Speaker anonymization aims to modify the speech signal in order to protect the identity of a speaker while preserving the linguistic content. Despite the increasing use of children's voices in educational applications, such as oral reading fluency (ORF) assessment, there is little work on the anonymization aspects. In this work, we investigate the effectiveness of available speaker anonymization methods drawing from traditional speech-production based approaches and a neural codec based method. We investigate the trade-off between privacy protection, measured as the degree of anonymity, and utility preservation, which in the current context of ORF assessment, includes the segmental and suprasegmental features of children’s read speech utterances. We report objective and subjective evaluations using two child-speaker datasets: MPS and SpeechOcean. Our objective evaluation results indicate that the speech-production based method of vocal tract length normalization coupled with pitch-transposition achieves the best balance between privacy and utility. Subjective listening results indicate that naturalness is achievable across methods while the neural method fails to preserve age characteristics, which are more easily controlled by the speech-production driven methods.
论文检索
输入标题、作者或关键词，从 3,069 篇学术成果中精准定位
Building AI Hardware Expertise: Edge AI Curriculum Design and Implementation in German Universities
PDF ↗Edge artificial intelligence (AI) redistributes AI computation from distant cloud to local processors for real-time processing and enhanced privacy. This fundamental shift underscores a critical gap in current university curricula, which predominantly focus on AI fundamentals and algorithms while often neglecting essential AI hardware topics. To address this deficiency, this paper presents Edge AI, a postgraduate curriculum co-designed by two universities in Germany. Guided by the Dagstuhl triangle, the curriculum is designed to comprehensively cover technical, sociocultural, and application perspectives. Courses are developed using the Four-Component Instructional Design model to encourage action-oriented skill development, with a Learning Management System template available to assist in the design of individual courses. Selected practical courses are formulated as self-managed projects and inverted classrooms, enabling students to learn at their own pace with just-in-time guidance. All curriculum materials are accessible online and maintained by the Open Science Framework to enhance collaboration across institutions and promote applicability in diverse domains. Evaluation results from 176 students over two years (2023-2025) demonstrate universal satisfaction across various curriculum components.
Large Language Models (LLMs) have shown remarkable success in supporting a wide range of knowledge-intensive tasks. In specialized domains, there is growing interest in leveraging LLMs to assist subject matter experts with domain-specific challenges. However, deploying LLMs as SaaS solutions raises data privacy concerns, while many open-source models demand significant computational resources for effective domain adaptation and deployment. A promising alternative is to develop smaller, domain-specialized LLMs, though this approach is often constrained by the lack of high-quality domain-specific training data. In this work, we address these limitations by presenting a cost-efficient and scalable training pipeline that combines guided synthetic data generation from a small seed corpus with bottom-up domain data curation. Our pipeline integrates Domain-Adaptive Pretraining (DAPT), Domain-specific Supervised Fine-tuning (DSFT), and Direct Preference Optimization (DPO) to train effective small-scale models for specialized use cases. We demonstrate this approach through DiagnosticSLM, a 3B-parameter domain-specific model tailored for fault diagnosis, root cause analysis, and repair recommendation in industrial settings. To evaluate model performance, we introduce four domain-specific benchmarks: multiple-choice questions (DiagnosticMCQ), question answering (DiagnosticQA), sentence completion (DiagnosticComp), and summarization (DiagnosticSum). DiagnosticSLM achieves up to 25% accuracy improvement over open-source models of comparable or larger size (2B-9B) on the MCQ task, while also outperforming or matching them in other tasks, demonstrating effective domain-specific reasoning and generalization capabilities.
Spatial Graph Attention Network Modeling for Neighborhood-Scale Lead Contamination Risk Prediction Using Publicly Available Data
PDF ↗Lead contamination in urban water systems remains a prevalent public health threat, affecting millions of American households and disproportionately endangering vulnerable population groups. Current municipal risk assessment and inspection strategies are overwhelmingly based on random sampling and complaint-driven protocols that overlook spatial complexity, reinforce inequities, and squander limited resources, leaving critical exposure areas unidentified. This paper presents a lead contamination risk prediction framework from socio-demographic housing features analytics, first of its kind, by drawing on partially anonymized residential testing data as ground truth and applying graph neural networks alongside gradient-boosted ensembles. Specifically, our method integrates spatial Deep Graph Attention Networks classifiers to capture inter-neighborhood contamination dependencies, fuse demographic and spatial evidence, and produce interpretable risk scores. Those scores are actionable by municipal water authorities at the intra-neighborhood level. Through extensive experiments on newly constructed Chicago block-group level datasets, our framework achieves a balanced accuracy of 84.8% and reduces false positive lead contamination by up to 44% versus spatial-only baselines and 21% over current practice, without sacrificing recall on contaminated blocks. Our approach not only extends technical boundaries in spatial-ensemble learning and privacy-preserving urban health modeling, but also provides policymakers and public health officials with a means to assess and address contamination risks, supporting efforts to protect community health and safety.
