arXiv subjects

Nikolaos Pappas

Publications and source records attributed to Nikolaos Pappas.

At least 55 records · Page 3Linked to original sources

arXiv2026

Semantics in Actuation Systems: From Age of Actuation to Age of Actuated Information

In this paper, we study the timeliness of actions in communication systems where actuation is constrained by control permissions or energy availability. Building on the Age of Actuation (AoA) metric, which quantifies the timeliness of actions independently of data freshness, we introduce a new metric, the \emph{Age of Actuated Information (AoAI)}. AoAI captures the end-to-end timeliness of actions by explicitly accounting for the age of the data packet at the moment it is actuated. We analyze and characterize both AoA and AoAI in discrete-time systems with data storage capabilities under multiple actuation scenarios. The actuator requires both a data packet and an actuation opportunity, which may be provided by a controller or enabled by harvested energy. Data packets may be stored either in a single-packet buffer or an infinite-capacity queue for future actuation. For these settings, we derive closed-form expressions for the average AoA and AoAI and investigate their structural differences. While AoA and AoAI coincide in instantaneous actuation systems, they differentiate when data buffering is present. Our results reveal counterintuitive regimes in which increasing update or actuation rates degrade action timeliness for both AoA and AoAI. Moreover, as part of the analysis, we obtain a novel closed-form characterization of the steady-state distribution of a Geo/Geo/1 queue operating under the FCFS discipline, expressed solely in terms of the queue length and the age of the head-of-line packet. The proposed metrics and analytical results provide new insights into the semantics of timeliness in systems where information ultimately serves the purpose of actuation.

cs.IT↗

arXiv2026

Balancing Classification and Calibration Performance in Decision-Making LLMs via Calibration Aware Reinforcement Learning

Large language models (LLMs) are increasingly deployed in decision-making tasks, where not only accuracy but also reliable confidence estimates are essential. Well-calibrated confidence enables downstream systems to decide when to trust a model and when to defer to fallback mechanisms. In this work, we conduct a systematic study of calibration in two widely used fine-tuning paradigms: supervised fine-tuning (SFT) and reinforcement learning with verifiable rewards (RLVR). We show that while RLVR improves task performance, it produces extremely overconfident models, whereas SFT yields substantially better calibration, even under distribution shift, though with smaller performance gains. Through targeted experiments, we diagnose RLVR's failure, showing that decision tokens act as extraction steps of the decision in reasoning traces and do not carry confidence information, which prevents reinforcement learning from surfacing calibrated alternatives. Based on this insight, we propose a calibration-aware reinforcement learning formulation that directly adjusts decision-token probabilities. Our method preserves RLVR's accuracy level while mitigating overconfidence, reducing ECE scores up to 9 points.

cs.LG↗

arXiv2026

From Timestamps to Versions: Version AoI in Single- and Multi-Hop Networks

Timely and informative data dissemination in communication networks is essential for enhancing system performance and energy efficiency, as it reduces the transmission of outdated or redundant data. Timeliness metrics, such as Age of Information (AoI), effectively quantify data freshness; however, these metrics fail to account for the intrinsic informativeness of the content itself. To address this limitation, content-based metrics have been proposed that combine both timeliness and informativeness. Nevertheless, existing studies have predominantly focused on evaluating average metric values, leaving the complete distribution-particularly in multi-hop network scenarios-largely unexplored. In this paper, we provide a comprehensive analysis of the stationary distribution of the Version Age of Information (VAoI), a content-based metric, under various scheduling policies, including randomized stationary, uniform, and threshold-based policies, with transmission constraints in single-hop and multi-hop networks. We derive closed-form expressions for the stationary distribution and average VAoI under these scheduling approaches. Furthermore, for threshold-based scheduling, we analytically determine the optimal threshold value that minimizes VAoI and derive the corresponding optimal VAoI in closed form. Numerical evaluations verify our analytical findings, providing valuable insights into leveraging VAoI in the design of efficient communication networks.

