Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks105 from public data
- A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deployment139
This paper introduces the concept of "full-stack" safety to systematically consider safety issues throughout the entire process of LLM training, deployment, and eventual commercialization, and represents the first safety survey to encompass the entire lifecycle of LLMs.
- A Survey on Trustworthy LLM Agents: Threats and Countermeasures137
The TrustAgent framework is proposed, a comprehensive study on the trustworthiness of agents, characterized by modular taxonomy, multi-dimensional connotations, and technical implementation, aiming to provide guidance for future endeavors.
- SafeGen: Mitigating Sexually Explicit Content Generation in Text-to-Image Models82
SafeGen is presented, a framework to mitigate sexual content generation by text-to-image models in a text-agnostic manner that outperforms eight state-of-the-art baseline methods and achieves 99.4% sexual content removal performance.
- SafeEar: Content Privacy-Preserving Audio Deepfake Detection70
SafeEar is proposed, a novel framework that aims to detect deepfake audios without relying on accessing the speech content within by devise a neural audio codec into a novel decoupling model that well separates the semantic and acoustic information from audio samples, and only use the acoustic information for deepfake detection.
- Attack Vulnerability of Network Controllability62
This paper systematically investigates the attack vulnerability of network controllability for the canonical model networks as well as the real-world networks subject to attacks on nodes and edges, and finds that the edge degree is not a good quantity to measure the importance of an edge in terms of network control.
- 57
- EarArray: Defending against DolphinAttack via Acoustic Attenuation53
EarArray is a lightweight method that can not only detect inaudible voice commands but also identify the direction of attackers without requiring any extra hardware or hardware modification.
- EV-Loc: integrating electronic and visual signals for accurate localization49
In EV-Loc, a new localization technique is proposed that uses visual signals to help improve the accuracy of wireless localization and fully leverages visual signals' high accuracy and electronic signals' pervasiveness.
- 46
- Teaching and Learning Computational Thinking Through Game-Based Learning: A Systematic Review43
Findings suggested that GBL has positive effects on CT, but has non-significant effects on some CT elements because of limited time to learn these elements and students’ preferences for using CT elements.
- 35
- 24
- Legilimens: Practical and Unified Content Moderation for Large Language Model Services23
It is revealed for the first time that effective and efficient content moderation can be achieved by extracting conceptual features from chat-oriented LLMs, despite their initial fine-tuning for conversation rather than content moderation.
- 23
- SurvSurf: human retrieval on large surveillance video data22
SurvSurf is designed, a human retrieval system on large surveillance video data that exploits characteristics of these data and big data processing tools and uses the MapReduce framework to process M-clips in parallel for human detection and appearance/motion feature extraction.
- Plant-assisted rhizoremediation of decabromodiphenyl ether for e-waste recycling area soil of Taizhou, China22
A possibility of phytoremediation for BDE-209-polluted soils by successive cultivation of rice followed by ryegrass coupling with suitable water management is illustrated, possibly through dissipation pathway of microbial reductive debromination and subsequent aerobic oxidative cleavage of benzene ring.
- 21
- 18
- Enrollment-Stage Backdoor Attacks on Speaker Recognition Systems via Adversarial Ultrasound17
Tuner is proposed, a new type of backdoor attack against the enrollment stage of SRS via adversarial ultrasound modulation, which is inaudible, synchronization-free, content-independent, and black-box, and can successfully bypass SRSs while remaining effective to various speakers, speech content, etc.
- 16
- 15
- The Eye of Sherlock Holmes: Uncovering User Private Attribute Profiling via Vision-Language Model Agentic Framework15
PAPI is constructed, the largest dataset for studying private attribute profiling in personal images, comprising 2,510 images from 251 individuals with 3,012 annotated privacy attributes and HolmesEye is proposed, a hybrid agentic framework that combines VLMs and LLMs to enhance privacy inference.
- Optimizing the controllability of arbitrary networks with genetic algorithm14
An efficient genetic algorithm oriented optimization tool to optimize the controllability of arbitrary networks consisting of both state nodes and control nodes under Popov–Belevitch–Hautus rank condition is developed.