Beyond Neuron-Level Sparsity: Achieving Faithful and Interpretable LLMs with Mixture of Decoders
PDF ↗As large language models (LLMs) scale, ensuring interpretability and privacy becomes critical. This talk addresses these interconnected challenges with novel approaches to model specialization and safety. First, we tackle the dense, distributed nature of LLM representations by casting Mixture-of-Experts (MoE) as a tensor decomposition, enabling specialized experts in a factorized space. Second, we argue that current neuron-level sparsity methods create a severe accuracy-sparsity trade-off, and we propose a paradigm shift to layer-level sparsity with the Mixture of Decoders (MxD). We explain how MxD uses tensor factorization to expand dense layers into thousands of specialized, full-rank sublayers, demonstrating how it significantly outperforms alternatives in preserving model faithfulness and performance across LLMs up to 3B parameters. Finally, we address privacy in open-weight models by proposing a scalable and certifiable algorithm that induces maximal uncertainty on protected instances, proving tight bounds that characterize the resulting privacy-utility tradeoff.
Large language models (LLMs) have achieved remarkable success in many domains, but concerns about data quality and privacy are growing. Federated Learning (FL) offers a privacy-preserving solution by training a model on local clients without sharing data. However, the impact of biased private data on LLMs fine-tuned through FL remains understudied. This work investigates how client-side biased data affects the global model during federated fine-tuning of LLMs. We simulate realistic scenarios where some clients possess datasets containing social biases (stereotypes, discriminatory language) while others have clean data through extensive experiments with popular FL algorithms (FedAvg, FedAdam and FedProx) and popular LLMs (LLaMA, Mistral, Phi-3 and Gemma) across datasets with varying bias proportions (33%, 66%, 100%). Our findings reveal that 1) FedAdam consistently shows the lowest bias propagation, reducing CrowS-Pairs scores by up to 15% compared to FedAvg; 2) Even small amounts of biased data (33%) can significantly influence global model bias; 3) Mixed biased and neutral data distributions lead to 5%-7% higher bias scores than segregated distributions. Additionally, we propose Bias-Aware Model Aggregation (BAMA), a novel debiasing method for federated fine-tuning that consistently reduces bias across various models and algorithms.
Graph neural networks (GNNs) excel at modeling graph-structured data but often inherit and amplify biases, leading to substantial efforts in developing fair GNNs. However, most existing approaches assume full access to sensitive attribute information, which is often impractical in real-world scenarios due to privacy concerns or risks of discrimination. To address this limitation, this paper focuses on graph fairness with limited sensitive attribute information, ensuring applicability to real-world contexts where current methods fall short. Specifically, we introduce an innovative fairness optimization strategy, propose a novel framework named FGLISA, and provide a theoretical perspective linking limited sensitive attribute information access to fairness objectives, thus enabling fair graph learning in real-world applications with limited sensitive attribute information. Experiments on diverse real-world datasets and tasks validate the effectiveness of our approach in achieving both fairness and predictive performance.
AXON: Action Characterization Through Cross-Modal Knowledge Distillation for Neurodiverse Individuals
PDF ↗Understanding the communicative behaviors of non- and minimally-speaking individuals with autism spectrum disorder (ASD) and complex neurodevelopmental disorders (NDDs) remains a critical challenge for both clinical support and machine learning (ML) research. However, developing automated systems for this task is hindered by data scarcity, privacy concerns, heterogeneous and idiosyncratic behaviors, and the significant domain shift from neurotypical to neurodiverse populations. To address these challenges, we first present a novel, large-scale, privacy-preserving action recognition dataset with 2,721 3D skeleton samples capturing in-home interactions of individuals with ASD and complex NDDs. Second, we propose AXON, a novel cross-modal knowledge distillation method that transfers the rich semantic understanding of a pre-trained CLIP model to a graph-based Hyperformer model, outperforming other cross-modal knowledge distillation baselines in action recognition. We further introduce a gradient-based interpretability method to characterize how individuals with ASD and complex NDDs perform communicative actions. Our analysis uncovers both individual- and population-level communicative styles, tendencies, and biases. Our foundational study helps spur the development of more adaptive and personalized augmentative technologies, aiming to foster greater communicative autonomy and understanding for this underserved population.
