Federated fine-tuning of pre-trained Large Language Models (LLMs) enables task-specific adaptation across diverse datasets while preserving data privacy. However, the large model size and heterogeneity in client resources pose significant computational and communication challenges. To address these issues, in this paper, we propose a novel Heterogeneous Adaptive Federated Low-Rank Adaptation (LoRA) fine-tuned LLM framework (HAFL). To accommodate client resource heterogeneity, we first introduce an importance-based parameter truncation scheme, which allows clients to have different LoRA ranks, and smoothed sensitivity scores are used as importance indicators. Despite its flexibility, the truncation process may cause performance degradation. To tackle this problem, we develop an importance-based parameter freezing scheme. In this approach, both the cloud server and clients maintain the same LoRA rank, while clients selectively update only the most important decomposed LoRA rank-1 matrices, keeping the rest frozen. To mitigate the information dilution caused by the zero-padding aggregation method, we propose an adaptive aggregation approach that operates at the decomposed rank-1 matrix level. Experiments on the 20 News Group classification task show that our method converges quickly with low communication size, and avoids performance degradation when distributing models to clients compared to truncation-based heterogeneous LoRA rank scheme. Additionally, our adaptive aggregation method achieves faster convergence compared to the zero-padding approach.

Droplet-based microfluidic devices have substantial promise as cost-effective alternatives to current assessment tools in biological research. Moreover, machine learning models that leverage tabular data, including input design parameters and their corresponding efficiency outputs, are increasingly utilised to automate the design process of these devices and to predict their performance. However, these models fail to fully leverage the data presented in the tables, neglecting crucial contextual information, including column headings and their associated descriptions. This study presents MicroFluidic-LLMs, a framework designed for processing and feature extraction, which effectively captures contextual information from tabular data formats. MicroFluidic-LLMs overcomes processing challenges by transforming the content into a linguistic format and leveraging pre-trained large language models (LLMs) for analysis. We evaluate our MicroFluidic-LLMs framework on 11 prediction tasks, covering aspects such as geometry, flow conditions, regimes, and performance, utilising a publicly available dataset on flow-focusing droplet microfluidics. We demonstrate that our MicroFluidic-LLMs framework can empower deep neural network models to be highly effective and straightforward while minimising the need for extensive data preprocessing. Moreover, the exceptional performance of deep neural network models, particularly when combined with advanced natural language processing models such as DistilBERT and GPT-2, reduces the mean absolute error in the droplet diameter and generation rate by nearly 5- and 7-fold, respectively, and enhances the regime classification accuracy by over 4%, compared with the performance reported in a previous study. This study lays the foundation for the huge potential applications of LLMs and machine learning in a wider spectrum of microfluidic applications.

The drive toward automating cellular network operations has grown with the increasing complexity of these systems. Despite advancements, full autonomy currently remains out of reach due to reliance on human intervention for modeling network behaviors and defining policies to meet target requirements. Network Digital Twins (NDTs) have shown promise in enhancing network intelligence, but the successful implementation of this technology is constrained by use case-specific architectures, limiting its role in advancing network autonomy. A more capable network intelligence, or "telecommunications brain", is needed to enable seamless, autonomous management of cellular network. Large Language Models (LLMs) have emerged as potential enablers for this vision but face challenges in network modeling, especially in reasoning and handling diverse data types. To address these gaps, we introduce Hermes, a chain of LLM agents that uses "blueprints" for constructing NDT instances through structured and explainable logical steps. Hermes allows automatic, reliable, and accurate network modeling of diverse use cases and configurations, thus marking progress toward fully autonomous network operations.

As the integration of the Large Language Models (LLMs) into various applications increases, so does their susceptibility to misuse, raising significant security concerns. Numerous jailbreak attacks have been proposed to assess the security defense of LLMs. Current jailbreak attacks mainly rely on scenario camouflage, prompt obfuscation, prompt optimization, and prompt iterative optimization to conceal malicious prompts. In particular, sequential prompt chains in a single query can lead LLMs to focus on certain prompts while ignoring others, facilitating context manipulation. This paper introduces SequentialBreak, a novel jailbreak attack that exploits this vulnerability. We discuss several scenarios, not limited to examples like Question Bank, Dialog Completion, and Game Environment, where the harmful prompt is embedded within benign ones that can fool LLMs into generating harmful responses. The distinct narrative structures of these scenarios show that SequentialBreak is flexible enough to adapt to various prompt formats beyond those discussed. Extensive experiments demonstrate that SequentialBreak uses only a single query to achieve a substantial gain of attack success rate over existing baselines against both open-source and closed-source models. Through our research, we highlight the urgent need for more robust and resilient safeguards to enhance LLM security and prevent potential misuse. All the result files and website associated with this research are available in this GitHub repository: https://anonymous.4open.science/r/JailBreakAttack-4F3B/.

