Hailu Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Hailu Xu has a total of 56 co-authors. Here are the top co-authors by number of shared publications:
Liting Hu
Department of Computer Science and Engineering, University of California at Santa Cruz, Santa Cruz, CA, USA
9 shared publicationsPinchao Liu
School of Computing and Information Sciences, Florida International University, Miami, FL, USA
9 shared publicationsBoyuan Guan
School of Computing and Information Sciences, Florida International University, Miami, FL, USA
5 shared publicationsQingyang Wang
Department of Computer Science and Engineering, Louisiana State University, Baton Rouge, LA, USA
5 shared publicationsShuai Xu
3 shared publications- Zhengxiong Li
University of Colorado Denver, Denver, USA
3 shared publications
Latest Publications
- Open Access
- 23 September 2026
- Published by Elsevier in SSRN Electronic Journal
Near-Duplicate Leakage in Brain Tumor MRI Classification Cluster-Disjoint Re-Evaluation with Parameter-Efficient Fine-Tuning
Near-ceiling accuracies are frequently reported on public brain tumor magnetic resonance imaging (MRI) classification benchmarks, yet the integrity of the underlying data partitions remains insufficiently examined. We analyze a widely used four-class dataset whose release documentation states that duplicate images and train/test overlap were removed. Although exact hashing finds no byte-identical files shared across the official partitions, a correlation-based analysis accounting for rotations and reflections identifies a near-duplicate training counterpart for 1,100 of the 1,600 test images (68.8%). On the official partition, a ResNet18 achieves an accuracy of 0.9973 on matched test images, compared with 0.8540 on unmatched images, a difference of 14.3 percentage points. We therefore construct a cluster-disjoint partition by assigning all images within each detected near-duplicate cluster exclusively to one subset. On the resulting test set, DINOv2 with low-rank adaptation (LoRA) achieves the highest observed accuracy of 0.9748 and a macro-F1 score of 0.9743 while updating only 1.02% of its parameters. Its performance is comparable to the 0.9721 accuracy of fully fine-tuned EfficientNet-B0, which requires 18 times more trainable parameters, and exceeds that of a frozen DINOv2 linear probe by 6.22 percentage points. These findings demonstrate the importance of near-duplicate-aware partitioning for credible benchmark evaluation and establish LoRA-adapted DINOv2 as an accurate, parameter-efficient baseline for brain tumor MRI classification.
- 23 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessNear-Duplicate Leakage in Brain Tumor MRI Classification Cluster-Disjoint Re-Evaluation with Parameter-Efficient Fine-Tuning
Near-ceiling accuracies are frequently reported on public brain tumor magnetic resonance imaging (MRI) classification benchmarks, yet the integrity of the underlying data partitions remains insufficiently examined. We analyze a widely used four-class dataset whose release documentation states that duplicate images and train/test overlap were removed. Although exact hashing finds no byte-identical files shared across the official partitions, a correlation-based analysis accounting for rotations and reflections identifies a near-duplicate training counterpart for 1,100 of the 1,600 test images (68.8%). On the official partition, a ResNet18 achieves an accuracy of 0.9973 on matched test images, compared with 0.8540 on unmatched images, a difference of 14.3 percentage points. We therefore construct a cluster-disjoint partition by assigning all images within each detected near-duplicate cluster exclusively to one subset. On the resulting test set, DINOv2 with low-rank adaptation (LoRA) achieves the highest observed accuracy of 0.9748 and a macro-F1 score of 0.9743 while updating only 1.02% of its parameters. Its performance is comparable to the 0.9721 accuracy of fully fine-tuned EfficientNet-B0, which requires 18 times more trainable parameters, and exceeds that of a frozen DINOv2 linear probe by 6.22 percentage points. These findings demonstrate the importance of near-duplicate-aware partitioning for credible benchmark evaluation and establish LoRA-adapted DINOv2 as an accurate, parameter-efficient baseline for brain tumor MRI classification.
