Xitong Gao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Xitong Gao has a total of 96 co-authors. Here are the top co-authors by number of shared publications:
- Cheng-Zhong Xu
State Key Laboratory of IoTSC, University of Macau, Macau, Macao
16 shared publications - Kejiang Ye
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences
16 shared publications - Juanjuan Zhao
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences
11 shared publications - Dongping Liao
Shenzhen Institutes of Advanced Technology Shenzhen Institutes of Advanced Technology Shenzhen Institutes of Advanced Technology University of Chinese Academy of Sciences Chinese Academy of Sciences University of Macau Imperial College London Chinese Academy of Sciences Shenzhen University of Advanced Technology University of Macau Chinese Academy of Sciences Shenzhen University of Advanced Technology
5 shared publications Kejiang Ye
Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences
5 shared publicationsYiren Zhao
Imperial College London
4 shared publications
Latest Publications
- 18 August 2026
- Published by Springer Nature
- p. 128-143
Asynchronous Speculative Sentence Decoding with Step-Level Alignment Classification
- 18 August 2026
- Published by Springer Nature
- p. 128-143
Asynchronous Speculative Sentence Decoding with Step-Level Alignment Classification
- Open Access
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Lyra: A Hardware-Accelerated RISC-V Verification Framework with Generative Model-Based Processor Fuzzing
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessLyra: A Hardware-Accelerated RISC-V Verification Framework with Generative Model-Based Processor Fuzzing
- Open Access
- 8 December 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 20 (1) , 1-23
Flareon: Stealthy All2all Backdoor Injection via Poisoned Augmentation
Open source software supply-chain attacks, once successful, can exact heavy costs in mission-critical applications. As open source ecosystems for deep learning flourish and become increasingly universal, they present attackers previously unexplored avenues to code-inject malicious backdoors in deep neural network models. This article proposes Flareon , a small, stealthy, seemingly harmless code modification that specifically targets the data augmentation pipeline with motion-based triggers. Flareon neither alters ground-truth labels, nor modifies the training loss objective, nor does it assume prior knowledge of the victim model architecture, training data, and training hyperparameters. Yet, it has a surprisingly large ramification on training—models trained under Flareon learn powerful target-conditioned (or “ all2all ”) backdoors. We also proposed a learnable variant of Flareon that is even stealthier in terms of added perturbations. The resulting models can exhibit high attack success rates for any target choices and better clean accuracies than backdoor attacks that not only seize greater control but also assume more restrictive attack capabilities. We also demonstrate the resilience of Flareon against a wide range of defenses. Flareon is fully open source and available online to the deep learning community.
- 8 December 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Knowledge Discovery From Data
- Vol. 20 (1) , 1-23
Open AccessFlareon: Stealthy All2all Backdoor Injection via Poisoned Augmentation
Open source software supply-chain attacks, once successful, can exact heavy costs in mission-critical applications. As open source ecosystems for deep learning flourish and become increasingly universal, they present attackers previously unexplored avenues to code-inject malicious backdoors in deep neural network models. This article proposes Flareon , a small, stealthy, seemingly harmless code modification that specifically targets the data augmentation pipeline with motion-based triggers. Flareon neither alters ground-truth labels, nor modifies the training loss objective, nor does it assume prior knowledge of the victim model architecture, training data, and training hyperparameters. Yet, it has a surprisingly large ramification on training—models trained under Flareon learn powerful target-conditioned (or “ all2all ”) backdoors. We also proposed a learnable variant of Flareon that is even stealthier in terms of added perturbations. The resulting models can exhibit high attack success rates for any target choices and better clean accuracies than backdoor attacks that not only seize greater control but also assume more restrictive attack capabilities. We also demonstrate the resilience of Flareon against a wide range of defenses. Flareon is fully open source and available online to the deep learning community.
