Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How does the robustness of federated large language models against label flipping attacks compare to centralized training when evaluated on reasoning benchmarks like GSM8K. Data poisoning and leakage risks impede…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How do alignment metrics for federated language models degrade under adversarial perturbations compared to centralized models when measured on safety evaluation datasets. Current research in adversarial…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the robustness of federated learning-based malware detectors against adversarial poisoning attacks compare to centralized models in terms of precision degradation. This work investigates the…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: To what extent do minimax-optimal personalized federated learning algorithms improve robustness against adversarial prompts in code generation tasks compared to purely local fine-tuning approaches. Although…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the effect of client update distance weighting on the code generation pass@k scores when federated fine-tuning is applied across heterogeneous programming language corpora. This paper provides an overview…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: What is the impact of adaptive sampling strategies on the inference latency and memory footprint of deployed personalized LLMs compared to random sampling in bandwidth-constrained federated networks. Abstract This…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: Does the stochastic control variate approach in WAFFLE improve inference efficiency and reduce latency variance in personalized multimodal models compared to standard FedAvg under straggler conditions. Federated…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the weighted averaging mechanism in WAFFLE impact the few-shot reasoning accuracy of personalized large language models under non-IID instruction tuning data distributions. This systematic literature…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: Can adaptive model pruning strategies in federated transfer learning improve inference efficiency and detection accuracy for code generation models deployed on resource-constrained devices. Successful integration…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does reducing federated aggregation frequency impact the convergence rate and final accuracy of lightweight neural networks for anomaly detection on edge devices. Federated learning (FL) is a machine learning…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the trade-off between communication overhead and model robustness against adversarial attacks in federated learning systems for IoT security. Federated learning (FL) is revolutionizing healthcare by…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: What is the trade-off between structured pruning ratios and code generation performance in federated learning setups with non-IID data distributions across IoT nodes. The use of artificial intelligence (AI) is…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the comparative robustness of supervised versus unsupervised federated models against adversarial poisoning attacks in cross-device IoT network traffic analysis. Abstract The integration of artificial…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the effect of asynchronous client participation rates on the robustness of federated learning algorithms against poisoning attacks in IoT malware detection scenarios. In this article, we present a…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How do different aggregation strategies in federated learning influence the trade-off between communication overhead and detection accuracy when scaling to heterogeneous IoT device networks. This paper provides…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: What is the impact of different federated learning aggregation techniques on the inference efficiency and latency of multimodal models deployed across heterogeneous edge devices. Abstract The rapid evolution of…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How does varying the degree of non-IID data distribution across IoT nodes impact the convergence rate and final F1-score of federated malware detection models compared to centralized baselines. In this article,…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How does batch size scaling impact the tokens-per-second inference efficiency of domain-adapted Baichuan-2 models on the FactCC hallucination detection benchmark. Abstract The rapid evolution of large language…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: Can the efficiency of cross-domain fine-tuning of Baichuan-2 be improved using gradient checkpointing, and how does this impact FactCC benchmark scores compared to full-precision training. Abstract The rapid…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: How does the fine-tuning of Baichuan-2 on in-domain legal datasets compare to biomedical datasets in terms of TruthfulQA alignment scores and reasoning accuracy on the HellaSwag benchmark. Large language models…

Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: To what extent do small language models evaluated in SLM-Bench maintain reasoning accuracy when subjected to adversarial prompt perturbations compared to larger LLM baselines. Large language models (LLMs) have…

Abstract: This report synthesises findings from 12 peer-reviewed papers addressing the following research question: How does inference-time scaling affect the cross-lingual consistency of factual knowledge in multilingual PLMs when evaluated using the RankC metric. Multilingual large-scale Pretrained Language Models (PLMs)…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: What is the trade-off between inference latency and cross-modality reasoning performance when applying weight-only quantization to LLaVA on the TextVQA dataset. Vision systems to see and reason about the…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: How does the inference throughput of small language models on SLM-Bench tasks vary across different quantization levels and hardware accelerators. Edge computing enables real-time data processing closer to its…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: Does the adoption of directional preference alignment improve robustness against diverse user preference shifts in code generation benchmarks without degrading model efficiency. Methods for detecting nucleotide…