- Open Access
- 31 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
A Survey on Multi-Robot Collaboration Systems: Architectures, Performances, and Applications
This survey provides a comprehensive review of the current landscape of multi-robot collaboration systems and platforms, exploring key technological advancements, architectural approaches, and deployment methodologies that have emerged over the past five years. We systematically analyze the evolution from centralized coordination architectures to distributed consensus mechanisms, the integration of artificial intelligence and machine learning techniques for autonomous decision-making, and the rise of human-robot collaborative frameworks that facilitate seamless interaction between human operators and robot teams. Our review categorizes existing platforms according to their coordination paradigms, while also assessing their scalability, fault tolerance, and real-world applicability. Additionally, we address critical challenges such as communication latency in distributed systems, computational resource allocation, safety assurance in mixed human-robot environments, and ongoing standardization efforts aimed at ensuring interoperability across diverse robot fleets.
- 31 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessA Survey on Multi-Robot Collaboration Systems: Architectures, Performances, and Applications
This survey provides a comprehensive review of the current landscape of multi-robot collaboration systems and platforms, exploring key technological advancements, architectural approaches, and deployment methodologies that have emerged over the past five years. We systematically analyze the evolution from centralized coordination architectures to distributed consensus mechanisms, the integration of artificial intelligence and machine learning techniques for autonomous decision-making, and the rise of human-robot collaborative frameworks that facilitate seamless interaction between human operators and robot teams. Our review categorizes existing platforms according to their coordination paradigms, while also assessing their scalability, fault tolerance, and real-world applicability. Additionally, we address critical challenges such as communication latency in distributed systems, computational resource allocation, safety assurance in mixed human-robot environments, and ongoing standardization efforts aimed at ensuring interoperability across diverse robot fleets.
- Open Access
- 2 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
LightLLM-Enhanced Multi-Robot Collaboration with Human-in-the-Loop
Human-in-the-Loop systems have become increasingly critical for multi-robot task planning, enabling robust coordination in complex and uncertain environments where full autonomy remains challenging. However, current approaches face significant computational overhead and limited scalability, with traditional LLM-based planners requiring substantial memory and long inference latencies. In this paper, we introduce a containerized lightweight LLM framework for Human-in-the-Loop multi-robot collaboration. This system employs a threetier hierarchical calibration architecture for uncertainty-aware decision making across task planning, action selection, and execution monitoring levels. Experimental evaluation with 12 simulated TurtleBot3 robots demonstrates significant improvements and the containerized deployment reduces configuration complexity while enabling scalable multi-robot coordination. The integrated calibration framework achieves effective balance between autonomous operation and human oversight for safetycritical multi-robot applications.
- 2 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessLightLLM-Enhanced Multi-Robot Collaboration with Human-in-the-Loop
Human-in-the-Loop systems have become increasingly critical for multi-robot task planning, enabling robust coordination in complex and uncertain environments where full autonomy remains challenging. However, current approaches face significant computational overhead and limited scalability, with traditional LLM-based planners requiring substantial memory and long inference latencies. In this paper, we introduce a containerized lightweight LLM framework for Human-in-the-Loop multi-robot collaboration. This system employs a threetier hierarchical calibration architecture for uncertainty-aware decision making across task planning, action selection, and execution monitoring levels. Experimental evaluation with 12 simulated TurtleBot3 robots demonstrates significant improvements and the containerized deployment reduces configuration complexity while enabling scalable multi-robot coordination. The integrated calibration framework achieves effective balance between autonomous operation and human oversight for safetycritical multi-robot applications.
- Open Access
Pronunciation Deviation Analysis Through Voice Cloning and Acoustic Comparison
This paper presents a novel approach for detecting mispronunciations by analyzing deviations between a user's original speech and their voice-cloned counterpart with corrected pronunciation. We hypothesize that regions with maximal acoustic deviation between the original and cloned utterances indicate potential mispronunciations. Our method leverages recent advances in voice cloning to generate a synthetic version of the user's voice with proper pronunciation, then performs frame-by-frame comparisons to identify problematic segments. Experimental results demonstrate the effectiveness of this approach in pinpointing specific pronunciation errors without requiring predefined phonetic rules or extensive training data for each target language.
