The aim of explainable artificial intelligence (XAI) is to address the black-box problem in high-stakes applications. However, transparency alone does not guarantee trust. This review examines a critical paradox in XAI research. While explanation methods can generate insights, three main challenges limit their effectiveness. Firstly, adversarial manipulations can exploit explanations by creating new attack surfaces with over ninety percent success while preserving model accuracy. Secondly, evaluation practices remain primarily computational. Only twenty-six percent of user studies follow human-centered protocols and fewer than twenty-three percent involve domain experts. Thirdly, regulatory requirements, such as the GDPR right to explanation, lack clear technical implementations, complicating compliance. We analyzed the literature across finance, healthcare, and cybersecurity and found that current research emphasizes algorithmic innovation over practical deployment. Moving toward reliable AI requires shifting from simple explanation methods (XAI 1.0) to systems that are aligned with human understanding, resistant to adversarial attacks, and compliant with legal requirements (XAI 2.0). This review provides guidance on key technical advances, evaluation strategies and regulatory clarifications necessary for deployment. trustworthy AI.

As 6G networks become increasingly complex and heterogeneous, effective classification of network slicing is essential for optimizing resources and managing quality of service. While recent advances demonstrate high accuracy under controlled laboratory conditions, a critical gap exists between algorithm performance evaluation under idealized conditions and their actual effectiveness in realistic deployment scenarios. This study presents a comprehensive comparative analysis of two distinct preprocessing methodologies for 6G network slicing classification: Pure Raw Data Analysis (PRDA) and Literature-Validated Realistic Transformations (LVRTs). We evaluate the impact of these strategies on algorithm performance, resilience characteristics, and practical deployment feasibility to bridge the laboratory–reality gap in 6G network optimization. Our experimental methodology involved testing eleven machine learning algorithms—including traditional ML, ensemble methods, and deep learning approaches—on a dataset comprising 10,000 network slicing samples (expanded to 21,033 through realistic transformations) across five network slice types. The LVRT methodology incorporates realistic operational impairments including market-driven class imbalance (9:1 ratio), multi-layer interference patterns, and systematic missing data reflecting authentic 6G deployment challenges. The experimental results revealed significant differences in algorithm behavior between the two preprocessing approaches. Under PRDA conditions, deep learning models achieved perfect accuracy (100% for CNN and FNN), while traditional algorithms ranged from 60.9% to 89.0%. However, LVRT results exposed dramatic performance variations, with accuracies spanning from 58.0% to 81.2%. Most significantly, we discovered that algorithms achieving excellent laboratory performance experience substantial degradation under realistic conditions, with CNNs showing an 18.8% accuracy loss (dropping from 100% to 81.2%), FNNs experiencing an 18.9% loss (declining from 100% to 81.1%), and Naive Bayes models suffering a 34.8% loss (falling from 89% to 58%). Conversely, SVM (RBF) and Logistic Regression demonstrated counter-intuitive resilience, improving by 14.1 and 10.3 percentage points, respectively, under operational stress, demonstrating superior adaptability to realistic network conditions. This study establishes a resilience-based classification framework enabling informed algorithm selection for diverse 6G deployment scenarios. Additionally, we introduce a comprehensive explainable artificial intelligence (XAI) framework using SHAP analysis to provide interpretable insights into algorithm decision-making processes. The XAI analysis reveals that Packet Loss Budget emerges as the dominant feature across all algorithms, while Slice Jitter and Slice Latency constitute secondary importance features. Cross-scenario interpretability consistency analysis demonstrates that CNN, LSTM, and Naive Bayes achieve perfect or near-perfect consistency scores (0.998–1.000), while SVM and Logistic Regression maintain high consistency (0.988–0.997), making them suitable for regulatory compliance scenarios. In contrast, XGBoost shows low consistency (0.106) despite high accuracy, requiring intensive monitoring for deployment. This research contributes essential insights for bridging the critical gap between algorithm development and deployment success in next-generation wireless networks, providing evidence-based guidelines for algorithm selection based on accuracy, resilience, and interpretability requirements. Our findings establish quantitative resilience boundaries: algorithms achieving >99% laboratory accuracy exhibit 58–81% performance under realistic conditions, with CNN and FNN maintaining the highest absolute accuracy (81.2% and 81.1%, respectively) despite experiencing significant degradation from laboratory conditions.

