
