Large language models (LLMs) have revolutionized scientific research with their exceptional capabilities and transformed various fields. Among their practical applications, LLMs have been playing a crucial role in mitigating threats to human life, infrastructure, and the environment. Despite growing research in disaster LLMs, there remains a lack of systematic review and in-depth analysis of LLMs for natural disaster management. To address the gap, this paper presents a comprehensive survey of existing LLMs in natural disaster management, along with a taxonomy that categorizes existing works based on disaster phases and application scenarios. By collecting public datasets and identifying key challenges and opportunities, this study aims to guide the professional community in developing advanced LLMs for disaster management to enhance the resilience against natural disasters.

Today a four-year-old child who does not know how to read or write can now create bedtime stories with graphical illustrations and narrated audio, using AI tools that seamlessly transform speech into text, generate visuals, and convert text back into speech in a natural and engaging manner. This remarkable example demonstrates why we are living in the age of AI applications. This paper examines contemporary leading AI applications and traces their historical development, highlighting the major advancements that have enabled their realization. Five key factors are identified: 1) The evolution of computational hardware (CPUs and GPUs), enabling the training of complex AI models 2) The vast digital archives provided by the World Wide Web, which serve as a foundational data resource for AI systems 3) The ubiquity of mobile computing, with smartphones acting as powerful, accessible small computers in the hands of billions 4) The rise of industrial-scale cloud infrastructures, offering elastic computational power for AI training and deployment 5) Breakthroughs in AI research, including neural networks, backpropagation, and the "Attention is All You Need" framework, which underpin modern AI capabilities. These innovations have elevated AI from solving narrow tasks to enabling applications like ChatGPT that are adaptable for numerous use cases, redefining human-computer interaction. By situating these developments within a historical context, the paper highlights the critical milestones that have made AI's current capabilities both possible and widely accessible, offering profound implications for society.

This paper introduces Shake-VLA, a Vision-Language-Action (VLA) model-based system designed to enable bimanual robotic manipulation for automated cocktail preparation. The system integrates a vision module for detecting ingredient bottles and reading labels, a speech-to-text module for interpreting user commands, and a language model to generate task-specific robotic instructions. Force Torque (FT) sensors are employed to precisely measure the quantity of liquid poured, ensuring accuracy in ingredient proportions during the mixing process. The system architecture includes a Retrieval-Augmented Generation (RAG) module for accessing and adapting recipes, an anomaly detection mechanism to address ingredient availability issues, and bimanual robotic arms for dexterous manipulation. Experimental evaluations demonstrated a high success rate across system components, with the speech-to-text module achieving a 93% success rate in noisy environments, the vision module attaining a 91% success rate in object and label detection in cluttered environment, the anomaly module successfully identified 95% of discrepancies between detected ingredients and recipe requirements, and the system achieved an overall success rate of 100% in preparing cocktails, from recipe formulation to action generation.

We consider the challenge of mitigating the generation of negative or toxic content by the Large Language Models (LLMs) in response to certain prompts. We propose integrating risk-averse principles into LLM fine-tuning to minimize the occurrence of harmful outputs, particularly rare but significant events. By optimizing the risk measure of Conditional Value at Risk (CVaR), our methodology trains LLMs to exhibit superior performance in avoiding toxic outputs while maintaining effectiveness in generative tasks. Empirical evaluations on sentiment modification and toxicity mitigation tasks demonstrate the efficacy of risk-averse reinforcement learning with human feedback (RLHF) in promoting a safer and more constructive online discourse environment. Trigger Warning: This paper contains prompts and model outputs that can be offensive in nature.

Limited availability of multilingual text corpora for training language models often leads to poor performance on downstream tasks due to undertrained representation spaces for languages other than English. This 'under-representation' has motivated recent cross-lingual transfer methods to leverage the English representation space by e.g. mixing English and 'non-English' tokens at the input level or extending model parameters to accommodate new languages. However, these approaches often come at the cost of increased computational complexity. We propose Fusion forLanguage Representations (FLARE) in adapters, a novel method that enhances representation quality and downstream performance for languages other than English while maintaining parameter efficiency. FLARE integrates source and target language representations within low-rank (LoRA) adapters using lightweight linear transformations, maintaining parameter efficiency while improving transfer performance. A series of experiments across representative cross-lingual natural language understanding tasks, including natural language inference, question-answering and sentiment analysis, demonstrate FLARE's effectiveness. FLARE achieves performance improvements of 4.9% for Llama 3.1 and 2.2% for Gemma~2 compared to standard LoRA fine-tuning on question-answering tasks, as measured by the exact match metric.

We analyze over 44,000 NBER and CEPR working papers from 1980 to 2023 using a custom language model to construct knowledge graphs that map economic concepts and their relationships. We distinguish between general claims and those documented via causal inference methods (e.g., DiD, IV, RDD, RCTs). We document a substantial rise in the share of causal claims-from roughly 4% in 1990 to nearly 28% in 2020-reflecting the growing influence of the "credibility revolution." We find that causal narrative complexity (e.g., the depth of causal chains) strongly predicts both publication in top-5 journals and higher citation counts, whereas non-causal complexity tends to be uncorrelated or negatively associated with these outcomes. Novelty is also pivotal for top-5 publication, but only when grounded in credible causal methods: introducing genuinely new causal edges or paths markedly increases both the likelihood of acceptance at leading outlets and long-run citations, while non-causal novelty exhibits weak or even negative effects. Papers engaging with central, widely recognized concepts tend to attract more citations, highlighting a divergence between factors driving publication success and long-term academic impact. Finally, bridging underexplored concept pairs is rewarded primarily when grounded in causal methods, yet such gap filling exhibits no consistent link with future citations. Overall, our findings suggest that methodological rigor and causal innovation are key drivers of academic recognition, but sustained impact may require balancing novel contributions with conceptual integration into established economic discourse.