The vision of 6G requires autonomous, intelligent, and adaptive systems that can manage complex and dynamic communication environments. In this context, artificial intelligence (AI)-based agents are seen as key enablers, combining perception, knowledge, decision, and execution functions to operate across distributed network layers. This study provides a comprehensive overview of agentic systems for 6G, examining their architecture, core capabilities, and enabling AI techniques. We propose a novel framework for agentic intelligence levels which conceptualizes agentic AI systems as a gradual and multidimensional spectrum. Additionally, trust and governance levels, which characterize system deployability independently of agentic competencies, are introduced. We analyze how methods such as reinforcement learning (RL), federated learning, graph neural networks, transfer learning (TL), and meta-learning can be integrated into agents to support tasks including network optimization, predictive traffic management, spectrum allocation, energy-efficient resource control, and security enhancement. The potential of foundation model-based agents in 6G networks is examined, comparing them to classical RL-based agents to highlight their complementarity. Practical 6G use cases—including smart resource orchestration, environmental coordination, industrial applications, reliability and security assurance, and situational awareness—are presented to demonstrate the applicability of agents in real-world scenarios. Their strengths, limitations, and complementary roles within agent-based frameworks are highlighted. The current technological and industrial maturity levels of use cases are mapped to the proposed levels of agentic intelligence and trust/governance. The study bridges theoretical advances and practical 6G implementations by linking AI methods with concrete deployment challenges. Finally, the study identifies challenges related to scalability, interoperability, explainability, and trust, and highlights future directions for building reliable, adaptive, and ethically aligned agent-based systems in 6G.

In addressing the multifaceted problem of multiple-input multiple-output (MIMO) detection in wireless communication systems, this work highlights the pressing need for enhanced detection reliability under variable channel conditions and MIMO antenna configurations. We propose a novel method that sets a new standard for deep unfolding in MIMO detection by integrating the iterative conjugate gradient method with Tikhonov regularization, combining the adaptability of modern deep learning techniques with the robustness of classical regularization. Unlike conventional techniques, our strategy treats the Tikhonov regularization parameter, as well as the step size and search direction coefficients of the conjugate gradient (CG) method, as trainable parameters within the deep learning framework. This enables dynamic adjustments based on varying channel conditions and MIMO antenna configurations. Detection performance is significantly improved by the proposed approach across a range of MIMO configurations and channel conditions, consistently achieving lower bit error rate (BER) and normalized minimum mean square error (NMSE) compared to well-known techniques like DetNet and CG. The proposed method has superior performance over CG and other model-based methods, especially with a small number of iterations. Consequently, the simulation results demonstrate the flexibility of the proposed approach, making it a viable choice for MIMO systems with a range of antenna configurations, modulation orders, and different channel conditions.

The sixth-generation (6G) wireless communication technology is not merely an incremental evolution of mobile systems, but a foundational infrastructure envisioned to support transformative applications ranging from tactile internet to brain-computer interfaces. This vision necessitates the widespread use of artificial intelligence (AI) and machine learning (ML)-based approaches across different components of the network architecture; however, it also introduces emerging and still poorly understood security vulnerabilities. In particular, adversarial attacks targeting AI components pose serious risks to network performance, reliability, and privacy across multiple layers of the 6G architecture. The fact that beneficial intelligence, expected to enhance network efficiency and adaptability under normal conditions, can exhibit hostile behavior under malicious inputs and interactions, potentially propagating across the system, emerges as a fundamental security concern. This paper is a survey that systematically examines adversarial threats reported in AI-native 6G architectures, covering a broad spectrum from the physical layer to the application layer. To organize the scattered literature, the survey adopts a taxonomy that brings together adversarial attacks from a holistic perspective. Attacks discussed in the literature are analyzed by considering where they occur, what they target, and how they propagate. Based on these dimensions, adversarial attacks are grouped into four main categories: physical-world attacks, input manipulation attacks, model-level attacks, and cross-domain attacks. This classification aims to support a more structured understanding of attack surfaces in the AI-intensive, cross-layer interactive, and distributed nature of 6G systems. The survey further reviews defense approaches proposed in the literature, outlining the threat types they address and their known limitations. Commonly used evaluation metrics and benchmark studies are discussed in terms of their strengths and shortcomings. In addition, research topics that have received limited attention so far, such as adversarial drift and cross-domain propagation, are highlighted. The aim of this study is to draw attention to vulnerability factors that should be considered in the design and security of future 6G networks and to raise awareness in this domain.

Target tracking in integrated sensing and communication (ISAC) systems faces critical challenges due to complex interference patterns and dynamic resource allocation between radar sensing and wireless communication functions. Classical tracking algorithms struggle with the non-Gaussian noise characteristics inherent in ISAC environments. This paper addresses these limitations through a novel hybrid ISAC-LSTM architecture that enhances Extended Kalman Filter performance using intelligent machine learning corrections. The approach processes comprehensive feature vectors including baseline EKF states, ISAC-specific interference indicators, and innovation-based statistical occlusion detection. ISAC systems exhibit fundamental symmetry through dual sensing–communication operations sharing identical spectral and hardware resources, requiring balanced resource allocation, where ${\alpha }_{\mathrm{sensing}}+{\alpha }_{\mathrm{comm}}=1$. The proposed hybrid architecture preserves this functional symmetry while achieving balanced performance across symmetric dual evaluation scenarios (normal and extreme conditions). Comprehensive evaluation across three realistic deployment scenarios demonstrates substantial performance improvements, achieving 21–24% RMSE reductions over classical methods (3.5–3.6 m vs. 4.6 m) with statistical significance confirmed through paired t-tests and cross-validation. The hybrid system incorporates fail-safe mechanisms ensuring reliable operation when machine learning components encounter errors, addressing critical deployment concerns for practical ISAC applications.