arXiv subjects

Yipeng Ji

Publications and source records attributed to Yipeng Ji.

4 recordsLinked to original sources

arXiv2025

Cross-Attention Speculative Decoding

Speculative decoding (SD) is a widely adopted approach for accelerating inference in large language models (LLMs), particularly when the draft and target models are well aligned. However, state-of-the-art SD methods typically rely on tightly coupled, self-attention-based Transformer decoders, often augmented with auxiliary pooling or fusion layers. This coupling makes them increasingly complex and harder to generalize across different models. We present Budget EAGLE (Beagle), the first, to our knowledge, cross-attention-based Transformer decoder SD model that achieves performance on par with leading self-attention SD models (EAGLE-v2) while eliminating the need for pooling or auxiliary components, simplifying the architecture, improving training efficiency, and maintaining stable memory usage during training-time simulation. To enable effective training of this novel architecture, we propose Two-Stage Block-Attention Training, a new method that achieves training stability and convergence efficiency in block-level attention scenarios. Extensive experiments across multiple LLMs and datasets show that Beagle achieves competitive inference speedups and higher training efficiency than EAGLE-v2, offering a strong alternative for architectures in speculative decoding.

cs.CL↗

arXiv2023

uxSense: Supporting User Experience Analysis with Visualization and Computer Vision

Analyzing user behavior from usability evaluation can be a challenging and time-consuming task, especially as the number of participants and the scale and complexity of the evaluation grows. We propose uxSense, a visual analytics system using machine learning methods to extract user behavior from audio and video recordings as parallel time-stamped data streams. Our implementation draws on pattern recognition, computer vision, natural language processing, and machine learning to extract user sentiment, actions, posture, spoken words, and other features from such recordings. These streams are visualized as parallel timelines in a web-based front-end, enabling the researcher to search, filter, and annotate data across time and space. We present the results of a user study involving professional UX researchers evaluating user data using uxSense. In fact, we used uxSense itself to evaluate their sessions.

cs.HC↗

arXiv2021

Relevance-Aware Anomalous Users Detection in Social Network via Graph Neural Network

Anomalous users detection in social network is an imperative task for security problems. Motivated by the great power of Graph Neural Networks(GNNs), many current researches adopt GNN-based detectors to reveal the anomalous users. However, the increasing scale of social activities, explosive growth of users and manifold technical disguise render the user detection a difficult task. In this paper, we propose an innovate Relevance-aware Anomalous Users Detection model (RAU-GNN) to obtain a fine-grained detection result. RAU-GNN first extracts multiple relations of all types of users in social network, including both benign and anomalous users, and accordingly constructs the multiple user relation graph. Secondly, we employ relevance-aware GNN framework to learn the hidden features of users, and discriminate the anomalous users after discriminating. Concretely, by integrating Graph Convolution Network(GCN) and Graph Attention Network(GAT), we design a GCN-based relation fusion layer to aggregate initial information from different relations, and a GAT-based embedding layer to obtain the high-level embeddings. Lastly, we feed the learned representations to the following GNN layer in order to consolidate the node embedding by aggregating the final users' embeddings. We conduct extensive experiment on real-world datasets. The experimental results show that our approach can achieve high accuracy for anomalous users detection.

cs.SI↗

arXiv2021

An Anomaly Event Detection Method Based on GNN Algorithm for Multi-data Sources

Anomaly event detection is crucial for critical infrastructure security(transportation system, social-ecological sector, insurance service, government sector etc.) due to its ability to reveal and address the potential cyber-threats in advance by analysing the data(messages, microblogs, logs etc.) from digital systems and networks. However, the convenience and applicability of smart devices and the maturity of connected technology make the social anomaly events data multi-source and dynamic, which result in the inadaptability for multi-source data detection and thus affect the critical infrastructure security. To effectively address the proposed problems, in this paper, we design a novel anomaly detection method based on multi-source data. First, we leverage spectral clustering algorithm for feature extraction and fusion of multiple data sources. Second, by harnessing the power of deep graph neural network(Deep-GNN), we perform a fine-gained anomaly social event detection, revealing the threatening events and guarantee the critical infrastructure security. Experimental results demonstrate that our framework outperforms other baseline anomaly event detection methods and shows high tracking accuracy, strong robustness and stability.

cs.SI↗