Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: How does directional preference alignment affect pass@1 scores on the HumanEval-Java subset compared to HumanEval-C++ and HumanEval-Javascript. As Large Language Models (LLMs) become increasingly integrated into…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: To what extent does partial client participation in federated learning degrade the multimodal alignment performance of vision-language models on standard VQA benchmarks. Federated learning (FL) is a machine…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: What is the impact of data heterogeneity under partial client participation on the code generation capabilities of federated LLMs as measured by HumanEval pass@k scores. Federated learning (FL) is a machine…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does the inference latency and token generation throughput of federatedly trained code models scale with client device diversity relative to centrally trained counterparts of similar size. Recent advancements…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the effect of non-IID data distribution across federated clients on the convergence rate and final accuracy of intrusion detection models trained on the Edge-IIoTset dataset. This paper presents a…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How does the inference latency of quantized large language models scale with the number of concurrent edge devices in a federated learning setup for real-time threat detection. Successful integration of deep…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the impact of random layer aggregation on inference efficiency and attack resilience in resource-constrained federated learning models. Federated Learning (FL) is increasingly applied in sectors like…

Abstract: This report synthesises findings from 9 peer-reviewed papers addressing the following research question: Can defense-free federated learning frameworks maintain competitive accuracy on standard benchmarks compared to Byzantine-robust aggregators under varying poisoning rates. Data poisoning is a type of adversarial…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: How does model scaling affect the robustness of federated learning systems against adversarial poisoning attacks in IoT security applications. Due to its distributed nature, federated learning is vulnerable to…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How do different aggregation rules in federated learning affect the inference efficiency and detection latency of deep learning-based intrusion detection systems on edge devices. Intrusion detection systems are…

Abstract: This report synthesises findings from 7 peer-reviewed papers addressing the following research question: What is the impact of non-IID data distribution on the robustness of federated learning models against label-flipping attacks in IoT cybersecurity applications. We present a federated learning approach for…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: How does the success rate of soft prompt attacks on safety alignment compare between quantized and full-precision open-source LLMs across standard harmlessness benchmarks. Current research in adversarial…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: To what extent does differential privacy noise in federated settings degrade the alignment performance of LLMs compared to centralized training on standard safety evaluation datasets. Federated Learning (FL)…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: Can adaptive federated aggregation strategies improve inference efficiency and robustness in distributed code generation models. Medical AI faces challenges in privacy-preserving collaborative learning while…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How does the robustness of federated learning approaches for multimodal sensing-communication alignment compare to centralized training under non-IID data distributions in 6G scenarios. Federated learning is a…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the inference efficiency and throughput comparison of code-generated anomaly detection pipelines versus pre-trained multimodal models for high-dimensional ISAC data streams. Large language models (LLMs)…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How do multimodal large language models perform in reasoning tasks when evaluating the alignment between wireless communication signals and sensing data in 6G integrated networks. Reasoning lies at the heart of…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the correlation between pruning ratios and reasoning accuracy in multimodal models trained via federated learning with skewed data distributions. Federated learning (FL) allows collaborative machine…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: Can adaptive aggregation rules in federated learning improve the resilience of unsupervised anomaly detection models against gradient poisoning in resource-constrained IoT environments. This work investigates the…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: How does structured pruning of code generation LLMs impact pass@1 scores on HumanEval under non-IID federated learning conditions. We introduce self-invoking code generation, a new task designed to evaluate the…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: What is the impact of non-IID data distribution on the convergence speed and detection accuracy of federated intrusion detection systems under label-flipping attacks. Cyber intrusion attacks that compromise the…

Abstract: This report synthesises findings from 5 peer-reviewed papers addressing the following research question: How do different aggregation strategies in federated learning influence the robustness and final performance metrics of malware detection models under highly skewed non-IID conditions. This work investigates the…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the impact of varying client availability schedules on the robustness of federated code generation models against backdoor injections. The delicate equilibrium between user privacy and the ability to…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the effect of asynchronous federated aggregation protocols on the inference latency and task accuracy of multimodal models running on resource-constrained IoT networks. This paper proposes a neural…

Abstract: This report synthesises findings from 6 peer-reviewed papers addressing the following research question: What is the effect of varying client heterogeneity on the inference efficiency and accuracy of federated deep learning systems in IoT cybersecurity applications. Federated Learning (FL) enables decentralized…