arXiv Computer Science @arxiv_cs@qoto.org

This study examines whether German X users would see politically balanced news feeds if they followed comparable leading politicians from each federal parliamentary party of Germany. We address this question using an algorithmic audit tool [1] and all publicly available posts published by 436 German politicians on X. We find that the default feed of X showed more content from far-right AfD than from other political parties. We analyze potential factors influencing feed content and the resulting political non-representativeness of X. Our findings suggest that engagement measures and unknown factors related to party affiliation contribute to the overrepresentation of extremes of the German political party spectrum in the default algorithmic feed of X.

Cascading failures (CF) entail component breakdowns spreading through infrastructure networks, causing system-wide collapse. Predicting CFs is of great importance for infrastructure stability and urban function. Despite extensive research on CFs in single networks such as electricity and road networks, interdependencies among diverse infrastructures remain overlooked, and capturing intra-infrastructure CF dynamics amid complex evolutions poses challenges. To address these gaps, we introduce the \textbf{I}ntegrated \textbf{I}nterdependent \textbf{I}nfrastructure CF model ($I^3$), designed to capture CF dynamics both within and across infrastructures. $I^3$ employs a dual GAE with global pooling for intra-infrastructure dynamics and a heterogeneous graph for inter-infrastructure interactions. An initial node enhancement pre-training strategy mitigates GCN-induced over-smoothing. Experiments demonstrate $I^3$ achieves a 31.94\% in terms of AUC, 18.03\% in terms of Precision, 29.17\% in terms of Recall, 22.73\% in terms of F1-score boost in predicting infrastructure failures, and a 28.52\% reduction in terms of RMSE for cascade volume forecasts compared to leading models. It accurately pinpoints phase transitions in interconnected and singular networks, rectifying biases in models tailored for singular networks. Access the code at https://github.com/tsinghua-fib-lab/Icube.

In recent years, vision transformers (ViTs) have emerged as powerful and promising techniques for computer vision tasks such as image classification, object detection, and segmentation. Unlike convolutional neural networks (CNNs), which rely on hierarchical feature extraction, ViTs treat images as sequences of patches and leverage self-attention mechanisms. However, their high computational complexity and memory demands pose significant challenges for deployment on resource-constrained edge devices. To address these limitations, extensive research has focused on model compression techniques and hardware-aware acceleration strategies. Nonetheless, a comprehensive review that systematically categorizes these techniques and their trade-offs in accuracy, efficiency, and hardware adaptability for edge deployment remains lacking. This survey bridges this gap by providing a structured analysis of model compression techniques, software tools for inference on edge, and hardware acceleration strategies for ViTs. We discuss their impact on accuracy, efficiency, and hardware adaptability, highlighting key challenges and emerging research directions to advance ViT deployment on edge platforms, including graphics processing units (GPUs), tensor processing units (TPUs), and field-programmable gate arrays (FPGAs). The goal is to inspire further research with a contemporary guide on optimizing ViTs for efficient deployment on edge devices.

High-quality annotations are essential for object detection models, but ensuring label accuracy - especially for bounding boxes - remains both challenging and costly. This paper introduces ClipGrader, a novel approach that leverages vision-language models to automatically assess the accuracy of bounding box annotations. By adapting CLIP (Contrastive Language-Image Pre-training) to evaluate both class label correctness and spatial precision of bounding box, ClipGrader offers an effective solution for grading object detection labels. Tested on modified object detection datasets with artificially disturbed bounding boxes, ClipGrader achieves 91% accuracy on COCO with a 1.8% false positive rate. Moreover, it maintains 87% accuracy with a 2.1% false positive rate when trained on just 10% of the COCO data. ClipGrader also scales effectively to larger datasets such as LVIS, achieving 79% accuracy across 1,203 classes. Our experiments demonstrate ClipGrader's ability to identify errors in existing COCO annotations, highlighting its potential for dataset refinement. When integrated into a semi-supervised object detection (SSOD) model, ClipGrader readily improves the pseudo label quality, helping achieve higher mAP (mean Average Precision) throughout the training process. ClipGrader thus provides a scalable AI-assisted tool for enhancing annotation quality control and verifying annotations in large-scale object detection datasets.