Open AccessPronunciation Deviation Analysis Through Voice Cloning and Acoustic Comparison
This paper presents a novel approach for detecting mispronunciations by analyzing deviations between a user's original speech and their voice-cloned counterpart with corrected pronunciation. We hypothesize that regions with maximal acoustic deviation between the original and cloned utterances indicate potential mispronunciations. Our method leverages recent advances in voice cloning to generate a synthetic version of the user's voice with proper pronunciation, then performs frame-by-frame comparisons to identify problematic segments. Experimental results demonstrate the effectiveness of this approach in pinpointing specific pronunciation errors without requiring predefined phonetic rules or extensive training data for each target language.
- Open Access
- 6 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 12-16
mmVanish: Extending the Vanish Attack for Multi-Radar Exploitation of mmWave Sensing with Meta-material Tags
- 6 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 12-16
Open AccessmmVanish: Extending the Vanish Attack for Multi-Radar Exploitation of mmWave Sensing with Meta-material Tags
- Open Access
- 6 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 110-113
Towards a Lightweight Platform for Human-Robot Interaction in Federated Edge and IoT Environments
- 6 May 2025
- Published by Association for Computing Machinery (ACM)
- p. 110-113
Open AccessTowards a Lightweight Platform for Human-Robot Interaction in Federated Edge and IoT Environments
- 29 July 2024
- Published by Elsevier in Neurocomputing
- Vol. 602
WSSGCN: Wide Sub-stage Graph Convolutional Networks
- 29 July 2024
- Published by Elsevier in Neurocomputing
- Vol. 602
WSSGCN: Wide Sub-stage Graph Convolutional Networks
- 21 March 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 34 (8) , 2464-2478
Adaptive Fragment-Based Parallel State Recovery for Stream Processing Systems
Today, large-scale cloud organizations are deploying datacenters and “edge” clusters globally to provide low-latency access to services. Running stream applications across geo-distributed sites are emerging as a daily requirement. However, existing efforts have dominantly centered around stateless stream processing , leaving another urgent trend- stateful stream processing -much less explored. A driving need is to store and update states during processing, and most importantly, successfully recover large distributed states when faults and failures happen. Existing studies exhibit major limitations including: (1) they mostly inherit MapReduce's “single master/many workers” architecture, where the central master can easily become ascalability bottleneck; (2) they offer state recovery mainly through three approaches: replication recovery, checkpointing recovery, and DStream-based lineage recovery, which are either slow, resource-expensive or failing to handle multiple failures; and (3) they are not adaptive to heterogeneous hardware settings. We present A-FP4S, a novel adaptive fragments-based parallel state recovery mechanism for stream processing systems. A-FP4S organizes stream operators into a distributed hash table based peer-to-peer overlay and divides each node's local state into many fragments. These fragments are periodically stored in node's multiple neighbors, ensuring different sets of available fragments can reconstruct failed states in parallel. This mechanism is extremely scalable to the lost state, significantly reduces failure recovery time, and can tolerate multiple node failures. A-FP4S is adaptive to heterogeneous hardware settings by automatic parameter tuning over phases. Compared to Apache Storm, A-FP4S achieves 31.8% to 50.5% reduction in recovery latency. Large-scale experiments using real-world datasets demonstrate A-FP4S's attractive scalability and adaptivity properties.