cs.NI↗

arXiv2025

Real-Time Remote Monitoring of Correlated Markovian Sources

We investigate real-time tracking of two correlated stochastic processes over a shared wireless channel. The joint evolution of the processes is modeled as a two-dimensional discrete-time Markov chain. Each process is observed by a dedicated sampler and independently reconstructed at a remote monitor according to a task-specific objective. Although both processes originate from a common underlying phenomenon (e.g., distinct features of the same source), each monitor is interested only in its corresponding feature. A reconstruction error is incurred when the true and reconstructed states mismatch at one or both monitors. To address this problem, we propose an error-aware joint sampling and transmission policy, under which each sampler probabilistically generates samples only when the current process state differs from the most recently reconstructed state at its corresponding monitor. We adopt the time-averaged reconstruction error as the primary performance metric and benchmark the proposed policy against state-of-the-art joint sampling and transmission schemes. For each policy, we derive closed-form expressions for the resulting time-averaged reconstruction error. We further formulate and solve an optimization problem that minimizes the time-averaged reconstruction error subject to an average sampling cost constraint. Analytical and numerical results demonstrate that the proposed error-aware policy achieves the minimum time-averaged reconstruction error among the considered schemes while efficiently utilizing the sampling budget. The performance gains are particularly pronounced in regimes with strong inter-process correlation and stringent tracking requirements, where frequent sampling by both samplers is necessary.

cs.IT↗

arXiv2025

Journey Before Destination: On the importance of Visual Faithfulness in Slow Thinking

Reasoning-augmented vision language models (VLMs) generate explicit chains of thought that promise greater capability and transparency but also introduce new failure modes: models may reach correct answers via visually unfaithful intermediate steps, or reason faithfully yet fail on the final prediction. Standard evaluations that only measure final-answer accuracy cannot distinguish these behaviors. We introduce the visual faithfulness of reasoning chains as a distinct evaluation dimension, focusing on whether the perception steps of a reasoning chain are grounded in the image. We propose a training- and reference-free framework that decomposes chains into perception versus reasoning steps and uses off-the-shelf VLM judges for step-level faithfulness, additionally verifying this approach through a human meta-evaluation. Building on this metric, we present a lightweight self-reflection procedure that detects and locally regenerates unfaithful perception steps without any training. Across multiple reasoning-trained VLMs and perception-heavy benchmarks, our method reduces Unfaithful Perception Rate while preserving final-answer accuracy, improving the reliability of multimodal reasoning.

cs.CV↗

arXiv2025

From Information Freshness to Semantics of Information and Goal-oriented Communications

Future wireless networks must support real-time, data-driven cyber-physical systems in which communication is tightly coupled with sensing, inference, control, and decision-making. Traditional communication paradigms centered on accuracy, throughput, and latency are increasingly inadequate for these systems, where the value of information depends on its semantic relevance to a specific task. This paper provides a unified exposition of the progression from classical distortion-based frameworks, through information freshness metrics such as the Age of Information (AoI) and its variants, to the emerging paradigm of goal-oriented semantics-aware communication. We organize and systematize existing semantics-aware metrics, including content- and version-aware measures, context-dependent distortion formulations, and history-dependent error persistence metrics that capture lasting impact and urgency. Within this framework, we highlight how these metrics address the limitations of purely accuracy- or freshness-centric designs, and how they collectively enable the selective generation and transmission of only task-relevant information. We further review analytical tools based on Markov decision process (MDP) and Lyapunov optimization methods that have been employed to characterize optimal or near-optimal timing and scheduling policies under semantic performance criteria and communication constraints. By synthesizing these developments into a coherent framework, the paper clarifies the design principles underlying goal-oriented, semantics-aware communication systems. It illustrates how they can significantly improve efficiency, reliability, and task performance. The presented perspective aims to serve as a bridge between information-theoretic, control-theoretic, and networking viewpoints, and to guide the design of semantic communication architectures for 6G and beyond.

cs.IT↗

arXiv2025

Feature-Based Semantics-Aware Scheduling for Energy-Harvesting Federated Learning

Federated Learning (FL) on resource-constrained edge devices faces a critical challenge: The computational energy required for training Deep Neural Networks (DNNs) often dominates communication costs. However, most existing Energy-Harvesting FL (EHFL) strategies fail to account for this reality, resulting in wasted energy due to redundant local computations. For efficient and proactive resource management, algorithms that predict local update contributions must be devised. We propose a lightweight client scheduling framework using the Version Age of Information (VAoI), a semantics-aware metric that quantifies update timeliness and significance. Crucially, we overcome VAoI's typical prohibitive computational cost, which requires statistical distance over the entire parameter space, by introducing a feature-based proxy. This proxy estimates model redundancy using intermediate-layer extraction from a single forward pass, dramatically reducing computational complexity. Experiments conducted under extreme non-IID data distributions and scarce energy availability demonstrate superior learning performance while achieving energy reduction compared to existing baseline selection policies. Our framework establishes semantics-aware scheduling as a practical and vital solution for EHFL in realistic scenarios where training costs dominate transmission costs.