Networks can be highly complex systems with numerous interconnected components and interactions. Granular computing offers a framework to manage this complexity by decomposing networks into smaller, more manageable components, or granules. In this article, we introduce metric-based granular computing technique to study networks. This technique can be applied to the analysis of networks where granules can represent subsets of nodes or edges and their interactions can be studied at different levels of granularity. We model the network as an information system and investigate its granular structures using metric representation. We establish that the concepts of reducts in rough set theory and resolving sets in networks are equivalent. Through this equivalence, we present a novel approach for computing all the minimal resolving sets of these networks.

Prosody plays a crucial role in speech perception, influencing both human understanding and automatic speech recognition (ASR) systems. Despite its importance, prosodic stress remains under-studied due to the challenge of efficiently analyzing it. This study explores fine-tuning OpenAI's Whisper large-v2 ASR model to recognize phrasal, lexical, and contrastive stress in speech. Using a dataset of 66 native English speakers, including male, female, neurotypical, and neurodivergent individuals, we assess the model's ability to generalize stress patterns and classify speakers by neurotype and gender based on brief speech samples. Our results highlight near-human accuracy in ASR performance across all three stress types and near-perfect precision in classifying gender and neurotype. By improving prosody-aware ASR, this work contributes to equitable and robust transcription technologies for diverse populations.

Gas leakage poses a significant hazard that requires prevention. Traditionally, human inspection has been used for detection, a slow and labour-intensive process. Recent research has applied machine learning techniques to this problem, yet there remains a shortage of high-quality, publicly available datasets. This paper introduces a synthetic dataset featuring diverse backgrounds, interfering foreground objects, diverse leak locations, and precise segmentation ground truth. We propose a zero-shot method that combines background subtraction, zero-shot object detection, filtering, and segmentation to leverage this dataset. Experimental results indicate that our approach significantly outperforms baseline methods based solely on background subtraction and zero-shot object detection with segmentation, reaching an IoU of 69\% overall. We also present an analysis of various prompt configurations and threshold settings to provide deeper insights into the performance of our method. The code and dataset will be released after publication.

Throughout this project, we have attempted to develop an automatic evaluator that determines whether Basque language compositions meet the C1 level. To achieve our goal, we obtained 10,000 transcribed compositions through an agreement between HABE and HiTZ to train our system. We have developed different techniques to avoid data scarcity and system overfitting: EDA, SCL and regulation; We have also conducted tests with different Language Models to analyze their behavior. Finally, we have also performed analyses of different system behaviors to measure model calibration and the impact of artifacts. -- Proiektu honetan zehar euskarazko idazlanek C1 maila duten edo ez zehazten duen ebaluatzaile automatiko bat garatzen saiatu gara. Gure helburua betetzeko HABE eta HiTZ arteko hitzarmenaren bitartez 10.000 transkribatutako idazlan eskuratu ditugu gure sistema entrenatzeko. Datu eskasia eta sistemaren gaindoitzea ekiditeko teknika ezberdinak landu ditugu: EDA, SCL eta erregulazioa; Hizkuntza Eredu ezberdinekin ere probak egin ditugu duten portaera aztertzeko. Azkenik, sistema ezberdinen portaeren analisiak ere egin ditugu, ereduen kalibrazioa eta artefaktuen eragina neurtzeko.

This paper presents the experimental validation of an interaction-aware model predictive decision-making (IAMPDM) approach in the course of a simulator study. The proposed IAMPDM uses a model of the pedestrian, which simultaneously predicts their future trajectories and characterizes the interaction between the pedestrian and the automated vehicle. The main benefit of the proposed concept and the experiment is that the interaction between the pedestrian and the socially compliant autonomous vehicle leads to smoother traffic. Furthermore, the experiment features a novel human-in-the-decision-loop aspect, meaning that the test subjects have no expected behavior or defined sequence of their actions, better imitating real traffic scenarios. Results show that intention-aware decision-making algorithms are more effective in realistic conditions and contribute to smoother traffic flow than state-of-the-art solutions. Furthermore, the findings emphasize the crucial impact of intention-aware decision-making on autonomous vehicle performance in urban areas and the need for further research.

This study investigates the machine unlearning techniques within the context of large language models (LLMs), referred to as \textit{LLM unlearning}. LLM unlearning offers a principled approach to removing the influence of undesirable data (e.g., sensitive or illegal information) from LLMs, while preserving their overall utility without requiring full retraining. Despite growing research interest, there is no comprehensive survey that systematically organizes existing work and distills key insights; here, we aim to bridge this gap. We begin by introducing the definition and the paradigms of LLM unlearning, followed by a comprehensive taxonomy of existing unlearning studies. Next, we categorize current unlearning approaches, summarizing their strengths and limitations. Additionally, we review evaluation metrics and benchmarks, providing a structured overview of current assessment methodologies. Finally, we outline promising directions for future research, highlighting key challenges and opportunities in the field.