- 17 November 2025
- Published by Springer Nature
- p. 301-316
StreamGuard: A Streaming-Based Defense Against Jailbreaking Attacks in Large Language Models
- 17 November 2025
- Published by Springer Nature
- p. 301-316
StreamGuard: A Streaming-Based Defense Against Jailbreaking Attacks in Large Language Models
- 30 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (11) , 19626-19644
Urban Transport Mode Split Prediction: A Hybrid Deep Learning Framework Considering Spatiotemporal Dependency
Transport Mode Split (TMS) represents the distribution of trips among transport modes between city regions. Accurate TMS prediction is crucial for urban planning and traffic management. Traditional methods, such as curve models and discrete choice models, often fail to capture user travel mode pReferences due to dataset limitations and inadequate spatiotemporal modeling. While deep learning models have been applied in related domains such as ride-sharing and traffic prediction, they typically focus on a single transport mode and do not model the competitive relationships between modes. We propose a novel Deep Learning-based Framework for Transport Mode Split Prediction (DMSP), integrating restricted-route (bus/subway) and free-route (taxi) data. It introduces two key innovations: 1) efficient extraction of TMS samples from sparse raw data, enriched with spatiotemporal context, and 2) a prediction model extending discrete choice theory via maximum utility principles. Mode utilities are estimated using CNNs for travel mode-specific effects and GNN for spatiotemporal dependencies. Experiments on six months of Shenzhen data show that DMSP reduces MAPE by over 4% compared to existing methods.
- 30 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 26 (11) , 19626-19644
Urban Transport Mode Split Prediction: A Hybrid Deep Learning Framework Considering Spatiotemporal Dependency
Transport Mode Split (TMS) represents the distribution of trips among transport modes between city regions. Accurate TMS prediction is crucial for urban planning and traffic management. Traditional methods, such as curve models and discrete choice models, often fail to capture user travel mode pReferences due to dataset limitations and inadequate spatiotemporal modeling. While deep learning models have been applied in related domains such as ride-sharing and traffic prediction, they typically focus on a single transport mode and do not model the competitive relationships between modes. We propose a novel Deep Learning-based Framework for Transport Mode Split Prediction (DMSP), integrating restricted-route (bus/subway) and free-route (taxi) data. It introduces two key innovations: 1) efficient extraction of TMS samples from sparse raw data, enriched with spatiotemporal context, and 2) a prediction model extending discrete choice theory via maximum utility principles. Mode utilities are estimated using CNNs for travel mode-specific effects and GNN for spatiotemporal dependencies. Experiments on six months of Shenzhen data show that DMSP reduces MAPE by over 4% compared to existing methods.
- Open Access
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1579-1587
BridgeNet: A Unified Multimodal Framework for Bridging 2D and 3D Industrial Anomaly Detection
- 27 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1579-1587
Open AccessBridgeNet: A Unified Multimodal Framework for Bridging 2D and 3D Industrial Anomaly Detection
- 22 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (3) , 4241-4252
MAIR: Model Agnostic Instance Reweighing for Heterogeneous Federated Learning
Federated learning (FL) enables collaborative training on decentralized data while preserving the data owners’ privacy, under the orchestration of a central server. FL has seen tremendous growth and advancements in recent years. Despite its progress, FL faces a significant challenge raised by data heterogeneity, leading to a slower convergence rate and a larger performance gap compared to centralized training. In this work, we empirically reveal that direct applying empirical risk minimizing (ERM) on skewed client training data causes the client model suffers from biased predictions towards majority classes. To address this problem, we propose a model agnostic instance reweighing method (MAIR). At a coarse-grained level, MAIR adjusts the logits predictions for each class to counteract the data heterogeneity. At a fine-grained level, it dynamically reweighs the importance of individual training samples with a predictive meta network. As a results, MAIR prevents client models from over-fitting on heterogeneous data and therefore substantially reduces client drift. Theoretically, we justify its non-convex convergence property. Extensive experiments demonstrate that MAIR reliably speeds up convergence and improves the quality of global models, outperforming its best competitor by a clear margin. It notably delivers$8.3\%$ improvements on ImageNet subset and achieves$67.6\%$ energy footprint reduction on CIFAR-100 over the FedAvg baseline. Our findings also suggest that improving the performance of FL-trained models necessitates rethinking clients’ local optimization objectives, and ERM should thus no longer be viewed as a de facto standard in FL under data heterogeneity.- 22 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Mobile Computing