AURA: Development and Validation of an Augmented Unplanned Removal Alert System Using Synthetic ICU Videos
PDF ↗Unplanned extubation (UE)—the unintended removal of an airway tube—remains a critical patient safety concern in intensive care units (ICUs), often leading to severe complications or death. Real-time UE detection has been limited, largely due to the ethical and privacy challenges of obtaining annotated ICU video data. We propose Augmented Unplanned Removal Alert (AURA), a vision-based risk detection system developed and validated entirely on a fully synthetic video dataset. By leveraging text-to-video diffusion, we generated diverse and clinically realistic ICU scenarios capturing a range of patient behaviors and care contexts. The system applies pose estimation to identify two high-risk movement patterns: collision, defined as hand entry into spatial zones near airway tubes, and agitation, quantified by the velocity of tracked anatomical keypoints. Expert assessments confirmed the realism of the synthetic data, and performance evaluations showed high accuracy for collision detection and moderate performance for agitation recognition.This work demonstrates a novel pathway for developing privacy-preserving, reproducible patient safety monitoring systems with potential for deployment in intensive care settings.
Personalized concept generation by tuning diffusion models with a few images raises potential legal and ethical concerns regarding privacy and intellectual property rights. Researchers attempt to prevent malicious personalization using adversarial perturbations. However, previous efforts have mainly focused on the effectiveness of protection while neglecting the visibility of perturbations. They utilize global adversarial perturbations, which introduce noticeable alterations to original images and significantly degrade visual quality. In this work, we propose the Visual-Friendly Concept Protection (VCPro) framework, which prioritizes the protection of key concepts chosen by the image owner through adversarial perturbations with lower perceptibility. To ensure these perturbations are as inconspicuous as possible, we introduce a relaxed optimization objective to identify the least perceptible yet effective adversarial perturbations, solved using the Lagrangian multiplier method. Qualitative and quantitative experiments validate that VCPro achieves a better trade-off between the visibility of perturbations and protection effectiveness, effectively prioritizing the protection of target concepts in images with less perceptible perturbations.
Predictive modeling in high-stakes domains often suffers from limited observed features due to ethical and practical constraints. To address this challenge, we propose a novel approach that formulates latent feature mining as a text-to-text propositional logic reasoning task, facilitating domain knowledge integration and improving the interpretability of latent features. We design FLAME, a domain knowledge-augmented reasoning framework for latent feature mining, offering an efficient training paradigm to strengthen the domain-specific reasoning capabilities of large language models (LLMs) for latent feature extraction. The goal of our framework is to augment observed features with inferred latent features, enhancing the performance of predictive models in downstream machine learning tasks. We validate our approach through two case studies: (1) the criminal justice system, where data collection is ethically challenging and inherently limited, and (2) the healthcare domain, where patient privacy concerns and the complexity of medical data restrict comprehensive feature collection. Experimental results demonstrate that the inferred latent features significantly enhance the performance of downstream classifiers by over 10%.
Fragile by Design: On the Limits of Adversarial Defenses in Personalized DreamBooth Generation
PDF ↗Personalized AI applications such as DreamBooth enable the generation of customized content from user images, but also raise significant privacy concerns, particularly the risk of facial identity leakage. Recent defense mechanisms like Anti-DreamBooth attempt to mitigate this risk by injecting adversarial perturbations into user photos to prevent successful personalization. However, we identify two critical yet overlooked limitations of these methods. First, the adversarial examples often exhibit perceptible artifacts such as conspicuous patterns or stripes, making them easily detectable as manipulated content. Second, the perturbations are highly fragile, as even a simple, non-learned filter can effectively remove them, thereby restoring the model's ability to memorize and reproduce user identity. To investigate this vulnerability, we propose a novel evaluation framework, AntiDB_Purify, to systematically evaluate existing defenses under realistic purification threats, including both traditional image filters and adversarial purification. Results reveal that none of the current methods maintains their protective effectiveness under such threats. These findings highlight that current defenses offer a false sense of security and underscore the urgent need for more imperceptible and robust protections to safeguard user identity in personalized generation.
Large Language Models (LLMs), pre-trained on massive text corpora, exhibit remarkable human-level language understanding, reasoning, and decision-making abilities. However, they tend to memorize unwanted information, such as private or copyrighted content, raising significant privacy and legal concerns. Unlearning has emerged as a promising solution, but existing methods face a significant challenge of over-forgetting. This issue arises because they indiscriminately suppress the generation of all the tokens in forget samples, leading to a substantial loss of model utility. To overcome this challenge, we introduce the Targeted Information Forgetting (TIF) framework, which consists of (1) a flexible targeted information identifier designed to differentiate between unwanted words (UW) and general words (GW) in the forget samples, and (2) a novel Targeted Preference Optimization approach that leverages Logit Preference Loss to unlearn unwanted information associated with UW and Preservation Loss to retain general information in GW, effectively improving the unlearning process while mitigating utility degradation. Extensive experiments on the TOFU and MUSE benchmarks demonstrate that the proposed TIF framework enhances unlearning effectiveness while preserving model utility and achieving state-of-the-art results.