This paper proposes a strictly predefined-time convergent and anti-noise fractional-order zeroing neural network (SPTC-AN-FOZNN) model, meticulously designed for addressing time-variant quadratic programming (TVQP) problems. This model marks the first variable-gain ZNN to collectively manifest strictly predefined-time convergence and noise resilience, specifically tailored for kinematic motion control of robots. The SPTC-AN-FOZNN advances traditional ZNNs by incorporating a conformable fractional derivative in accordance with the Leibniz rule, a compliance not commonly achieved by other fractional derivative definitions. It also features a novel activation function designed to ensure favorable convergence independent of the model's order. When compared to five recently published recurrent neural networks (RNNs), the SPTC-AN-FOZNN, configured with $0<α\leq 1$, exhibits superior positional accuracy and robustness against additive noises for TVQP applications. Extensive empirical evaluations, including simulations with two types of robotic manipulators and experiments with a Flexiv Rizon robot, have validated the SPTC-AN-FOZNN's effectiveness in precise tracking and computational efficiency, establishing its utility for robust kinematic control.

It has been 100 years since statistical process control (SPC) or statistical process monitoring (SPM) was first introduced for production processes and later applied to service, healthcare, and other industries. The techniques applied to SPM applications are mostly statistically oriented. Recent advances in Artificial Intelligence (AI) have reinvigorated the imagination of adopting AI for SPM applications. This manuscript begins with a concise review of the historical development of the statistically based SPM methods. Next, this manuscript explores AI and Machine Learning (ML) algorithms and methods applied in various SPM applications, addressing quality characteristics of univariate, multivariate, profile, and image. These AI methods can be classified into the following categories: classification, pattern recognition, time series applications, and generative AI. Specifically, different kinds of neural networks, such as artificial neural networks (ANN), convolutional neural networks (CNN), recurrent neural networks (RNN), and generative adversarial networks (GAN), are among the most implemented AI methods impacting SPM. Finally, this manuscript outlines a couple of future directions that harness the potential of the Large Multimodal Model (LMM) for advancing SPM research and applications in complex systems. The ultimate objective is to transform statistical process monitoring (SPM) into smart process control (SMPC), where corrective actions are autonomously implemented to either prevent quality issues or restore process performance.

Advances in Large Language Models revolutionized medical education by enabling scalable and efficient learning solutions. This paper presents a pipeline employing Retrieval-Augmented Generation (RAG) system to prepare comments generation for Poland's State Specialization Examination (PES) based on verified resources. The system integrates these generated comments and source documents with a spaced repetition learning algorithm to enhance knowledge retention while minimizing cognitive overload. By employing a refined retrieval system, query rephraser, and an advanced reranker, our modified RAG solution promotes accuracy more than efficiency. Rigorous evaluation by medical annotators demonstrates improvements in key metrics such as document relevance, credibility, and logical coherence of generated content, proven by a series of experiments presented in the paper. This study highlights the potential of RAG systems to provide scalable, high-quality, and individualized educational resources, addressing non-English speaking users.

This paper presents our ongoing work toward developing an enterprise-ready Computer Using Generalist Agent (CUGA) system. Our research highlights the evolutionary nature of building agentic systems suitable for enterprise environments. By integrating state-of-the-art agentic AI techniques with a systematic approach to iterative evaluation, analysis, and refinement, we have achieved rapid and cost-effective performance gains, notably reaching a new state-of-the-art performance on the WebArena benchmark. We detail our development roadmap, the methodology and tools that facilitated rapid learning from failures and continuous system refinement, and discuss key lessons learned and future challenges for enterprise adoption.