cs.LG↗

arXiv2025

Rethinking LLM Uncertainty: A Multi-Agent Approach to Estimating Black-Box Model Uncertainty

Quantifying uncertainty in black-box LLMs is vital for reliable responses and scalable oversight. Existing methods, which gauge a model's uncertainty through evaluating self-consistency in responses to the target query, can be misleading: an LLM may confidently provide an incorrect answer to a target query, yet give a confident and accurate answer to that same target query when answering a knowledge-preserving perturbation of the query. We systematically analyze the model behaviors and demonstrate that this discrepancy stems from suboptimal retrieval of parametric knowledge, often due to contextual biases that prevent consistent access to stored knowledge. We then introduce DiverseAgentEntropy, a novel, theoretically-grounded method employing multi-agent interaction across diverse query variations for uncertainty estimation of black-box LLMs. This approach more accurately assesses an LLM's true uncertainty and improves hallucination detection, outperforming existing self-consistency based techniques.

cs.CL↗

arXiv2025

Age of Information-Aware Cognitive Shared Access Networks with Energy Harvesting

This study investigates a cognitive shared access network with energy harvesting capabilities operating under Age of Information (AoI) constraints for the primary user. Secondary transmitters are spatially distributed according to a homogeneous Poisson Point Process (PPP), while the primary user is located at a fixed position. The primary transmitter handles bursty packet arrivals, whereas secondary users operate under saturated traffic conditions. To manage interference and energy, two distinct zones are introduced: an energy harvesting zone around the primary transmitter and a guard zone around the primary receiver, within which secondary transmissions are prohibited. Secondary users access the channel probabilistically, with access decisions depending on their current battery state (charged or empty) and their location relative to the guard zone. Our objective is to analyze the primary user's AoI performance under three distinct packet management policies.

cs.NI↗

arXiv2025

DeAL: Decoding-time Alignment for Large Language Models

Large Language Models (LLMs) are nowadays expected to generate content aligned with human preferences. Current work focuses on alignment at model training time, through techniques such as Reinforcement Learning with Human Feedback (RLHF). However, it is unclear if such methods are an effective choice to teach alignment objectives to the model. First, the inability to incorporate multiple, custom rewards and reliance on a model developer's view of universal and static principles are key limitations. Second, the reliability of such approaches is also questionable (e.g. susceptibility to jailbreaking even after safety training). To address these issues, we propose DeAL, a framework that allows the user to customize reward functions and enables Decoding-time Alignment of LLMs (DeAL). At its core, we view decoding as a heuristic-guided search process and facilitate the use of a wide variety of alignment objectives. Our experiments with programmatic constraints such as keyword and length constraints, and abstract objectives such as harmlessness and helpfulness, show that we can DeAL with fine-grained trade-offs and improve adherence to alignment objectives. Lastly, we demonstrate that DeAL is largely complementary to existing alignment strategies, and can be effectively paired with RLHF and prompting techniques to achieve better alignment.

cs.AI↗

arXiv2025

Cross-Modal Content Optimization for Steering Web Agent Preferences

Vision-language model (VLM)-based web agents increasingly power high-stakes selection tasks like content recommendation or product ranking by combining multimodal perception with preference reasoning. Recent studies reveal that these agents are vulnerable against attackers who can bias selection outcomes through preference manipulations using adversarial pop-ups, image perturbations, or content tweaks. Existing work, however, either assumes strong white-box access, with limited single-modal perturbations, or uses impractical settings. In this paper, we demonstrate, for the first time, that joint exploitation of visual and textual channels yields significantly more powerful preference manipulations under realistic attacker capabilities. We introduce Cross-Modal Preference Steering (CPS) that jointly optimizes imperceptible modifications to an item's visual and natural language descriptions, exploiting CLIP-transferable image perturbations and RLHF-induced linguistic biases to steer agent decisions. In contrast to prior studies that assume gradient access, or control over webpages, or agent memory, we adopt a realistic black-box threat setup: a non-privileged adversary can edit only their own listing's images and textual metadata, with no insight into the agent's model internals. We evaluate CPS on agents powered by state-of-the-art proprietary and open source VLMs including GPT-4.1, Qwen-2.5VL and Pixtral-Large on both movie selection and e-commerce tasks. Our results show that CPS is significantly more effective than leading baseline methods. For instance, our results show that CPS consistently outperforms baselines across all models while maintaining 70% lower detection rates, demonstrating both effectiveness and stealth. These findings highlight an urgent need for robust defenses as agentic systems play an increasingly consequential role in society.