Large language models have gained widespread attention recently, but their potential security vulnerabilities, especially privacy leakage, are also becoming apparent. To test and evaluate for data extraction risks in LLMs, we propose CoSPED, short for Consistent Soft Prompt Targeted Data Extraction and Defense. We introduce several innovative components, including Dynamic Loss, Additive Loss, Common Loss, and Self Consistency Decoding Strategy, and tested to enhance the consistency of the soft prompt tuning process. Through extensive experimentation with various combinations, we achieved an extraction rate of 65.2% at a 50-token prefix comparison. Our comparisons of CoSPED with other reference works confirm our superior extraction rates. We evaluate CoSPED on more scenarios, achieving Pythia model extraction rate of 51.7% and introducing cross-model comparison. Finally, we explore defense through Rank-One Model Editing and achieve a reduction in the extraction rate to 1.6%, which proves that our analysis of extraction mechanisms can directly inform effective mitigation strategies against soft prompt-based attacks.
Recent advancements in multimodal large language models for video understanding (videoLLMs) have enhanced their capacity to process complex spatiotemporal data. However, challenges such as factual inaccuracies, harmful content, biases, hallucinations, and privacy risks compromise their reliability. This study introduces Trust-videoLLMs, a first comprehensive benchmark evaluating 23 state-of-the-art videoLLMs (5 commercial, 18 open-source) across five critical dimensions: truthfulness, robustness, safety, fairness, and privacy. Comprising 30 tasks with adapted, synthetic, and annotated videos, the framework assesses spatiotemporal risks, temporal consistency and cross-modal impact. Results reveal significant limitations in dynamic scene comprehension, cross-modal perturbation resilience and real-world risk mitigation. While open-source models occasionally outperform, proprietary models generally exhibit superior credibility, though scaling does not consistently improve performance. These findings underscore the need for enhanced training datat diversity and robust multimodal alignment. Trust-videoLLMs provides a publicly available, extensible toolkit for standardized trustworthiness assessments, addressing the critical gap between accuracy-focused benchmarks and demands for robustness, safety, fairness, and privacy.
Towards Benchmarking Privacy Vulnerabilities in Selective Forgetting with Large Language Models
PDF ↗The rapid advancements in artificial intelligence (AI) have primarily focused on the process of learning from data to acquire knowledgeable learning systems. As these systems are increasingly deployed in critical areas, ensuring their privacy and alignment with human values is paramount. Recently, selective forgetting (also known as machine unlearning) has shown promise for privacy and data removal tasks, and has emerged as a transformative paradigm shift in the field of AI. It refers to the ability of a model to selectively erase the influence of previously seen data, which is especially important for compliance with modern data protection regulations and for aligning models with human values. Despite its promise, selective forgetting raises significant privacy concerns, especially when the data involved come from sensitive domains. While new unlearning-induced privacy attacks are continuously proposed, each is shown to outperform its predecessors using different experimental settings, which can lead to overly optimistic and potentially unfair assessments that may disproportionately favor one particular attack over the others. In this work, we present the first comprehensive benchmark for evaluating privacy vulnerabilities in selective forgetting. We extensively investigate privacy vulnerabilities of machine unlearning techniques and benchmark privacy leakage across a wide range of victim data, state-of-the-art unlearning privacy attacks, unlearning methods, and model architectures. We systematically evaluate and identify critical factors related to unlearning-induced privacy leakage. With our novel insights, we aim to provide a standardized tool for practitioners seeking to deploy customized unlearning applications with faithful privacy assessments.
As LLMs (large language models) are increasingly used to generate synthetic personas, particularly in data-limited domains such as health, privacy, and HCI, it becomes necessary to understand how these narratives represent identity, especially that of minority communities. In this paper, we audit synthetic personas generated by 3 LLMs (GPT4o, Gemini 1.5 Pro, Deepseek v2.5) through the lens of representational harm, focusing specifically on racial identity. Using a mixed-methods approach combining close reading, lexical analysis, and a parameterized creativity framework, we compare 1,512 LLM-generated persona to human-authored responses. Our findings reveal that LLMs disproportionately foreground racial markers, overproduce culturally coded language, and construct personas that are syntactically elaborate yet narratively reductive. These patterns result in a range of sociotechnical harms, including stereotyping, exoticism, erasure, and benevolent bias, that are often obfuscated by superficially positive narrations. We formalize this phenomenon as algorithmic othering, where minoritized identities are rendered hypervisible but less authentic.