- 13
- 12
- 11
- Learning to Rank Chain-of-Thought: Using a Small Model10
EORM effectively leverages a given pool of candidate solutions to match or exceed the performance of brute-force sampling, thereby enhancing LLM reasoning outcome reliability through its streamlined post-hoc verification process.
- Data-driven few-shot crop pest detection based on object pyramid for smart agriculture10
A less-shot object detection method in crop pest detection based on an object pyramid called multiscale positive sample refinement is introduced, which can give some new insight to the future scholars, which is of great significance to the development of smart agriculture.
- The Silent Manipulator: A Practical and Inaudible Backdoor Attack against Speech Recognition Systems10
SMA is proposed, an inaudible grey-box backdoor attack that can be generalized to real-world scenarios where victims are present by exploiting both the vulnerability of microphones and neural networks, and which can resist typical defense countermeasures to backdoor attacks.
- 10
- 9
- Niche Genetic Algorithm Based on Sexual Reproduction and Multimodal Function Optimization Problem9
The testing for Rastrigin function and Shubert function proves that the niche genetic algorithm based on sexual reproduction is effective and can combat premature convergence and keep the diversity of population.
- Efficient Reasoning via Chain of Unconscious Thought8
This work proposes a new reasoning paradigm, termed Chain of Unconscious Thought (CoUT), to improve the token efficiency of large Reasoning Models by guiding them to mimic human unconscious thought and internalize reasoning processes.
- DynamicNER: A Dynamic, Multilingual, and Fine-Grained Dataset for LLM-based Named Entity Recognition7
DynamicNER is proposed, the first NER dataset designed for LLM-based methods with dynamic categorization, introducing various entity types and entity type lists for the same entity in different context, leveraging the generalization of LLM-based NER better.
- The Controllability of Power Grids in Comparison with Classical Complex Network Models7
The main conclusion is that most driver nodes of power grids are hub-free nodes with low nodal degree values of 1 or 2, and power grids is harder to control than WS networks and CM networks while easier than BA networks.
- 7
- 6
- 6
- 6
- EmoRAG: Evaluating RAG Robustness to Symbolic Perturbations5
It is demonstrated that injecting a single emoticon into a query makes it nearly 100% likely to retrieve semantically unrelated texts, which contain a matching emoticon, termed EmoRAG.
- Mitigating Over-Refusal in Aligned Large Language Models via Inference-Time Activation Energy4
This work establishes a promising paradigm for building LLMs that simultaneously achieve high safety and low false refusal rates, and establishes a novel, fine-tuning free framework designed to resolve this challenge through dynamic, inference-time intervention.
- 4
- PATFinger: Prompt-Adapted Transferable Fingerprinting against Unauthorized Multimodal Dataset Usage4
A novel prompt-adapted transferable fingerprinting scheme from a training-free perspective, called PATFinger, which incorporates the global optimal perturbation and the adaptive prompts to capture dataset-specific distribution characteristics and allows the dataset owner to check the usage of datasets by observing specific prediction behaviors linked to the PATFinger during retrieval queries.
- 4
- 4
- 4
- Controllability of deterministic complex networks4
This paper examines six typical deterministic networks and investigates the characteristics of driver nodes in deterministic complex networks, finding that the driver nodes tend to avoid high degree nodes but to have high clustering coefficients.
- 3
- 3
- 3
- How Centrality of Driver Nodes Affects Controllability of Complex Networks3
This paper systematically investigates the nodal centrality of driver nodes in controlling complex networks and finds that the driver nodes tend to be low in-degree but high out-degree nodes, and most of driver nodes tend to have low betweenness centrality but relatively high closeness centrality.
- 3
- 2
- BIGbench: A Unified Benchmark for Evaluating Multi-dimensional Social Biases in Text-to-Image Models2
- Edge features extraction from 3D laser point cloud based on corresponding images2
The experimental results demonstrate that the method guarantees accurate fine extraction of edge features from 3D laser point cloud based on corresponding images using gray moment algorithm.
