Fei Sun ( 孙飞 鲁迅的字 Lu Xun's Calligraphy )

Despite near-perfect results reported in the literature, the effectiveness of model editing in real-world applications remains unclear. To bridge this gap, we introduce QAEdit, a new benchmark aligned with widely used question answering (QA) datasets, and WILD, a task-agnostic evaluation framework designed to better reflect real-world usage of model editing. Our single editing experiments show that current editing methods perform substantially worse than previously reported (38.5% vs. 96.8%). We demonstrate that it stems from issues in the synthetic evaluation practices of prior work. Among them, the most severe is the use of teacher forcing during testing, which leaks both content and length of the ground truth, leading to overestimated performance. Furthermore, we simulate practical deployment by sequential editing, revealing that current approaches fail drastically with only 1000 edits. This work calls for a shift in model editing research toward rigorous evaluation and the development of robust, scalable methods that can reliably update knowledge in LLMs for real-world use.

Despite significant progress in model editing methods, their application in real-world scenarios remains challenging as they often cause large language models (LLMs) to collapse. Among them, ROME is particularly concerning, as it could disrupt LLMs with only a single edit. In this paper, we study the root causes of such collapse. Through extensive analysis, we identify two primary factors that contribute to the collapse: i) inconsistent handling of prefixed and unprefixed keys in the parameter update equation may result in very small denominators, causing excessively large parameter updates; ii) the subject of collapse cases is usually the first token, whose unprefixed key distribution significantly differs from the prefixed key distribution in autoregressive transformers, causing the aforementioned issue to materialize. To validate our findings, we propose a simple yet effective approach: uniformly using prefixed keys during editing phase and adding prefixes during testing phase to ensure the consistency between training and testing. The experimental results show that the proposed solution can prevent model collapse while maintaining the effectiveness of the edits.

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.

With the explosive growth of online information, recommender systems play a key role to alleviate such information overload. Due to the important application value of recommender systems, there have always been emerging works in this field. In recommender systems, the main challenge is to learn the effective user/item representations from their interactions and side information (if any). Recently, graph neural network (GNN) techniques have been widely utilized in recommender systems since most of the information in recommender systems essentially has graph structure and GNN has superiority in graph representation learning. This article aims to provide a comprehensive review of recent research efforts on GNN-based recommender systems. Specifically, we provide a taxonomy of GNN-based recommendation models according to the types of information used and recommendation tasks. Moreover, we systematically analyze the challenges of applying GNN on different types of data and discuss how existing works in this field address these challenges. Furthermore, we state new perspectives pertaining to the development of this field. We collect the representative papers along with their open-source implementations in .

Modeling users’ dynamic preferences from their historical behaviors is challenging and crucial for recommendation systems. Previous methods employ sequential neural networks to encode users’ historical interactions from left to right into hidden representations for making recommendations. Despite their effectiveness, we argue that such left-to-right unidirectional models are sub-optimal due to the limitations including: begin enumerate* [label=seriesitshapealph*upshape)] item unidirectional architectures restrict the power of hidden representation in users’ behavior sequences; item they often assume a rigidly ordered sequence which is not always practical. end enumerate* To address these limitations, we proposed a sequential recommendation model called BERT4Rec, which employs the deep bidirectional self-attention to model user behavior sequences. To avoid the information leakage and efficiently train the bidirectional model, we adopt the Cloze objective to sequential recommendation, predicting the random masked items in the sequence by jointly conditioning on their left and right context. In this way, we learn a bidirectional representation model to make recommendations by allowing each item in user historical behaviors to fuse information from both left and right sides. Extensive experiments on four benchmark datasets show that our model outperforms various state-of-the-art sequential models consistently.