A two-part affine approximation has been found to be a good approximation for transformer computations over certain subject object relations. Adapting the Bigger Analogy Test Set, we show that the linear transformation Ws, where s is a middle layer representation of a subject token and W is derived from model derivatives, is also able to accurately reproduce final object states for many relations. This linear technique is able to achieve 90% faithfulness on morphological relations, and we show similar findings multi-lingually and across models. Our findings indicate that some conceptual relationships in language models, such as morphology, are readily interpretable from latent space, and are sparsely encoded by cross-layer linear transformations.

The development of satellite-augmented low-altitude economy and terrestrial networks (SLAETNs) demands intelligent and autonomous systems that can operate reliably across heterogeneous, dynamic, and mission-critical environments. To address these challenges, this survey focuses on enabling agentic artificial intelligence (AI), that is, artificial agents capable of perceiving, reasoning, and acting, through generative AI (GAI) and large language models (LLMs). We begin by introducing the architecture and characteristics of SLAETNs, and analyzing the challenges that arise in integrating satellite, aerial, and terrestrial components. Then, we present a model-driven foundation by systematically reviewing five major categories of generative models: variational autoencoders (VAEs), generative adversarial networks (GANs), generative diffusion models (GDMs), transformer-based models (TBMs), and LLMs. Moreover, we provide a comparative analysis to highlight their generative mechanisms, capabilities, and deployment trade-offs within SLAETNs. Building on this foundation, we examine how these models empower agentic functions across three domains: communication enhancement, security and privacy protection, and intelligent satellite tasks. Finally, we outline key future directions for building scalable, adaptive, and trustworthy generative agents in SLAETNs. This survey aims to provide a unified understanding and actionable reference for advancing agentic AI in next-generation integrated networks.

Large Language Models (LLMs) face significant challenges in specialized domains like law, where precision and domain-specific knowledge are critical. This paper presents a streamlined two-stage framework consisting of Retrieval and Re-ranking to enhance legal document retrieval efficiency and accuracy. Our approach employs a fine-tuned Bi-Encoder for rapid candidate retrieval, followed by a Cross-Encoder for precise re-ranking, both optimized through strategic negative example mining. Key innovations include the introduction of the Exist@m metric to evaluate retrieval effectiveness and the use of semi-hard negatives to mitigate training bias, which significantly improved re-ranking performance. Evaluated on the SoICT Hackathon 2024 for Legal Document Retrieval, our team, 4Huiter, achieved a top-three position. While top-performing teams employed ensemble models and iterative self-training on large bge-m3 architectures, our lightweight, single-pass approach offered a competitive alternative with far fewer parameters. The framework demonstrates that optimized data processing, tailored loss functions, and balanced negative sampling are pivotal for building robust retrieval-augmented systems in legal contexts.

Large Language Models (LLMs) hold promise for improving healthcare access in low-resource settings, but their effectiveness in African primary care contexts remains under-explored. We present a rigorous methodology for creating a benchmark dataset and evaluation framework focused on Kenyan Level 2-3 (dispensary and health center) clinical care. Our approach leverages retrieval-augmented generation (RAG) to ground questions and answers in Kenya's national clinical guidelines, ensuring content aligns with local standard-of-care. The guidelines were digitised, chunked, and indexed for efficient semantic retrieval. Gemini Flash 2.0 Lite was then prompted with relevant guideline excerpts to generate realistic clinical questions, multiple - choice answers, and reasoning scenarios with source citations in English and Swahili. We engaged Kenyan physicians in a co - creation process to refine the dataset's relevance and fairness, and instituted a blinded expert validation pipeline to review for clinical accuracy, clarity, and cultural appropriateness. The resulting Alama Health QA dataset comprises thousands of regulator-aligned question-answer pairs spanning common outpatient conditions in English and Swahili. Beyond standard accuracy metrics, we propose innovative evaluation measures targeting clinical reasoning, safety, and adaptability (e.g. Initial results highlight significant performance gaps in state - of-the - art LLMs when confronted with localized scenarios, echoing recent findings that LLM accuracy on African medical questions lags behind performance on U.S. benchmarks. Our work demonstrates a pathway for dynamic, locally-grounded benchmarks that can evolve with guidelines, providing a crucial tool for safe and effective deployment of AI in African healthcare. Advances in large language models have spurred interest in their potential to augment medical services, especially in low-and middle -income countries facing clinician shortages(Bekbolatova et al., 2024). By handling routine queries or providing decision support, LLMs might help bridge gaps in healthcare access across Africa.

Numerous domain-specific machine learning tasks struggle with data scarcity and class imbalance. This paper systematically explores data augmentation methods for NLP, particularly through large language models like GPT. The purpose of this paper is to examine and evaluate whether traditional methods such as paraphrasing and backtranslation can leverage a new generation of models to achieve comparable performance to purely generative methods. Methods aimed at solving the problem of data scarcity and utilizing ChatGPT were chosen, as well as an exemplary dataset. We conducted a series of experiments comparing four different approaches to data augmentation in multiple experimental setups. We then evaluated the results both in terms of the quality of generated data and its impact on classification performance. The key findings indicate that backtranslation and paraphrasing can yield comparable or even better results than zero and a few-shot generation of examples.

