arXiv subjects

Cui Jiangtao

Publications and source records attributed to Cui Jiangtao.

2 recordsLinked to original sources

arXiv2026

Towards Better Evolution Modeling for Temporal Knowledge Graphs

Temporal knowledge graphs (TKGs) structurally preserve evolving human knowledge. Recent research has focused on designing models to learn the evolutionary nature of TKGs to predict future facts, achieving impressive results. For instance, Hits@10 scores over 0.9 on YAGO dataset. However, we find that existing benchmarks inadvertently introduce a shortcut. Near state-of-the-art performance can be simply achieved by counting co-occurrences, without using any temporal information. In this work, we examine the root cause of this issue, identifying inherent biases in current datasets and over simplified form of evaluation task that can be exploited by these biases. Through this analysis, we further uncover additional limitations of existing benchmarks, including unreasonable formatting of time-interval knowledge, ignorance of learning knowledge obsolescence, and insufficient information for precise evolution understanding, all of which can amplify the shortcut and hinder a fair assessment. Therefore, we introduce the TKG evolution benchmark. It includes four bias-corrected datasets and two novel tasks closely aligned with the evolution process, promoting a more accurate understanding of the challenges in TKG evolution modeling. Benchmark is available at: https://github.com/zjs123/TKG-Benchmark.

cs.AI↗

arXiv2021

Revisiting $k$-Nearest Neighbor Graph Construction on High-Dimensional Data : Experiments and Analyses

The $k$-nearest neighbor graph (KNNG) on high-dimensional data is a data structure widely used in many applications such as similarity search, dimension reduction and clustering. Due to its increasing popularity, several methods under the same framework have been proposed in the past decade. This framework contains two steps, i.e. building an initial KNNG (denoted as \texttt{INIT}) and then refining it by neighborhood propagation (denoted as \texttt{NBPG}). However, there remain several questions to be answered. First, it lacks a comprehensive experimental comparison among representative solutions in the literature. Second, some recently proposed indexing structures, e.g., SW and HNSW, have not been used or tested for building an initial KNNG. Third, the relationship between the data property and the effectiveness of \texttt{NBPG} is still not clear. To address these issues, we comprehensively compare the representative approaches on real-world high-dimensional data sets to provide practical and insightful suggestions for users. As the first attempt, we take SW and HNSW as the alternatives of \texttt{INIT} in our experiments. Moreover, we investigate the effectiveness of \texttt{NBPG} and find the strong correlation between the huness phenomenon and the performance of \texttt{NBPG}.

cs.DS↗