We explore the impact of pre-training data composition on the performance of small language models in a sample-efficient setting. Using datasets limited to 10 million words, we evaluate several dataset sources, including child-directed speech (CHILDES), classic books (Gutenberg), synthetic data (TinyStories), and a mix of these (Mix) across different model sizes ranging from 18 million to 705 million parameters. Our experiments show that smaller models (e.g., GPT2-97M, GPT2-705M, Llama-360M) perform better when trained on more complex and rich datasets like Gutenberg. Models trained on the CHILDES and TinyStories datasets underperformed across all model sizes. These findings suggest that the optimal dataset for sample efficient training depends on the model size, and that neither child-directed speech nor simplified stories are optimal for language models of all sizes. We highlight the importance of considering both dataset composition and model capacity for effective sample efficient language model training.

With the rapid advancement of large language models (LLMs), aligning policy models with human preferences has become increasingly critical. Direct Preference Optimization (DPO) has emerged as a promising approach for alignment, acting as an RL-free alternative to Reinforcement Learning from Human Feedback (RLHF). Despite DPO's various advancements and inherent limitations, an in-depth review of these aspects is currently lacking in the literature. In this work, we present a comprehensive review of the challenges and opportunities in DPO, covering theoretical analyses, variants, relevant preference datasets, and applications. Specifically, we categorize recent studies on DPO based on key research questions to provide a thorough understanding of DPO's current landscape. Additionally, we propose several future research directions to offer insights on model alignment for the research community.

This study compares Sporo Health's AI Scribe, a proprietary model fine-tuned for medical scribing, with various LLMs (GPT-4o, GPT-3.5, Gemma-9B, and Llama-3.2-3B) in clinical documentation. We analyzed de-identified patient transcripts from partner clinics, using clinician-provided SOAP notes as the ground truth. Each model generated SOAP summaries using zero-shot prompting, with performance assessed via recall, precision, and F1 scores. Sporo outperformed all models, achieving the highest recall (73.3%), precision (78.6%), and F1 score (75.3%) with the lowest performance variance. Statistically significant differences (p < 0.05) were found between Sporo and the other models, with post-hoc tests showing significant improvements over GPT-3.5, Gemma-9B, and Llama 3.2-3B. While Sporo outperformed GPT-4o by up to 10%, the difference was not statistically significant (p = 0.25). Clinical user satisfaction, measured with a modified PDQI-9 inventory, favored Sporo. Evaluations indicated Sporo's outputs were more accurate and relevant. This highlights the potential of Sporo's multi-agentic architecture to improve clinical workflows.

The rapid evolution of artificial intelligence, particularly large language models, presents unprecedented opportunities for materials science research. We proposed and developed an AI materials scientist named MatPilot, which has shown encouraging abilities in the discovery of new materials. The core strength of MatPilot is its natural language interactive human-machine collaboration, which augments the research capabilities of human scientist teams through a multi-agent system. MatPilot integrates unique cognitive abilities, extensive accumulated experience, and ongoing curiosity of human-beings with the AI agents' capabilities of advanced abstraction, complex knowledge storage and high-dimensional information processing. It could generate scientific hypotheses and experimental schemes, and employ predictive models and optimization algorithms to drive an automated experimental platform for experiments. It turns out that our system demonstrates capabilities for efficient validation, continuous learning, and iterative optimization.

Reasoning is a central capability of human intelligence. In recent years, with the advent of large-scale datasets, pretrained large language models have emerged with new capabilities, including reasoning. However, these models still struggle with long-term, complex reasoning tasks, such as playing chess. Based on the observation that expert chess players employ a dual approach combining long-term strategic play with short-term tactical play along with language explanation, we propose improving the reasoning capability of large language models in chess by integrating annotated strategy and tactic. Specifically, we collect a dataset named MATE, which consists of 1 million chess positions with candidate moves annotated by chess experts for strategy and tactics. We finetune the LLaMA-3-8B model and compare it against state-of-the-art commercial language models in the task of selecting better chess moves. Our experiments show that our models perform better than GPT, Claude, and Gemini models. We find that language explanations can enhance the reasoning capability of large language models.

In large language model (LLM) training, several parallelization strategies, including Tensor Parallelism (TP), Pipeline Parallelism (PP), Data Parallelism (DP), as well as Sequence Parallelism (SP) and Context Parallelism (CP), are employed to distribute model parameters, activations, and optimizer states across devices. Identifying the optimal parallelization configuration for each environment while avoiding GPU memory overflow remains a challenging task. In this study, we provide precise formulas to estimate the memory consumed by parameters, gradients, optimizer states, and activations for 4D parallel training (DP, TP, PP, CP) in the Llama architecture. We conducted 454 experiments on A100 and H100 GPUs, incorporating often neglected factors such as temporary buffers and memory fragmentation into our analysis. Results indicate that when the estimated memory usage is below 80\% of the available GPU memory, the training never encounters out-of-memory errors. This simple yet effective formula allows us to identify parallelization configurations that could lead to memory overflow in advance, significantly reducing the configuration search space. Additionally, through a comprehensive exploration of optimal configurations in 4D parallelism, our analysis of the 454 experimental results provides empirical insights into optimal 4D parallelism configurations.