cs.AI↗

arXiv2025

Optimizing Version AoI in Energy-Harvesting IoT: Model-Based and Learning-Based Approaches

Efficient data transmission in resource-constrained Internet of Things (IoT) systems requires semantics-aware management that maximizes the delivery of timely and informative data. This paper investigates the optimization of the semantic metric Version Age of Information (VAoI) in a status update system comprising an energy-harvesting (EH) sensor and a destination monitoring node. We consider three levels of knowledge about the system model -- fully known, partially known, and unknown -- and propose corresponding optimization strategies: model-based, estimation-based, and model-free methods. By employing Markov Decision Process (MDP) and Reinforcement Learning (RL) frameworks, we analyze performance trade-offs under varying degrees of model information. Our findings provide guidance for designing efficient and adaptive semantics-aware policies in both known and unknown IoT environments.

cs.NI↗

arXiv2025

RANGAN: GAN-empowered Anomaly Detection in 5G Cloud RAN

Radio Access Network (RAN) systems are inherently complex, requiring continuous monitoring to prevent performance degradation and ensure optimal user experience. The RAN leverages numerous key performance indicators (KPIs) to evaluate system performance, generating vast amounts of data each second. This immense data volume can make troubleshooting and accurate diagnosis of performance anomalies more difficult. Furthermore, the highly dynamic nature of RAN performance demands adaptive methodologies capable of capturing temporal dependencies to detect anomalies reliably. In response to these challenges, we introduce \textbf{RANGAN}, an anomaly detection framework that integrates a Generative Adversarial Network (GAN) with a transformer architecture. To enhance the capability of capturing temporal dependencies within the data, RANGAN employs a sliding window approach during data preprocessing. We rigorously evaluated RANGAN using the publicly available RAN performance dataset from the Spotlight project \cite{sun-2024}. Experimental results demonstrate that RANGAN achieves promising detection accuracy, notably attaining an F1-score of up to $83\%$ in identifying network contention issues.

cs.NI↗

arXiv2025

Rethinking Federated Learning Over the Air: The Blessing of Scaling Up

Federated learning facilitates collaborative model training across multiple clients while preserving data privacy. However, its performance is often constrained by limited communication resources, particularly in systems supporting a large number of clients. To address this challenge, integrating over-the-air computations into the training process has emerged as a promising solution to alleviate communication bottlenecks. The system significantly increases the number of clients it can support in each communication round by transmitting intermediate parameters via analog signals rather than digital ones. This improvement, however, comes at the cost of channel-induced distortions, such as fading and noise, which affect the aggregated global parameters. To elucidate these effects, this paper develops a theoretical framework to analyze the performance of over-the-air federated learning in large-scale client scenarios. Our analysis reveals three key advantages of scaling up the number of participating clients: (1) Enhanced Privacy: The mutual information between a client's local gradient and the server's aggregated gradient diminishes, effectively reducing privacy leakage. (2) Mitigation of Channel Fading: The channel hardening effect eliminates the impact of small-scale fading in the noisy global gradient. (3) Improved Convergence: Reduced thermal noise and gradient estimation errors benefit the convergence rate. These findings solidify over-the-air model training as a viable approach for federated learning in networks with a large number of clients. The theoretical insights are further substantiated through extensive experimental evaluations.

cs.LG↗

arXiv2025

Age of Information Minimization in Goal-Oriented Communication with Processing and Cost of Actuation Error Constraints