With the advent of Vision-Language Models (VLMs), medical artificial intelligence (AI) has experienced significant technological progress and paradigm shifts. This survey provides an extensive review of recent advancements in Medical Vision-Language Models (Med-VLMs), which integrate visual and textual data to enhance healthcare outcomes. We discuss the foundational technology behind Med-VLMs, illustrating how general models are adapted for complex medical tasks, and examine their applications in healthcare. The transformative impact of Med-VLMs on clinical practice, education, and patient care is highlighted, alongside challenges such as data scarcity, narrow task generalization, interpretability issues, and ethical concerns like fairness, accountability, and privacy. These limitations are exacerbated by uneven dataset distribution, computational demands, and regulatory hurdles. Rigorous evaluation methods and robust regulatory frameworks are essential for safe integration into healthcare workflows. Future directions include leveraging large-scale, diverse datasets, improving cross-modal generalization, and enhancing interpretability. Innovations like federated learning, lightweight architectures, and Electronic Health Record (EHR) integration are explored as pathways to democratize access and improve clinical relevance. This review aims to provide a comprehensive understanding of Med-VLMs' strengths and limitations, fostering their ethical and balanced adoption in healthcare.

Preference learning is critical for aligning large language models (LLMs) with human values, with the quality of preference datasets playing a crucial role in this process. While existing metrics primarily assess data quality based on either explicit or implicit reward margins, they often provide contradictory evaluations for the same data. To address this issue, we introduce the alignment potential metric, which quantifies the gap from the model's current implicit reward margin to the target explicit reward margin, thereby estimating the model's potential to align with the preference data. Empirical results demonstrate that training on data selected by this metric consistently enhances alignment performance, surpassing existing metrics across different base models and optimization objectives. Furthermore, our method extends to self-play data generation frameworks, where the metric is used to identify high-quality data within the self-generated content by LLMs. Under this data generation scenario, our method surpasses current state-of-the-art (SOTA) results across various training settings and demonstrates continuous improvements in alignment performance as dataset size and training iterations increase.

Large Language Models (LLMs) have been previously explored for mental healthcare training and therapy client simulation, but they still fall short in authentically capturing diverse client traits and psychological conditions. We introduce \textbf{Eeyore}, an 8B model optimized for realistic depression simulation through a structured alignment framework, incorporating expert input at every stage. First, we systematically curate real-world depression-related conversations, extracting depressive traits to guide data filtering and psychological profile construction, and use this dataset to instruction-tune Eeyore for profile adherence. Next, to further enhance realism, Eeyore undergoes iterative preference optimization -- first leveraging model-generated preferences and then calibrating with a small set of expert-annotated preferences. Throughout the entire pipeline, we actively collaborate with domain experts, developing interactive interfaces to validate trait extraction and iteratively refine structured psychological profiles for clinically meaningful role-play customization. Despite its smaller model size, the Eeyore depression simulation outperforms GPT-4o with SOTA prompting strategies, both in linguistic authenticity and profile adherence.

Novel research aimed at text-to-image (T2I) generative AI safety often relies on publicly available datasets for training and evaluation, making the quality and composition of these datasets crucial. This paper presents a comprehensive review of the key datasets used in the T2I research, detailing their collection methods, compositions, semantic and syntactic diversity of prompts and the quality, coverage, and distribution of harm types in the datasets. By highlighting the strengths and limitations of the datasets, this study enables researchers to find the most relevant datasets for a use case, critically assess the downstream impacts of their work given the dataset distribution, particularly regarding model safety and ethical considerations, and also identify the gaps in dataset coverage and quality that future research may address.

Key-value (KV) caching has emerged as a crucial optimization technique for accelerating inference in large language models (LLMs). By allowing the attention operation to scale linearly rather than quadratically with the total sequence length, KV caching significantly enhances generation throughput. However, due to large context lengths in the modern LLMs, the memory footprint of the KV is a huge bottleneck for model deployment directly impacting the model's batch size, hindering its ability to deliver high-throughput. Existing research addresses this challenge using several techniques, such as discarding low-attention tokens, quantization, and matrix approximation which typically lead to a negative impact on the model accuracy. In this paper, We propose KVCrush technology which can be combined with many KV compression technologies to improve the model accuracy at a much smaller memory. KVCrush provides an alternate representation scheme for key-value states, along with a low-overhead token pruning algorithm that accounts for the token distribution in the KV cache, which in turn allows for a a smaller footprint while maintaining the accuracy of the model. Based on our results, KVCrush reduces LongBench KV Cache size by 4x with less than 1% accuracy drop and achieves state-of-the-art average accuracy with minimal overhead, incurring less than 0.5% total inference latency. KVCrush not only outperforms the accuracy of state-of-the-art importance-based token retention schemes but is also compatible with typical practical LLM deployments using KV cache paging schemes such as vLLM and mixed precision quantization.