Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Shaofei Lu has a total of 49 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Causal-driven Large Language Models with Faithful Reasoning for Knowledge Question Answering

    by Jiawei Wang, Da Cao, Shaofei Lu, Zhanchang Ma, Junbin Xiao, Tat-Seng Chua

    In Large Language Models (LLMs), text generation that involves knowledge representation is often fraught with the risk of "hallucinations'', where models confidently produce erroneous or fabricated content. These inaccuracies often stem from intrinsic biases in the pre-training stage or from the incorporation of human preference biases during the fine-tuning process. To mitigate these issues, we take inspiration from Goldman's causal theory of knowledge, which asserts that knowledge is not merely about having a true belief but also involves a causal connection between the belief and the truth of the proposition. We instantiate this theory within the context of Knowledge Question Answering (KQA) by constructing a causal graph that delineates the pathways between the candidate knowledge and belief. Through the application of the do-calculus rules from structural causal models, we devise an unbiased estimation framework based on this causal graph, thereby establishing a methodology for knowledge modeling grounded in causal inference. The resulting CORE framework (short for "Causal knOwledge REasoning'') is comprised of four essential components: question answering, causal reasoning, belief scoring, and refinement. Together, they synergistically improve the KQA system by fostering faithful reasoning and introspection. Extensive experiments are conducted on ScienceQA and HotpotQA datasets, which demonstrate the effectiveness and rationality of the CORE framework.


  • Graph Reasoning With Supervised Contrastive Learning for Legal Judgment Prediction

    by Jiawei Wang, Yuquan Le, Da Cao, Shaofei Lu, Zhe Quan, Meng Wang

    Given the fact descriptions of legal cases, the legal judgment prediction (LJP) problem aims to determine three judgment tasks of law articles, charges, and the term of penalty. Most existing studies have considered task dependencies while neglecting the prior dependencies of labels among different tasks. Therefore, how to make better use of the information on the relation dependencies among tasks and labels becomes a crucial issue. To this end, we transform the text classification problem into a node classification framework based on graph reasoning and supervised contrastive learning (SCL) techniques, named GraSCL. Specifically, we first design a graph reasoning network to model the potential dependency structures and facilitate relational learning under various graph topologies. Then, we introduce the SCL method for the LJP task to further leverage the label relation on the graph. To accommodate the node classification settings, we extend the traditional SCL method to novel variants for SCL at the node level, which allows the GraSCL framework to be trained efficiently even with small batches. Furthermore, to recognize the importance of hard negative samples in contrastive learning, we introduce a simple yet effective technique called online hard negative mining (OHNM) to enhance our SCL approach. This technique complements our SCL method and enables us to control the number and complexity of negative samples, leading to further improvements in the model’s performance. Finally, extensive experiments are conducted on two well-known benchmarks, demonstrating the effectiveness and rationality of our proposed SCL approach as compared to the state-of-the-art competitors.