We study a goal-oriented communication system in which a source monitors an environment that evolves as a discrete-time, two-state Markov chain. At each time slot, a controller decides whether to sample the environment and if so whether to transmit a raw or processed sample, to the controller. Processing improves transmission reliability over an unreliable wireless channel, but incurs an additional cost. The objective is to minimize the long-term average age of information (AoI), subject to constraints on the costs incurred at the source and the cost of actuation error (CAE), a semantic metric that assigns different penalties to different actuation errors. Although reducing AoI can potentially help reduce CAE, optimizing AoI alone is insufficient, as it overlooks the evolution of the underlying process. For instance, faster source dynamics lead to higher CAE for the same average AoI, and different AoI trajectories can result in markedly different CAE under identical average AoI. To address this, we propose a stationary randomized policy that achieves an average AoI within a bounded multiplicative factor of the optimal among all feasible policies. Extensive numerical experiments are conducted to characterize system behavior under a range of parameters. These results offer insights into the feasibility of the optimization problem, the structure of near-optimal actions, and the fundamental trade-offs between AoI, CAE, and the costs involved.

cs.IT↗

arXiv2025

Pull-Based Query Scheduling for Goal-Oriented Semantic Communication

This paper addresses query scheduling for goal-oriented semantic communication in pull-based status update systems. We consider a system where multiple sensing agents (SAs) observe a source characterized by various attributes and provide updates to multiple actuation agents (AAs), which act upon the received information to fulfill their heterogeneous goals at the endpoint. A hub serves as an intermediary, querying the SAs for updates on observed attributes and maintaining a knowledge base, which is then broadcast to the AAs. The AAs leverage the knowledge to perform their actions effectively. To quantify the semantic value of updates, we introduce a grade of effectiveness (GoE) metric. Furthermore, we integrate cumulative perspective theory (CPT) into the long-term effectiveness analysis to account for risk awareness and loss aversion in the system. Leveraging this framework, we compute effect-aware scheduling policies aimed at maximizing the expected discounted sum of CPT-based total GoE provided by the transmitted updates while complying with a given query cost constraint. To achieve this, we propose a model-based solution based on dynamic programming and model-free solutions employing state-of-the-art deep reinforcement learning (DRL) algorithms. Our findings demonstrate that effect-aware scheduling significantly enhances the effectiveness of communicated updates compared to benchmark scheduling methods, particularly in settings with stringent cost constraints where optimal query scheduling is vital for system performance and overall effectiveness.

cs.IT↗

arXiv2025

Diffusion-Driven Semantic Communication for Generative Models with Bandwidth Constraints

Diffusion models have been extensively utilized in AI-generated content (AIGC) in recent years, thanks to the superior generation capabilities. Combining with semantic communications, diffusion models are used for tasks such as denoising, data reconstruction, and content generation. However, existing diffusion-based generative models do not consider the stringent bandwidth limitation, which limits its application in wireless communication. This paper introduces a diffusion-driven semantic communication framework with advanced VAE-based compression for bandwidth-constrained generative model. Our designed architecture utilizes the diffusion model, where the signal transmission process through the wireless channel acts as the forward process in diffusion. To reduce bandwidth requirements, we incorporate a downsampling module and a paired upsampling module based on a variational auto-encoder with reparameterization at the receiver to ensure that the recovered features conform to the Gaussian distribution. Furthermore, we derive the loss function for our proposed system and evaluate its performance through comprehensive experiments. Our experimental results demonstrate significant improvements in pixel-level metrics such as peak signal to noise ratio (PSNR) and semantic metrics like learned perceptual image patch similarity (LPIPS). These enhancements are more profound regarding the compression rates and SNR compared to deep joint source-channel coding (DJSCC). We release the code at https://github.com/import-sudo/Diffusion-Driven-Semantic-Communication.

cs.LG↗

arXiv2025

Age-Aware CSI Acquisition of a Finite-State Markovian Channel

The Age of Information (AoI) has emerged as a critical metric for quantifying information freshness; however, its interplay with channel estimation in partially observable wireless systems remains underexplored. This work considers a transmitter-receiver pair communicating over an unreliable channel with time-varying reliability levels. The transmitter observes the instantaneous link reliability through a channel state information acquisition procedure, during which the data transmission is interrupted. This leads to a fundamental trade-off between utilizing limited network resources for either data transmission or channel state information acquisition to combat the channel aging effect. Assuming the wireless channel is modeled as a finite-state Markovian channel, we formulate an optimization problem as a partially observable Markov decision process (POMDP), obtain the optimal policy through the relative value iteration algorithm, and demonstrate the efficiency of our solution through simulations. To the best of our knowledge, this is the first work to aim for an optimal scheduling policy for data transmissions while considering the effect of channel state information aging.

cs.IT↗