- Enclave: Promoting Unobtrusive and Secure Mobile Communications with a Ubiquitous Electronic World2
This work proposes Enclave, a delegate wireless device that helps people’s smartphones communicate unobtrusively and demonstrates its potential for smartphone data protection and unobtrusive and secure communication.
- 2
- 1
- 1
- 1
- 1
- Assessment and Prediction of Soil Fertility in Urban Areas of the Loess Plateau Based on Machine Learning Methods1
A theoretical basis and scientific support for large‐scale SFI prediction is provided and normalized difference vegetation index (NDVI) and soil organic carbon density (SOCD) emerged as the dominant predictors, collectively contributing over 80% to SFI prediction accuracy.
- 1
- Heuristic-Induced Multimodal Risk Distribution Jailbreak Attack for Multimodal Large Language Models1
- 1
- 1
- 1
- 1
- Time-sensitive Information Communication, Sensing, and Computing in Cyber-Physical Systems1
This dissertation proposes TurfCast, a novel information dissemination service that selectively broadcasts messages based on receivers’ lingering time and introduces TurfBurst, which exploits the Shannon bound to differentiate among receivers based on spatial areas.
- Simulation of Tumor Detection Based on Bioelectrical Impedance Measurement1
A simplified tumor model is developed, the electrical impedance value changes in the model’s different levels are measured, and the theoretical basis for the tumor impedance detection system is provided.
- Study of digital watermark on wavelet transform and the realization on Matlab simulation1
The status of the technique of digital watermark and its fundamental is summed up, the advantage of the digital water mark on wavelet transform is analyzed and the process of embed or get watermark is expatiated.
- Cybersecurity vulnerabilities in IoT devices–
This Review provides a comprehensive overview of IoT device vulnerabilities and countermeasures, contrasting their characteristics with traditional computer vulnerabilities, and proposes a new taxonomy that categorizes IoT device vulnerabilities into in-band and out-of-band classifications, emphasizing emerging out-of-band vulnerabilities.
- HolmesEye: Profiling Private Attributes with Vision-Language Agents–
This work builds PAPI, a benchmark with 2,510 images from 251 individuals and 3,012 annotations over 12 attributes, and proposes HolmesEye, a VLM–LLM agentic framework that extracts image-level cues, reasons over cross-image patterns, asks targeted visual questions, and produces person-level predictions.
- –
- –
- AP60: A Taxonomy-Guided Benchmark Dataset for Fine-Grained Pest Recognition with Feature-Level Confusion Analysis–
A feature-level confusion analysis framework is introduced to characterize the intrinsic visual structure of AP60 from two complementary aspects: intra-class consistency and inter-class overlap, which suggests that higher intra-class similarity is associated with better class-level accuracy, whereas lower inter-class separability is associated with increased misclassification.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- A New Automatic Quality Control System for Fiber Ring Production–
A new methodology for automatic production quality control is proposed which uses the computer vision technology to apply a system which can distinguish real-time fiber ring production images, applying pattern recognition of data mining technology for understanding of the type of fault production image.
- Research on loss recovery of application layer multicast–
This paper analyzes the error and delivery features of ALM trees, and further presents a data loss recovery solution (called HBHLR) for application layer multicast.
- Encoding Techniques in Delay/Disruption Tolerant Networks–
This paper first introduces the concepts related to a space information network, and then focuses on one of its branches, i.e., delay/disruption-tolerant networks (DTN), and emphasizes on introducing the coding techniques in DTN.
- –
- –
- –
- Analysis of factors influencing physical health standard of ordinary university students——Taking Henan province as an example–
The factor of physical teaching plays the most important role in improving university students' physical health standard, and then the factor of students themselves, the factors of students' family and thefactor of social environment are analyzed.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Consideration on Development and Reform of Teaching Material of Competitive Events in Our College Sport from the Prospect of Health First–
Based on teaching practice, the author discusses the corresponding measurement in order to provide a unique research perspective for college physical education reform.
- –
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.