- 21 March 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 34 (8) , 2464-2478
Adaptive Fragment-Based Parallel State Recovery for Stream Processing Systems
Today, large-scale cloud organizations are deploying datacenters and “edge” clusters globally to provide low-latency access to services. Running stream applications across geo-distributed sites are emerging as a daily requirement. However, existing efforts have dominantly centered around stateless stream processing , leaving another urgent trend- stateful stream processing -much less explored. A driving need is to store and update states during processing, and most importantly, successfully recover large distributed states when faults and failures happen. Existing studies exhibit major limitations including: (1) they mostly inherit MapReduce's “single master/many workers” architecture, where the central master can easily become ascalability bottleneck; (2) they offer state recovery mainly through three approaches: replication recovery, checkpointing recovery, and DStream-based lineage recovery, which are either slow, resource-expensive or failing to handle multiple failures; and (3) they are not adaptive to heterogeneous hardware settings. We present A-FP4S, a novel adaptive fragments-based parallel state recovery mechanism for stream processing systems. A-FP4S organizes stream operators into a distributed hash table based peer-to-peer overlay and divides each node's local state into many fragments. These fragments are periodically stored in node's multiple neighbors, ensuring different sets of available fragments can reconstruct failed states in parallel. This mechanism is extremely scalable to the lost state, significantly reduces failure recovery time, and can tolerate multiple node failures. A-FP4S is adaptive to heterogeneous hardware settings by automatic parameter tuning over phases. Compared to Apache Storm, A-FP4S achieves 31.8% to 50.5% reduction in recovery latency. Large-scale experiments using real-world datasets demonstrate A-FP4S's attractive scalability and adaptivity properties.
- Open Access
- 12 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 7768-7781
Achieving Online and Scalable Information Integrity by Harnessing Social Spam Correlations
Malicious web links, social rumors, fraudulent advertisements, faked comments, and biased propaganda are overwhelmingly influencing online social networks. Enabling information integrity is a hot topic in both academia and industry. Traditional social spam detection techniques rely on centralized processing, focusing only on one specific set of data sources, thereby ignoring the social spam correlations between distributed data sources. In this paper, we propose an online and scalable misinformation detection system, named Spiral, to uncover social spam by leveraging the correlations between different social data sources in geo-distributed sites. The key insight in our approach is to amplify the effectiveness of state-of-the-art techniques to detect inappropriate posts by enabling the efficient large-scale propagation of detection information across domains. The novelty of our design lies in three key components: (1) a decentralized distributed hash-table-based tree overlay deployment for harvesting and uncovering deceptive information spreading in multiple online social networks communities; (2) a progressive aggregation tree for collecting the properties of these posts and creating new classifiers to actively filter out the propagation of inappropriate posts; and (3) a group communication structure that allows multiple groups to exchange the correlations among distributed social data sources. We designed and implemented a prototype of the Spiral system. Our large-scale experiments, using real-world social data, demonstrate Spiral’s scalability, effective load-balancing, and efficiency in online spam detection for social networks.
- 12 January 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 7768-7781
Open AccessAchieving Online and Scalable Information Integrity by Harnessing Social Spam Correlations
Malicious web links, social rumors, fraudulent advertisements, faked comments, and biased propaganda are overwhelmingly influencing online social networks. Enabling information integrity is a hot topic in both academia and industry. Traditional social spam detection techniques rely on centralized processing, focusing only on one specific set of data sources, thereby ignoring the social spam correlations between distributed data sources. In this paper, we propose an online and scalable misinformation detection system, named Spiral, to uncover social spam by leveraging the correlations between different social data sources in geo-distributed sites. The key insight in our approach is to amplify the effectiveness of state-of-the-art techniques to detect inappropriate posts by enabling the efficient large-scale propagation of detection information across domains. The novelty of our design lies in three key components: (1) a decentralized distributed hash-table-based tree overlay deployment for harvesting and uncovering deceptive information spreading in multiple online social networks communities; (2) a progressive aggregation tree for collecting the properties of these posts and creating new classifiers to actively filter out the propagation of inappropriate posts; and (3) a group communication structure that allows multiple groups to exchange the correlations among distributed social data sources. We designed and implemented a prototype of the Spiral system. Our large-scale experiments, using real-world social data, demonstrate Spiral’s scalability, effective load-balancing, and efficiency in online spam detection for social networks.
- 14 December 2022
- Published by Springer Nature
- p. 13-21
Towards an Efficient Client Selection System for Federated Learning
- 14 December 2022
- Published by Springer Nature
- p. 13-21
Towards an Efficient Client Selection System for Federated Learning