  • Open Access

    P-Raft: An Efficient and Robust Consensus Mechanism for Consortium Blockchains

    by Shaofei Lu, Xuyang Zhang, Renke Zhao, Lizhi Chen, Junyi Li, Guanzhong Yang

    With the rise in blockchain technology, consortium blockchains have garnered increasing attention in practical applications due to their decentralization and immutability. However, the performance of current consortium blockchains remains a significant obstacle to large-scale commercial adoption. The consensus algorithm, as a fundamental component of blockchain technology, plays a critical role in ensuring both security and efficiency. Unfortunately, most existing consensus algorithms for consortium blockchains are vote-based consensus algorithms, and the performance of vote-based consensus algorithms is largely limited by the performance of the leader node. Therefore, we present P-Raft: a high-performance consensus algorithm that builds upon the Raft algorithm and leverages node server performance evaluations. The primary objectives of this article included enhancing the efficiency of Leader processing, promoting the utilization of the consortium blockchain, and ensuring the robustness of Leader election. Specifically designed to meet the service requirements of consortium blockchain’s consensus mechanism, the P-Raft incorporated the Yasa model, which evaluated the instant machine performance of each node. The performance of each node is then associated with the election timeout, ensuring that nodes with superior performance are more likely to be chosen as Leaders. Additionally, we implemented a leader verification mechanism based on the Bohen-Lynn-Shacham (BLS) signature, which prevented malicious Byzantine nodes from becoming Leaders without receiving enough votes. Empirical findings show that the P-Raft can swiftly designate high-performing nodes as Leaders, thereby greatly improving service efficiency in the consensus process and the overall performance of the consensus mechanism. Ultimately, P-Raft is better equipped to meet the demands of consortium blockchain applications for large-scale transactions.


  • Open Access

    CCIO: A Cross-Chain Interoperability Approach for Consortium Blockchains Based on Oracle

    by Shaofei Lu, Jingru Pei, Renke Zhao, Xiaochen Yu, Xuyang Zhang, Junyi Li, Guanzhong Yang

    Cross-chain interoperability can expand the ability of data interaction and value circulation between different blockchains, especially the value interaction and information sharing between industry consortium blockchains. However, some current public blockchain cross-chain technologies or data migration schemes between consortium blockchains need help to meet the consortium blockchain requirements for efficient two-way data interaction. The critical issue to solve in cross-chain technology is improving the efficiency of cross-chain exchange while ensuring the security of data transmission outside the consortium blockchain. In this article, we design a cross-chain architecture based on blockchain oracle technology. Then, we propose a bidirectional information cross-chain interaction approach (CCIO) based on the former architecture, we novelly improve three traditional blockchain oracle patterns, and we combine a mixture of symmetric and asymmetric keys to encrypt private information to ensure cross-chain data security. The experimental results demonstrate that the proposed CCIO approach can achieve efficient and secure two-way cross-chain data interactions and better meet the application needs of large-scale consortium blockchains.


  • A DRL-Based Decentralized Computation Offloading Method: An Example of an Intelligent Manufacturing Scenario

    by Shaofei Lu, Shen Liu, Yajun Zhu, Wei Liang, Kuanching Li, Yingping Lu

    With the development of edge computing and 5G, the demand for resources-limited devices to execute computation-intensive tasks can be effectively alleviated. The research of computation offloading lays an essential foundation for realizing mobile edge computing, and Deep Reinforcement Learning (DRL) has become an emerging technique to address the computation offloading problem. This paper utilizes a DRL-based algorithm to design a decentralized computation offloading framework aimed at minimizing the computational cost. We employ a multi-user system model with a single-edge server suitable for industrial scenarios. Then we propose a Dual-Critic Deep Deterministic Policy Gradient (DC-DDPG) algorithm based on Deep Deterministic Policy Gradient (DDPG) algorithm to tackle computation offloading and resource allocation problems for all users, respectively. DC-DDPG adopts two critic nets in both the primary net and the target net to fit the action-value of two different optimization objectives, which expedites the convergence during the training process and reduces the computational cost of the edge computation system during operation. Compared with other deep reinforcement learning methods such as Deep Q-Network (DQN) and DDPG, numerical results demonstrate that the proposed DC-DDPG algorithm has a faster convergence speed and performs significantly better than other DRL-based algorithms in terms of system computational cost in computing-intensive tasks, which is more suitable for industrial intelligent manufacturing scenarios with large data volume.


  • Long Short-Term Memory in Chemistry Dynamics Simulation

    by Heng Wu, Shaofei Lu, Colmenares-Diaz Eduardo, Junbin Liang, Jingke She, Xiaolin Tan

  • Research on Efficient and Fuzzy Matching Algorithm in Information Dissemination System

    by Qinwen Zuo, Fred Wu, Fei Yan, Shaofei Lu, Colmenares-Diaz Eduardo, Junbin Liang