Despite the artificial intelligence (AI) revolution, deep learning has yet to achieve much success with tabular data due to heterogeneous feature space and limited sample sizes without viable transfer learning. The new era of generative AI, powered by large language models (LLM), brings unprecedented learning opportunities to diverse data and domains. This paper investigates the effectiveness of an LLM application programming interface (API) and transfer learning of LLM in tabular data classification. LLM APIs respond to input text prompts with tokenized data and instructions, whereas transfer learning finetunes an LLM for a target classification task. This paper proposes an end-to-end finetuning of LLM to demonstrate cross-data transfer learning on ten benchmark data sets when large pre-trained tabular data models do not exist to facilitate transfer learning. The proposed LLM finetuning method outperforms state-of-the-art machine and deep learning methods on tabular data with less than ten features - a standard feature size for tabular data sets. The transfer learning approach uses a fraction of the computational cost of other deep learning or API-based solutions while ensuring competitive or superior classification performance.

Mental health disorders pose a growing public health concern in the Arab world, emphasizing the need for accessible diagnostic and intervention tools. Large language models (LLMs) offer a promising approach, but their application in Arabic contexts faces challenges including limited labeled datasets, linguistic complexity, and translation biases. This study comprehensively evaluates 8 LLMs, including general multi-lingual models, as well as bi-lingual ones, on diverse mental health datasets (such as AraDepSu, Dreaddit, MedMCQA), investigating the impact of prompt design, language configuration (native Arabic vs. translated English, and vice versa), and few-shot prompting on diagnostic performance. We find that prompt engineering significantly influences LLM scores mainly due to reduced instruction following, with our structured prompt outperforming a less structured variant on multi-class datasets, with an average difference of 14.5\%. While language influence on performance was modest, model selection proved crucial: Phi-3.5 MoE excelled in balanced accuracy, particularly for binary classification, while Mistral NeMo showed superior performance in mean absolute error for severity prediction tasks. Few-shot prompting consistently improved performance, with particularly substantial gains observed for GPT-4o Mini on multi-class classification, boosting accuracy by an average factor of 1.58. These findings underscore the importance of prompt optimization, multilingual analysis, and few-shot learning for developing culturally sensitive and effective LLM-based mental health tools for Arabic-speaking populations.

Large Language Models (LLMs) have demonstrated exceptional performance across diverse tasks, yet their training remains highly resource-intensive and susceptible to critical challenges such as training instability. A predominant source of this instability stems from gradient and loss spikes, which disrupt the learning process, often leading to costly interventions like checkpoint recovery and experiment restarts, further amplifying inefficiencies. This paper presents a comprehensive investigation into gradient spikes observed during LLM training, revealing their prevalence across multiple architectures and datasets. Our analysis shows that these spikes can be up to $1000\times$ larger than typical gradients, substantially deteriorating model performance. To address this issue, we propose Spike-Aware Adam with Momentum Reset SPAM, a novel optimizer designed to counteract gradient spikes through momentum reset and spike-aware gradient clipping. Extensive experiments, including both pre-training and fine-tuning, demonstrate that SPAM consistently surpasses Adam and its variants across various tasks, including (1) LLM pre-training from 60M to 1B, (2) 4-bit LLM pre-training,(3) reinforcement learning, and (4) Time Series Forecasting. Additionally, SPAM facilitates memory-efficient training by enabling sparse momentum, where only a subset of momentum terms are maintained and updated. When operating under memory constraints, SPAM outperforms state-of-the-art memory-efficient optimizers such as GaLore and Adam-Mini. Our work underscores the importance of mitigating gradient spikes in LLM training and introduces an effective optimization strategy that enhances both training stability and resource efficiency at scale. Code is available at https://github.com/TianjinYellow/SPAM-Optimizer.git

This paper presents a novel approach to represent enterprise web application structures using Large Language Models (LLMs) to enable intelligent quality engineering at scale. We introduce a hierarchical representation methodology that optimizes the few-shot learning capabilities of LLMs while preserving the complex relationships and interactions within web applications. The approach encompasses five key phases: comprehensive DOM analysis, multi-page synthesis, test suite generation, execution, and result analysis. Our methodology addresses existing challenges around usage of Generative AI techniques in automated software testing by developing a structured format that enables LLMs to understand web application architecture through in-context learning. We evaluated our approach using two distinct web applications: an e-commerce platform (Swag Labs) and a healthcare application (MediBox) which is deployed within Atalgo engineering environment. The results demonstrate success rates of 90\% and 70\%, respectively, in achieving automated testing, with high relevance scores for test cases across multiple evaluation criteria. The findings suggest that our representation approach significantly enhances LLMs' ability to generate contextually relevant test cases and provide better quality assurance overall, while reducing the time and effort required for testing.