Abstract: This report synthesises findings from 13 peer-reviewed papers addressing the following research question: How does the few-shot accuracy of Flamingo compare to domain-adapted models on multimodal code vulnerability detection benchmarks when subjected to adversarial perturbations in natural language. In this report,…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: To what extent do Visual Language Models like Flamingo generalize to unseen CWE categories in zero-shot settings compared to fine-tuned code-specific multimodal models. A convolutional neural network (CNN) is one…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: What is the comparative robustness of Llama3 versus GRU models against varying levels of injected noise in multivariate time series forecasting tasks measured by RMSE. Gradient Boosted Decision Trees (GBDT's) are…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: How do non-negative activation constraints in multimodal evidential transformers impact inference throughput compared to standard ReLU baselines on image-text retrieval benchmarks. Abstract Plant diseases cause…

Abstract: This report synthesises findings from 9 peer-reviewed papers addressing the following research question: How does Multi-Objective Reinforcement Learning impact pass@k scores on HumanEval-Java and HumanEval-JavaScript compared to single-objective PPO under varying user preference distributions. Abstract The rapid…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: What is the comparative impact of Q-shaping with learned heuristics versus standard RLHF on the pass@1 accuracy and latency of LLMs across Python, Java, and C++ subsets of the HumanEval benchmark. Large Language…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: Can Directional Preference Alignment maintain higher inference stability than RLHF under edge-case input perturbations in Python code generation evaluations. Fine-grained control over large language models (LLMs)…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: What is the impact of prompt ambiguity on the pass@1 accuracy of LLMs aligned via Directional Preference Alignment versus RLHF in standard code benchmarks. We introduce self-invoking code generation, a new task…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: What is the difference in inference latency and token throughput between Directional Preference Alignment and RLHF fine-tuned models on the DS-1000 data science code generation tasks. Fine-grained control over…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How does the pass@k metric for Directional Preference Alignment compare to RLHF on the MBPP benchmark when scaling model parameters from 7B to 70B. Fine-grained control over large language models (LLMs) remains a…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: How does the inference latency of Directional Preference Alignment compare to standard RLHF pipelines when generating Python solutions on the HumanEval benchmark. This paper studies the alignment process of…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: Does Directional Preference Alignment reduce the computational overhead per token during code synthesis compared to traditional reward modeling approaches in multi-language scenarios. Artificial intelligence (AI)…

Abstract: This report synthesises findings from 10 peer-reviewed papers addressing the following research question: What is the impact of Directional Preference Alignment on cross-lingual code generation accuracy across Java, C++, and JavaScript subsets of the HumanEval dataset. Abstract The rapid evolution of large language…

Abstract: This report synthesises findings from 4 peer-reviewed papers addressing the following research question: What is the correlation between client heterogeneity in federated learning setups and the convergence speed of intrusion detection models trained on the Edge-IIoTset dataset. Federated Learning (FL) is a learning…

Abstract: This report synthesises findings from 16 peer-reviewed papers addressing the following research question: What is the impact of non-IID data distributions across different IoT device types on the robustness and false positive rates of federated intrusion detection models compared to centralized training. In this…

Abstract: This report synthesises findings from 11 peer-reviewed papers addressing the following research question: How does the communication overhead in federated learning scale with the number of participating IoT devices when deploying real-time malware detection models on edge infrastructure. With the rapid development of…

Abstract: This report synthesises findings from 8 peer-reviewed papers addressing the following research question: How does communication efficiency in federated learning for code generation models scale with model size and client heterogeneity relative to centralized distributed training approaches. Federated learning (FL) is…

Abstract: This report synthesises findings from 9 peer-reviewed papers addressing the following research question: How does increasing the number of federated clients impact the inference latency and throughput of lightweight malware detection models on edge devices when evaluated on the N-BaIoT dataset. In this paper, we…

Abstract: This report synthesises findings from 1 peer-reviewed paper addressing the following research question: What is the impact of model scaling on the robustness of federated learning systems against adversarial poisoning attacks in IoT security applications. Federated learning (FL) is a privacypreserving method for…

Abstract: This report synthesises findings from 7 peer-reviewed papers addressing the following research question: How does the participation rate of malicious clients influence the convergence speed and final classification accuracy of personalized federated intrusion detection systems under coordinated. In this article, we…

Abstract: This report synthesises findings from 3 peer-reviewed papers addressing the following research question: What is the trade-off between inference latency and model robustness against adversarial attacks when applying differential privacy mechanisms to federated fine-tuning of code generation models. Federated learning…

Abstract: This report synthesises findings from 15 peer-reviewed papers addressing the following research question: What is the trade-off between communication overhead and F1-score stability when applying differential privacy to federated learning rounds on IoT security datasets. This paper provides a comprehensive study of…

Abstract: This report synthesises findings from 16 peer-reviewed papers addressing the following research question: Can adaptive federated averaging strategies reduce the number of communication rounds required to achieve peak inference efficiency in multimodal IoT intrusion detection systems. Communication overhead in…

Abstract: This report synthesises findings from 16 peer-reviewed papers addressing the following research question: How does varying the degree of data heterogeneity (non-IID) across clients impact the robustness of personalized federated learning models against label-flipping poisoning attacks in intrusion. Federated learning…

Abstract: This report synthesises findings from 14 peer-reviewed papers addressing the following research question: What is the comparative effectiveness of adaptive aggregation strategies versus static weighting in mitigating Byzantine failures within personalized federated learning frameworks for network anomaly. Given the…