- Vol. 25 (3) , 4241-4252
MAIR: Model Agnostic Instance Reweighing for Heterogeneous Federated Learning
Federated learning (FL) enables collaborative training on decentralized data while preserving the data owners’ privacy, under the orchestration of a central server. FL has seen tremendous growth and advancements in recent years. Despite its progress, FL faces a significant challenge raised by data heterogeneity, leading to a slower convergence rate and a larger performance gap compared to centralized training. In this work, we empirically reveal that direct applying empirical risk minimizing (ERM) on skewed client training data causes the client model suffers from biased predictions towards majority classes. To address this problem, we propose a model agnostic instance reweighing method (MAIR). At a coarse-grained level, MAIR adjusts the logits predictions for each class to counteract the data heterogeneity. At a fine-grained level, it dynamically reweighs the importance of individual training samples with a predictive meta network. As a results, MAIR prevents client models from over-fitting on heterogeneous data and therefore substantially reduces client drift. Theoretically, we justify its non-convex convergence property. Extensive experiments demonstrate that MAIR reliably speeds up convergence and improves the quality of global models, outperforming its best competitor by a clear margin. It notably delivers$8.3\%$ improvements on ImageNet subset and achieves$67.6\%$ energy footprint reduction on CIFAR-100 over the FedAvg baseline. Our findings also suggest that improving the performance of FL-trained models necessitates rethinking clients’ local optimization objectives, and ERM should thus no longer be viewed as a de facto standard in FL under data heterogeneity.- Open Access
Flash-Searcher: Fast and Effective Web Agents via DAG-Based Parallel Execution
Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks when equipped with external tools. However, current frameworks predominantly rely on sequential processing, leading to inefficient execution particularly for tasks requiring extensive tool interaction. This paper introduces Flash-Searcher, a novel parallel agent reasoning framework that fundamentally reimagines the execution paradigm from sequential chains to directed acyclic graphs (DAGs). Flash-Searcher decomposes complex tasks into subtasks with explicit dependencies, enabling concurrent execution of independent reasoning paths while maintaining logical constraints. Through dynamic workflow optimization, our framework continuously refines the execution graph based on intermediate results, effectively integrating summary module. Comprehensive evaluations across multiple benchmarks demonstrate that Flash-Searcher consistently outperforms existing approaches. Specifically, it achieves 67.7% accuracy on BrowseComp and 83% on xbench-DeepSearch, while reducing agent execution steps by up to 35% compared to current frameworks. Furthermore, when distilling this parallel reasoning pipeline into single models, we observe substantial performance gains across diverse backbone architectures, underscoring the generalizability of our methodology. Our work thus represents a significant advance in agent architecture design, offering a more scalable and efficient paradigm for complex reasoning tasks.
Open AccessFlash-Searcher: Fast and Effective Web Agents via DAG-Based Parallel Execution
Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks when equipped with external tools. However, current frameworks predominantly rely on sequential processing, leading to inefficient execution particularly for tasks requiring extensive tool interaction. This paper introduces Flash-Searcher, a novel parallel agent reasoning framework that fundamentally reimagines the execution paradigm from sequential chains to directed acyclic graphs (DAGs). Flash-Searcher decomposes complex tasks into subtasks with explicit dependencies, enabling concurrent execution of independent reasoning paths while maintaining logical constraints. Through dynamic workflow optimization, our framework continuously refines the execution graph based on intermediate results, effectively integrating summary module. Comprehensive evaluations across multiple benchmarks demonstrate that Flash-Searcher consistently outperforms existing approaches. Specifically, it achieves 67.7% accuracy on BrowseComp and 83% on xbench-DeepSearch, while reducing agent execution steps by up to 35% compared to current frameworks. Furthermore, when distilling this parallel reasoning pipeline into single models, we observe substantial performance gains across diverse backbone architectures, underscoring the generalizability of our methodology. Our work thus represents a significant advance in agent architecture design, offering a more scalable and efficient paradigm for complex reasoning tasks.
- Open Access
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1118-1127
On the Adversarial Robustness of Visual-Language Chat Models
- 30 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1118-1127
Open AccessOn the Adversarial Robustness of Visual-Language Chat Models