Large Language Models (LLMs) are increasingly applied in the fields of mechanical engineering and materials science. As models that establish connections through the interface of language, LLMs can be applied for step-wise reasoning through the Processing-Structure-Property-Performance chain of material science and engineering. Current LLMs are built for adequately representing a dataset, which is the most part of the accessible internet. However, the internet mostly contains non-scientific content. If LLMs should be applied for engineering purposes, it is valuable to investigate models for their intrinsic knowledge -- here: the capacity to generate correct information about materials. In the current work, for the example of the Periodic Table of Elements, we highlight the role of vocabulary and tokenization for the uniqueness of material fingerprints, and the LLMs' capabilities of generating factually correct output of different state-of-the-art open models. This leads to a material knowledge benchmark for an informed choice, for which steps in the PSPP chain LLMs are applicable, and where specialized models are required.

The use of Bidirectional Encoder Representations from Transformers (BERT) model and its variants for classifying collaborative problem solving (CPS) has been extensively explored within the AI in Education community. However, limited attention has been given to understanding how individual tokenised words in the dataset contribute to the model's classification decisions. Enhancing the explainability of BERT-based CPS diagnostics is essential to better inform end users such as teachers, thereby fostering greater trust and facilitating wider adoption in education. This study undertook a preliminary step towards model transparency and explainability by using SHapley Additive exPlanations (SHAP) to examine how different tokenised words in transcription data contributed to a BERT model's classification of CPS processes. The findings suggested that well-performing classifications did not necessarily equate to a reasonable explanation for the classification decisions. Particular tokenised words were used frequently to affect classifications. The analysis also identified a spurious word, which contributed positively to the classification but was not semantically meaningful to the class. While such model transparency is unlikely to be useful to an end user to improve their practice, it can help them not to overrely on LLM diagnostics and ignore their human expertise. We conclude the workshop paper by noting that the extent to which the model appropriately uses the tokens for its classification is associated with the number of classes involved. It calls for an investigation into the exploration of ensemble model architectures and the involvement of human-AI complementarity for CPS diagnosis, since considerable human reasoning is still required for fine-grained discrimination of CPS subskills.

Ontology evaluation through functional requirements--such as testing via competency question (CQ) verification--is a well-established yet costly, labour-intensive, and error-prone endeavour, even for ontology engineering experts. In this work, we introduce OE-Assist, a novel framework designed to assist ontology evaluation through automated and semi-automated CQ verification. By presenting and leveraging a dataset of 1,393 CQs paired with corresponding ontologies and ontology stories, our contributions present, to our knowledge, the first systematic investigation into large language model (LLM)-assisted ontology evaluation, and include: (i) evaluating the effectiveness of a LLM-based approach for automatically performing CQ verification against a manually created gold standard, and (ii) developing and assessing an LLM-powered framework to assist CQ verification with Protégé, by providing suggestions. We found that automated LLM-based evaluation with o1-preview and o3-mini perform at a similar level to the average user's performance. Through a user study on the framework with 19 knowledge engineers from eight international institutions, we also show that LLMs can assist manual CQ verification and improve user accuracy, especially when suggestions are correct. Additionally, participants reported a marked decrease in perceived task difficulty. However, we also observed a reduction in human performance when the LLM provided incorrect guidance, showing a critical trade-off between efficiency and accuracy in assisted ontology evaluation.

Recent advances in large language models (LLMs) and autonomous agents have enabled systems capable of performing complex tasks across domains such as human-computer interaction, planning, and web navigation. However, many existing frameworks struggle in real-world or resource-constrained environments due to their reliance on cloud-based computation, limited robustness in dynamic contexts, and lack of persistent autonomy and environmental awareness. We present Amico, a modular, event-driven framework for building autonomous agents optimized for embedded systems. Written in Rust for safety and performance, Amico supports reactive, persistent agents that operate efficiently across embedded platforms and browser environments via WebAssembly. It provides clean abstractions for event handling, state management, behavior execution, and integration with reasoning modules. Amico delivers a unified infrastructure for constructing resilient, interactive agents suitable for deployment in settings with limited compute and intermittent connectivity.

In this paper, we formulate the knowledge distillation (KD) as a conditional generative problem and propose the \textit{Generative Distribution Distillation (GenDD)} framework. A naive \textit{GenDD} baseline encounters two major challenges: the curse of high-dimensional optimization and the lack of semantic supervision from labels. To address these issues, we introduce a \textit{Split Tokenization} strategy, achieving stable and effective unsupervised KD. Additionally, we develop the \textit{Distribution Contraction} technique to integrate label supervision into the reconstruction objective. Our theoretical proof demonstrates that \textit{GenDD} with \textit{Distribution Contraction} serves as a gradient-level surrogate for multi-task learning, realizing efficient supervised training without explicit classification loss on multi-step sampling image representations. To evaluate the effectiveness of our method, we conduct experiments on balanced, imbalanced, and unlabeled data. Experimental results show that \textit{GenDD} performs competitively in the unsupervised setting, significantly surpassing KL baseline by \textbf{16.29\%} on ImageNet validation set. With label supervision, our ResNet-50 achieves \textbf{82.28\%} top-1 accuracy on ImageNet in 600 epochs training, establishing a new state-of-the-art.