Islamic University of Technology

Efficient debugging is essential for software maintenance, as identifying, reproducing, and correcting bugs are critical to ensuring software reliability. However, debugging remains a time-consuming and cost-sensitive activity due to the complexities involved in analyzing and resolving bugs throughout the software development lifecycle. Recent advancements in Artificial Intelligence, Machine Learning, and particularly Natural Language Processing (NLP), and Large Language Models (LLM) offer promising opportunities to enhance debugging process. Since bug reports are typically written in natural language, NLP techniques can streamline debugging tasks, such as bug categorization, localization, and resolution. To assess the potential of NLP in debugging, we conduct a Systematic Literature Review (SLR) of 87 research papers published between 2011 and 2025. Our study presents a comprehensive taxonomy of research efforts, evaluates the effectiveness and limitations of various NLP methods, and highlights best practices across key software debugging tasks, including bug reproduction, bug localization, and bug report analysis. Furthermore, we identify significant challenges in real-world applications, including issues with scalability, accuracy, and adaptability to diverse software environments. Finally, we derive seven findings from our result analysis and propose future research directions corresponding to the findings, to advance NLP-driven debugging practices. This study provides a holistic overview of NLP in debugging for researchers, practitioners, and tool developers, helping to uncover trends, address gaps, and inspire new approaches to improving software maintenance workflows.

Robust visual recognition remains a significant challenge in underwater environments due to complex distortions such as turbidity, low illumination, and occlusion, which severely degrade the performance of standard vision systems. This study introduces AQUA20, a dataset comprising 8,171 images of 20 ecologically significant marine species. AQUA20 provides greater multi-species diversity and real-world variability than other public datasets, such as Fish4Knowledge and WildFish, which are limited in terms of taxonomic scope or environmental complexity. We benchmarked this dataset by evaluating models ranging from lightweight CNNs suitable for edge deployment (e.g., SqueezeNet and MobileNetV2) to modern vision transformers (e.g., ViT and ConvNeXt) for marine species classification under challenging conditions. ConvNeXt achieved the best performance, with a Top-3 accuracy of 98.82 https://huggingface.co/datasets/taufiktrf/AQUA20 .

Accurate short-term photovoltaic (PV) power forecasting is essential for grid stability and efficient PV-grid coordination. However, many conventional learning pipelines remain vulnerable to pervasive missing data, irregular sampling, and the absence of calibrated uncertainty estimates. This paper proposes a physics-informed hybrid forecasting framework that couples Extreme Gradient Boosting (XGBoost) for feature-level learning with a Long Short-Term Memory (LSTM) network for residual correction and temporal dependency modeling. To improve robustness under real-world data conditions, the pipeline incorporates irradiance- guided resampling and domain-guided imputation based on PV operational status. Predictive reliability is further enhanced via Monte Carlo ensemble calibration and conformal prediction, enabling probabilistic forecasts and prediction intervals that are assessed using standard calibration metrics (e.g., Prediction Interval Coverage Probability, Continuous Ranked Probability Score). Experiments on the large-scale UNISOLAR dataset (over 2.7 million samples at 15-minute resolution from 42 PV sites worldwide) show that the proposed hybrid model achieves an RMSE of 2.57 kW h and an R2 of 0.934 on held-out test data, corresponding to a 7.05 % reduction in RMSE relative to the next-best baseline (TCN) and a 17.84 % improvement over a standalone LSTM. An ablation study confirms the critical role of historical lag features, whose removal increases RMSE by over 280 %. The framework also provides well-calibrated uncertainty estimates, with conformal prediction achieving 93.0 % coverage at the 95 % confidence level. Computational profiling confirms the hybrid model’s efficiency, requiring 2529 s training time and 3396 MB memory on a free-tier Google Colab CPU, making it suitable for real-time deployment.

Large language models (LLMs) remain unreliable for high-stakes claim verification due to hallucinations and shallow reasoning. While retrieval-augmented generation (RAG) and multi-agent debate (MAD) address this, they are limited by one-pass retrieval and unstructured debate dynamics. We propose a courtroom-style multi-agent framework, PROClaim, that reformulates verification as a structured, adversarial deliberation. Our approach integrates specialized roles (e.g., Plaintiff, Defense, Judge) with Progressive RAG (P-RAG) to dynamically expand and refine the evidence pool during the debate. Furthermore, we employ evidence negotiation, self-reflection, and heterogeneous multi-judge aggregation to enforce calibration, robustness, and diversity. In zero-shot evaluations on the Check-COVID benchmark, PROClaim achieves 81.7

Multi-agent systems achieve state-of-the-art outcomes through peer collaboration. However, when an agent in the pipeline silently drops a constraint, the system's final output may look correct even though the reasoning chain was quietly corrupted, and existing outcome-based evaluations are blind to such multi-hop process failures. To make these vulnerabilities measurable before deployment, we introduce AgentCollabBench, a diagnostic benchmark of 900 human-validated tasks spanning software engineering, DevOps, and data engineering. Each task isolates one of four behavioral risks: instruction decay (does a constraint survive peer pressure?), false-belief contagion (does a falsehood spread through consensus?), context leakage (does information bleed between tasks?), and tracer durability (does marked data reach the final agent?). Evaluating four modern LLMs (GPT 4.1 mini, Gemini 2.5 Flash Lite, Qwen-3.5-35B-A3B, and Llama 3.1 8B Instruct), we expose model-specific vulnerability profiles invisible to outcome-only evaluation; Qwen-3.5-35B-A3B, for example, leads on tracer durability and instruction stability, while GPT 4.1 mini leads on leakage containment and false-belief resistance. Beyond per-model differences, communication topology emerges as a primary risk factor that explains 7-40% of the variance in multi-hop information survival. The effect traces to a synthesis bottleneck specific to converging-DAG nodes: an agent weighing competing parent inputs discards constraints carried by a minority branch, a bottleneck structurally absent from linear chains. AgentCollabBench demonstrates that suboptimal topology can silently erase the safeguards of highly capable models, arguing that multi-agent reliability is fundamentally a structural problem and that scaling model intelligence alone is no substitute for architecture.