Resilience in Ambient Multi-Agent LLMs via Decentralized Bio-Autonomic Control and Immune-Inspired Anomaly Detection
PDF ↗Large Language Model (LLM) agents are now widely deployed in Ambient Intelligence (AmI) environments, where autonomous agents must sense, act, and coordinate at scale. As agent capabilities and interdependence increase, traditional reliability strategies such as isolated adaptive control, anomaly detection, or trust modeling have proven inadequate due to their fragmented and scenario-specific nature. Comprehensive architectures that enable integrated self-management, collective anomaly response, robust information dissemination, and privacy-preserving adaptation remain scarce. We propose a bio-autonomic framework for decentralized resilience in multi-agent LLM systems where a unified architecture systematically applies principles from biological autonomic systems to LLM-based multi-agent environments. Specifically, each agent implements an autonomic control loop, formally structured as Monitor-Analyze-Plan-Execute over a shared Knowledge base (MAPE-K), for self-regulation. At the system level, the framework integrates immune-inspired anomaly detection using the Dendritic Cell Algorithm, probabilistic computational trust, decentralized gossip for robust information sharing, and federated learning with homomorphic encryption for collaborative, privacy-preserving adaptation. This holistic approach enables LLM agent ecosystems to self-organize, detect and isolate faults, and collectively adapt as system complexity increases. Empirical evaluations show that our framework achieves substantially improved resilience and recovery compared to state-of-the-art multi-agent baselines.
We consider the problem of offline reinforcement learning from human feedback (RLHF) with pairwise comparisons, where the implicit reward is a linear function of an unknown parameter. Given an offline dataset, our objective is to identify the optimal action for each state, with the ultimate goal of minimizing the simple regret. We propose an algorithm, Reinforcement Learning with Locally Optimal Weights (RL-LOW), which achieves an exponential rate of simple regret that decays exponentially with the ratio of the number of data samples to an instance-dependent hardness parameter. This hardness parameter depends explicitly on the suboptimality gap of each action. Furthermore, we derive the first instance-dependent lower bound for offline RLHF with pairwise comparisons. Interestingly, the lower and upper bounds on the simple regret match in an order-wise sense in the exponent, demonstrating the order-wise optimality of RL-LOW. Motivated by privacy considerations in practical applications, we further extend RL-LOW to the setting of differential privacy and show, somewhat surprisingly, that the hardness parameter remains unchanged in the asymptotic regime as the number of data samples tends to infinity. This result highlights the inherent efficiency of RL-LOW in preserving the privacy of the observed rewards. By establishing instance-dependent bounds with exponential convergence rates, our work fills an important gap in the existing literature, which has primarily focused on worst-case regret bounds with inverse polynomial convergence rates for offline RLHF with pairwise comparisons.
Rep Deep & Machine Learning: Exemplar-Free Continual Video Action Recognition via Slow-Fast Collaborative Learning
PDF ↗In real-world applications, video action recognition models must continuously learn new action categories while retaining previously acquired knowledge. However, most existing approaches rely on storing historical data for replay, which introduces storage burdens and raises data privacy concerns. To address these challenges, we investigate the problem of Exemplar-Free Continual Video Action Recognition (EF-CVAR) and propose a novel framework named Slow-Fast Collaborative Learning (SFCL). SFCL integrates two complementary learning paradigms: a slow branch based on gradient-driven deep learning, which provides strong adaptability to new tasks, and a fast branch based on analytic learning (e.g., Recursive Least Squares), which efficiently preserves old knowledge without requiring access to past samples. To enable effective collaboration between the two branches, we design the Slow-Fast Dynamic Re-parameterization (SFDR) mechanism for adaptive fusion, and the Knowledge Reflection Mechanism (KRM), which mitigates forgetting and task-recency bias via pseudo-feature generation and dual-level knowledge distillation. Extensive experiments on UCF101, HMDB51, and Something-Something V2 demonstrate that SFCL achieves superior performance compared to existing replay-based methods, despite being exemplar-free. Notably, in long-duration continual learning scenarios, SFCL exhibits remarkable robustness, achieving up to a 30.39\% improvement in accuracy over baselines while maintaining a low forgetting rate, highlighting its scalability and effectiveness in real-world video recognition tasks.